AuroraMiddleware/gtk
1
0
Fork 1
mirror of https://gitlab.gnome.org/GNOME/gtk.git synced 2024-12-27 14:10:30 +00:00
Code Issues Projects Releases Wiki Activity
79ebd76ac8
gtk/gsk/gl/gskglrenderjob.c

4598 lines
177 KiB
C
Raw Normal View History

Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
/* gskglrenderjob.c
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
*
* Copyright 2017 Timm Bäder <mail@baedert.org>
* Copyright 2018 Matthias Clasen <mclasen@redhat.com>
* Copyright 2018 Alexander Larsson <alexl@redhat.com>
* Copyright 2020 Christian Hergert <chergert@redhat.com>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this program. If not, see <http://www.gnu.org/licenses/>.
*
* SPDX-License-Identifier: LGPL-2.1-or-later
*/
#include "config.h"
#include <gdk/gdkglcontextprivate.h>
#include <gdk/gdkprofilerprivate.h>
#include <gdk/gdkrgbaprivate.h>
#include <gsk/gskrendernodeprivate.h>
#include <gsk/gskglshaderprivate.h>
#include <gdk/gdktextureprivate.h>
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
#include <gdk/gdkmemorytextureprivate.h>
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
#include <gsk/gsktransformprivate.h>
#include <gsk/gskroundedrectprivate.h>
gsk: Add gskrectprivate.h Add a bunch of inline functions for graphene_rectangle_t. We use those quite extensively in tight loops so making them as fast as possible via inlining has massive benefits. The current render-heavy benchmark I am playing (th paris-30k in node-editor) went from 49fps to 85fps on my AMD.
2023-07-20 20:25:20 +00:00
#include <gsk/gskrectprivate.h>
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
#include <math.h>
#include <string.h>
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
#include "gskglcommandqueueprivate.h"
#include "gskgldriverprivate.h"
#include "gskglglyphlibraryprivate.h"
#include "gskgliconlibraryprivate.h"
#include "gskglprogramprivate.h"
#include "gskglrenderjobprivate.h"
#include "gskglshadowlibraryprivate.h"
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
#include "ninesliceprivate.h"
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
#include "fp16private.h"
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
#define ORTHO_NEAR_PLANE -10000
#define ORTHO_FAR_PLANE 10000
#define MAX_GRADIENT_STOPS 6
#define SHADOW_EXTRA_SIZE 4
/* Make sure gradient stops fits in packed array_count */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
G_STATIC_ASSERT ((MAX_GRADIENT_STOPS * 5) < (1 << GSK_GL_UNIFORM_ARRAY_BITS));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
#define ALPHA_IS_CLEAR(alpha) ((alpha) < ((float) 0x00ff / (float) 0xffff))
#define RGBA_IS_CLEAR(rgba) ALPHA_IS_CLEAR((rgba)->alpha)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
typedef struct _GskGLRenderClip
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
GskRoundedRect rect;
guint is_rectilinear : 1;
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
guint is_fully_contained : 1;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
} GskGLRenderClip;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
typedef struct _GskGLRenderModelview
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
GskTransform *transform;
float scale_x;
float scale_y;
gsk: Add a shortcut for a common transform case Add a private getter for the translation part of a transform node, to avoid pulling it out of the transform over and over.
2021-07-24 14:48:20 +00:00
float dx;
float dy;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
float offset_x_before;
float offset_y_before;
graphene_matrix_t matrix;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
} GskGLRenderModelview;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
struct _GskGLRenderJob
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
/* The context containing the framebuffer we are drawing to. Generally this
* is the context of the surface but may be a shared context if rendering to
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
* an offscreen texture such as gsk_gl_renderer_render_texture().
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
*/
GdkGLContext *context;
/* The driver to be used. This is shared among all the renderers on a given
* GdkDisplay and uses the shared GL context to send commands.
*/
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLDriver *driver;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* The command queue (which is just a faster pointer to the driver's
* command queue.
*/
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLCommandQueue *command_queue;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* The region that we are clipping. Normalized to a single rectangle region. */
cairo_region_t *region;
/* The framebuffer to draw to in the @context GL context. So 0 would be the
* default framebuffer of @context. This is important to note as many other
* operations could be done using objects shared from the command queues
* GL context.
*/
guint framebuffer;
gsk/gl: support non-standard default framebuffer There are situations where our "default framebuffer" is not actually zero, yet we still want to apply a scissor rect. Generally, 0 is the default framebuffer. But on platforms where we need to bind a platform-specific feature to a GL_FRAMEBUFFER, we might have a default that is not 0. For example, on macOS we bind an IOSurfaceRef to a GL_TEXTURE_RECTANGLE which then is assigned as the backing store for a framebuffer. This is different than using gsk_gl_renderer_render_texture() in that we don't want to incur an extra copy to the destination surface nor do we even have a way to pass a texture_id into render_texture().
2022-02-11 05:51:15 +00:00
guint default_framebuffer;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* The viewport we are using. This state is updated as we process render
* nodes in the specific visitor callbacks.
*/
graphene_rect_t viewport;
/* The current projection, updated as we process nodes */
graphene_matrix_t projection;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
/* An array of GskGLRenderModelview updated as nodes are processed. The
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
* current modelview is the last element.
*/
GArray *modelview;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
/* An array of GskGLRenderClip updated as nodes are processed. The
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
* current clip is the last element.
*/
GArray *clip;
/* Our current alpha state as we process nodes */
float alpha;
/* Offset (delta x,y) as we process nodes. Occasionally this is merged into
* a transform that is referenced from child transform nodes.
*/
float offset_x;
float offset_y;
/* The scale we are processing, possibly updated by transforms */
float scale_x;
float scale_y;
/* Cached pointers */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
const GskGLRenderClip *current_clip;
const GskGLRenderModelview *current_modelview;
GskGLProgram *current_program;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
guint source_is_glyph_atlas : 1;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* If we should be rendering red zones over fallback nodes */
guint debug_fallback : 1;
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
gsk/gl: avoid clearing opaque regions If the rendering operation is over an opaque region, we can potentially avoid clearing a large section of the framebuffer destination. Some cases you do want to clear, such as when clearing the whole contents as some drivers have fast paths for that to avoid bringing data back into the framebuffer.
2022-02-22 07:27:42 +00:00
/* In some cases we might want to avoid clearing the framebuffer
* because we're going to render over the existing contents.
*/
guint clear_framebuffer : 1;
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
/* Format we want to use for intermediate textures, determined by
* looking at the format of the framebuffer we are rendering on.
*/
int target_format;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
};
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
typedef struct _GskGLRenderOffscreen
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
ngl: Add sonme comments The GskNglRenderOffscreen struct is a mix of in and out arguments. Better annotate that a little bit.
2021-03-31 18:32:47 +00:00
/* The bounds to render */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
const graphene_rect_t *bounds;
ngl: Add sonme comments The GskNglRenderOffscreen struct is a mix of in and out arguments. Better annotate that a little bit.
2021-03-31 18:32:47 +00:00
/* Return location for texture coordinates */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
struct {
float x;
float y;
float x2;
float y2;
} area;
ngl: Add sonme comments The GskNglRenderOffscreen struct is a mix of in and out arguments. Better annotate that a little bit.
2021-03-31 18:32:47 +00:00
/* Return location for texture ID */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
guint texture_id;
gsk: Synchronize when using textures Pass the GLsync object from texture into our command queue, and when executing the queue, wait on the sync object the first time we use its associated texture.
2023-01-28 20:20:26 +00:00
gpointer sync;
ngl: Add sonme comments The GskNglRenderOffscreen struct is a mix of in and out arguments. Better annotate that a little bit.
2021-03-31 18:32:47 +00:00
/* Whether to force creating a new texture, even if the
* input already is a texture
*/
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
guint force_offscreen : 1;
guint reset_clip : 1;
guint do_not_cache : 1;
ngl: Add sonme comments The GskNglRenderOffscreen struct is a mix of in and out arguments. Better annotate that a little bit.
2021-03-31 18:32:47 +00:00
/* Return location for whether we created a texture */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
guint was_offscreen : 1;
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
guint has_mipmap : 1;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
} GskGLRenderOffscreen;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
static void gsk_gl_render_job_visit_node (GskGLRenderJob *job,
const GskRenderNode *node);
static gboolean gsk_gl_render_job_visit_node_with_offscreen (GskGLRenderJob *job,
const GskRenderNode *node,
GskGLRenderOffscreen *offscreen);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
static inline int
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
get_target_format (GskGLRenderJob *job,
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
const GskRenderNode *node)
{
rendernode: Change to gsk_render_node_get_preferred_depth() Instead of just tracking preferred_high_depth(), track the actual depth we'd like to have.
2023-06-15 18:27:13 +00:00
if (gsk_render_node_get_preferred_depth (node) != GDK_MEMORY_U8)
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
return job->target_format;
return GL_RGBA8;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
init_full_texture_region (GskGLRenderOffscreen *offscreen)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
offscreen->area.x = 0;
offscreen->area.y = 0;
offscreen->area.x2 = 1;
offscreen->area.y2 = 1;
}
static inline gboolean G_GNUC_PURE
node_is_invisible (const GskRenderNode *node)
{
return node->bounds.size.width == 0.0f ||
node->bounds.size.height == 0.0f;
}
ngl: Small optimization We really want equality here.
2021-04-11 00:26:45 +00:00
static inline gboolean G_GNUC_PURE
rounded_rect_equal (const GskRoundedRect *r1,
const GskRoundedRect *r2)
{
return memcmp (r1, r2, sizeof (GskRoundedRect)) == 0;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
gsk_rounded_rect_shrink_to_minimum (GskRoundedRect *self)
{
self->bounds.size.width = MAX (self->corner[0].width + self->corner[1].width,
self->corner[3].width + self->corner[2].width);
self->bounds.size.height = MAX (self->corner[0].height + self->corner[3].height,
self->corner[1].height + self->corner[2].height);
}
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
static inline gboolean G_GNUC_PURE
node_supports_2d_transform (const GskRenderNode *node)
{
gsk: Use explicit switches This makes it harder to forget to add new node types here.
2023-04-19 07:19:12 +00:00
switch (gsk_render_node_get_node_type (node))
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
{
case GSK_COLOR_NODE:
case GSK_OPACITY_NODE:
case GSK_COLOR_MATRIX_NODE:
case GSK_TEXTURE_NODE:
gsk: Fix rotated texture scale nodes This was causing us to render rotated texture scale nodes to an unclipped offscreen, bringing 'giant intermediate texture' problems back.
2023-03-13 16:59:04 +00:00
case GSK_TEXTURE_SCALE_NODE:
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
case GSK_CROSS_FADE_NODE:
case GSK_LINEAR_GRADIENT_NODE:
case GSK_REPEATING_LINEAR_GRADIENT_NODE:
case GSK_CONIC_GRADIENT_NODE:
case GSK_RADIAL_GRADIENT_NODE:
case GSK_REPEATING_RADIAL_GRADIENT_NODE:
case GSK_DEBUG_NODE:
case GSK_TEXT_NODE:
case GSK_CAIRO_NODE:
case GSK_BLEND_NODE:
case GSK_BLUR_NODE:
gsk: Mask nodes support 2d transforms When adding mask nodes, I overlooked that we have two separate functions for determining what transforms a node supports without offlines. Since we claim that mask nodes support general transform, they must certainly support 2d transforms as well.
2023-04-19 06:13:55 +00:00
case GSK_MASK_NODE:
gsk: Add GskFillNode Take a rendernode as source and a GskPath and fill the region inside the path with the source, just like cairo_fill() would.
2020-11-10 23:44:32 +00:00
case GSK_FILL_NODE:
gsk: Add GskStrokeNode Take a rendernode as source and a GskPath and GskStroke, and fill the area that is covered when stroking the path with the given stroke parameters, like cairo_stroke() would.
2020-11-11 06:36:04 +00:00
case GSK_STROKE_NODE:
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
return TRUE;
case GSK_SHADOW_NODE:
return node_supports_2d_transform (gsk_shadow_node_get_child (node));
case GSK_TRANSFORM_NODE:
return node_supports_2d_transform (gsk_transform_node_get_child (node));
case GSK_CONTAINER_NODE:
for (guint i = 0, p = gsk_container_node_get_n_children (node); i < p; i++)
{
if (!node_supports_2d_transform (gsk_container_node_get_child (node, i)))
return FALSE;
}
return TRUE;
gsk: Use explicit switches This makes it harder to forget to add new node types here.
2023-04-19 07:19:12 +00:00
case GSK_BORDER_NODE:
case GSK_INSET_SHADOW_NODE:
case GSK_OUTSET_SHADOW_NODE:
case GSK_REPEAT_NODE:
case GSK_CLIP_NODE:
case GSK_ROUNDED_CLIP_NODE:
case GSK_GL_SHADER_NODE:
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
return FALSE;
gsk: Use explicit switches This makes it harder to forget to add new node types here.
2023-04-19 07:19:12 +00:00
case GSK_NOT_A_RENDER_NODE:
default:
g_assert_not_reached ();
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
}
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline gboolean G_GNUC_PURE
node_supports_transform (const GskRenderNode *node)
{
/* Some nodes can't handle non-trivial transforms without being
* rendered to a texture (e.g. rotated clips, etc.). Some however work
* just fine, mostly because they already draw their child to a
* texture and just render the texture manipulated in some way, think
* opacity or color matrix.
*/
gsk: Use explicit switches This makes it harder to forget to add new node types here.
2023-04-19 07:19:12 +00:00
switch (gsk_render_node_get_node_type (node))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
case GSK_COLOR_NODE:
case GSK_OPACITY_NODE:
case GSK_COLOR_MATRIX_NODE:
case GSK_TEXTURE_NODE:
case GSK_CROSS_FADE_NODE:
case GSK_DEBUG_NODE:
case GSK_TEXT_NODE:
case GSK_CAIRO_NODE:
case GSK_BLEND_NODE:
case GSK_BLUR_NODE:
gsk: Mark mask nodes as 'supports transforms' Just like blur nodes, the renderer can handle these without creating an extra offscreen.
2023-03-13 17:10:53 +00:00
case GSK_MASK_NODE:
gsk: Add GskFillNode Take a rendernode as source and a GskPath and fill the region inside the path with the source, just like cairo_fill() would.
2020-11-10 23:44:32 +00:00
case GSK_FILL_NODE:
gsk: Add GskStrokeNode Take a rendernode as source and a GskPath and GskStroke, and fill the area that is covered when stroking the path with the given stroke parameters, like cairo_stroke() would.
2020-11-11 06:36:04 +00:00
case GSK_STROKE_NODE:
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
return TRUE;
case GSK_SHADOW_NODE:
return node_supports_transform (gsk_shadow_node_get_child (node));
case GSK_TRANSFORM_NODE:
return node_supports_transform (gsk_transform_node_get_child (node));
gsk: Use explicit switches This makes it harder to forget to add new node types here.
2023-04-19 07:19:12 +00:00
case GSK_CONTAINER_NODE:
case GSK_LINEAR_GRADIENT_NODE:
case GSK_REPEATING_LINEAR_GRADIENT_NODE:
case GSK_RADIAL_GRADIENT_NODE:
case GSK_REPEATING_RADIAL_GRADIENT_NODE:
case GSK_CONIC_GRADIENT_NODE:
case GSK_BORDER_NODE:
case GSK_INSET_SHADOW_NODE:
case GSK_OUTSET_SHADOW_NODE:
case GSK_REPEAT_NODE:
case GSK_CLIP_NODE:
case GSK_ROUNDED_CLIP_NODE:
case GSK_GL_SHADER_NODE:
case GSK_TEXTURE_SCALE_NODE:
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
return FALSE;
gsk: Use explicit switches This makes it harder to forget to add new node types here.
2023-04-19 07:19:12 +00:00
case GSK_NOT_A_RENDER_NODE:
default:
g_assert_not_reached ();
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
static inline gboolean G_GNUC_PURE
color_matrix_modifies_alpha (const GskRenderNode *node)
{
const graphene_matrix_t *matrix = gsk_color_matrix_node_get_color_matrix (node);
const graphene_vec4_t *offset = gsk_color_matrix_node_get_color_offset (node);
graphene_vec4_t row3;
if (graphene_vec4_get_w (offset) != 0.0f)
return TRUE;
graphene_matrix_get_row (matrix, 3, &row3);
return !graphene_vec4_equal (graphene_vec4_w_axis (), &row3);
}
static inline void
init_projection_matrix (graphene_matrix_t *projection,
const graphene_rect_t *viewport)
{
graphene_matrix_init_ortho (projection,
viewport->origin.x,
viewport->origin.x + viewport->size.width,
viewport->origin.y,
viewport->origin.y + viewport->size.height,
ORTHO_NEAR_PLANE,
ORTHO_FAR_PLANE);
graphene_matrix_scale (projection, 1, -1, 1);
}
static inline float
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_alpha (GskGLRenderJob *job,
float alpha)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
if (job->alpha != alpha)
{
float ret = job->alpha;
job->alpha = alpha;
job->driver->stamps[UNIFORM_SHARED_ALPHA]++;
return ret;
}
return alpha;
}
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
extract_matrix_metadata (GskGLRenderModelview *modelview)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk_transform_to_matrix (modelview->transform, &modelview->matrix);
switch (gsk_transform_get_category (modelview->transform))
{
case GSK_TRANSFORM_CATEGORY_IDENTITY:
gsk: Add a shortcut for a common transform case Add a private getter for the translation part of a transform node, to avoid pulling it out of the transform over and over.
2021-07-24 14:48:20 +00:00
modelview->scale_x = 1;
modelview->scale_y = 1;
modelview->dx = 0;
modelview->dy = 0;
break;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
case GSK_TRANSFORM_CATEGORY_2D_TRANSLATE:
modelview->scale_x = 1;
modelview->scale_y = 1;
gsk: Add a shortcut for a common transform case Add a private getter for the translation part of a transform node, to avoid pulling it out of the transform over and over.
2021-07-24 14:48:20 +00:00
gsk_transform_to_translate (modelview->transform,
&modelview->dx, &modelview->dy);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_TRANSFORM_CATEGORY_2D_AFFINE:
gsk_transform_to_affine (modelview->transform,
&modelview->scale_x, &modelview->scale_y,
gsk: Add a shortcut for a common transform case Add a private getter for the translation part of a transform node, to avoid pulling it out of the transform over and over.
2021-07-24 14:48:20 +00:00
&modelview->dx, &modelview->dy);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
gsk: Handle 2D transforms better For 2D transforms, we can read the scale factors more directly off the matrix. This should eventually be moved out into a function to decompose a 2D transform into scale + rotation + skew + translation.
2021-09-17 13:42:20 +00:00
case GSK_TRANSFORM_CATEGORY_2D:
{
gsk: Extract scales properly from 2D transforms We have the code to do it properly.
2023-05-01 17:28:27 +00:00
float skew_x, skew_y, angle, dx, dy;
gsk_transform_to_2d_components (modelview->transform,
&skew_x, &skew_y,
&modelview->scale_x, &modelview->scale_y,
&angle, &dx, &dy);
modelview->dx = 0;
modelview->dy = 0;
gsk: Handle 2D transforms better For 2D transforms, we can read the scale factors more directly off the matrix. This should eventually be moved out into a function to decompose a 2D transform into scale + rotation + skew + translation.
2021-09-17 13:42:20 +00:00
}
break;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
case GSK_TRANSFORM_CATEGORY_UNKNOWN:
case GSK_TRANSFORM_CATEGORY_ANY:
case GSK_TRANSFORM_CATEGORY_3D:
{
gsk/gl/renderjob: Decompose matrix for 3D transforms While this can be more expensive, it's also correct.
2023-04-01 14:38:22 +00:00
graphene_quaternion_t rotation;
graphene_vec4_t perspective;
graphene_vec3_t translation;
graphene_vec3_t scale;
graphene_vec3_t shear;
graphene_matrix_decompose (&modelview->matrix,
&translation,
&scale,
&rotation,
&shear,
&perspective);
modelview->scale_x = graphene_vec3_get_x (&scale);
modelview->scale_y = graphene_vec3_get_y (&scale);
gsk: Add a shortcut for a common transform case Add a private getter for the translation part of a transform node, to avoid pulling it out of the transform over and over.
2021-07-24 14:48:20 +00:00
modelview->dx = 0;
modelview->dy = 0;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
break;
default:
break;
}
}
gsk: Fix a refcounting mishap This may fix #5823
2023-05-09 10:51:53 +00:00
/* takes ownership of transform */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_modelview (GskGLRenderJob *job,
GskTransform *transform)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderModelview *modelview;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (job != NULL);
g_assert (job->modelview != NULL);
job->driver->stamps[UNIFORM_SHARED_MODELVIEW]++;
g_array_set_size (job->modelview, job->modelview->len + 1);
modelview = &g_array_index (job->modelview,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderModelview,
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->modelview->len - 1);
modelview->transform = transform;
modelview->offset_x_before = job->offset_x;
modelview->offset_y_before = job->offset_y;
extract_matrix_metadata (modelview);
job->offset_x = 0;
job->offset_y = 0;
job->scale_x = modelview->scale_x;
job->scale_y = modelview->scale_y;
job->current_modelview = modelview;
}
gsk: Fix a refcounting mishap This may fix #5823
2023-05-09 10:51:53 +00:00
/* doesn't take ownership of transform */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_modelview (GskGLRenderJob *job,
GskTransform *transform)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderModelview *modelview;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (job != NULL);
g_assert (job->modelview != NULL);
g_assert (transform != NULL);
job->driver->stamps[UNIFORM_SHARED_MODELVIEW]++;
g_array_set_size (job->modelview, job->modelview->len + 1);
modelview = &g_array_index (job->modelview,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderModelview,
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->modelview->len - 1);
if G_LIKELY (job->modelview->len > 1)
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderModelview *last;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
GskTransform *t = NULL;
last = &g_array_index (job->modelview,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderModelview,
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->modelview->len - 2);
Improve transformed offscreen rendering Preserve the scale for 2D transforms to avoid a pixellated appearance.
2021-05-13 21:32:52 +00:00
/* Multiply given matrix with our previous modelview */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
t = gsk_transform_translate (gsk_transform_ref (last->transform),
&(graphene_point_t) {
job->offset_x,
job->offset_y
});
t = gsk_transform_transform (t, transform);
modelview->transform = t;
}
else
{
modelview->transform = gsk_transform_ref (transform);
}
modelview->offset_x_before = job->offset_x;
modelview->offset_y_before = job->offset_y;
extract_matrix_metadata (modelview);
job->offset_x = 0;
job->offset_y = 0;
job->scale_x = modelview->scale_x;
job->scale_y = modelview->scale_y;
job->current_modelview = modelview;
}
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_modelview (GskGLRenderJob *job)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
const GskGLRenderModelview *head;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (job != NULL);
g_assert (job->modelview);
g_assert (job->modelview->len > 0);
job->driver->stamps[UNIFORM_SHARED_MODELVIEW]++;
head = job->current_modelview;
job->offset_x = head->offset_x_before;
job->offset_y = head->offset_y_before;
gsk_transform_unref (head->transform);
job->modelview->len--;
if (job->modelview->len >= 1)
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
head = &g_array_index (job->modelview, GskGLRenderModelview, job->modelview->len - 1);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->scale_x = head->scale_x;
job->scale_y = head->scale_y;
job->current_modelview = head;
}
else
{
job->current_modelview = NULL;
}
}
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (GskGLRenderJob *job,
const GskRoundedRect *rect)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderClip *clip;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (job != NULL);
g_assert (job->clip != NULL);
g_assert (rect != NULL);
job->driver->stamps[UNIFORM_SHARED_CLIP_RECT]++;
g_array_set_size (job->clip, job->clip->len + 1);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
clip = &g_array_index (job->clip, GskGLRenderClip, job->clip->len - 1);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
memcpy (&clip->rect, rect, sizeof *rect);
clip->is_rectilinear = gsk_rounded_rect_is_rectilinear (rect);
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
clip->is_fully_contained = FALSE;
job->current_clip = clip;
}
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_contained_clip (GskGLRenderJob *job)
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderClip *clip;
GskGLRenderClip *old_clip;
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
g_assert (job != NULL);
g_assert (job->clip != NULL);
g_assert (job->clip->len > 0);
job->driver->stamps[UNIFORM_SHARED_CLIP_RECT]++;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
old_clip = &g_array_index (job->clip, GskGLRenderClip, job->clip->len - 1);
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
g_array_set_size (job->clip, job->clip->len + 1);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
clip = &g_array_index (job->clip, GskGLRenderClip, job->clip->len - 1);
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
memcpy (&clip->rect.bounds, &old_clip->rect.bounds, sizeof (graphene_rect_t));
memset (clip->rect.corner, 0, sizeof clip->rect.corner);
clip->is_rectilinear = TRUE;
clip->is_fully_contained = TRUE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->current_clip = clip;
}
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_clip (GskGLRenderJob *job)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
g_assert (job != NULL);
g_assert (job->clip != NULL);
g_assert (job->clip->len > 0);
job->driver->stamps[UNIFORM_SHARED_CLIP_RECT]++;
job->current_clip--;
job->clip->len--;
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_offset (GskGLRenderJob *job,
float offset_x,
float offset_y)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
if (offset_x || offset_y)
{
job->offset_x += offset_x;
job->offset_y += offset_y;
}
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_projection (GskGLRenderJob *job,
const graphene_matrix_t *projection)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
memcpy (&job->projection, projection, sizeof job->projection);
job->driver->stamps[UNIFORM_SHARED_PROJECTION]++;
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_projection_from_rect (GskGLRenderJob *job,
const graphene_rect_t *rect,
graphene_matrix_t *prev_projection)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
if (prev_projection)
memcpy (prev_projection, &job->projection, sizeof *prev_projection);
init_projection_matrix (&job->projection, rect);
job->driver->stamps[UNIFORM_SHARED_PROJECTION]++;
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_projection_for_size (GskGLRenderJob *job,
float width,
float height,
graphene_matrix_t *prev_projection)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
if (prev_projection)
memcpy (prev_projection, &job->projection, sizeof *prev_projection);
graphene_matrix_init_ortho (&job->projection, 0, width, 0, height, ORTHO_NEAR_PLANE, ORTHO_FAR_PLANE);
graphene_matrix_scale (&job->projection, 1, -1, 1);
job->driver->stamps[UNIFORM_SHARED_PROJECTION]++;
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_viewport (GskGLRenderJob *job,
const graphene_rect_t *viewport,
graphene_rect_t *prev_viewport)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
if (prev_viewport)
memcpy (prev_viewport, &job->viewport, sizeof *prev_viewport);
memcpy (&job->viewport, viewport, sizeof job->viewport);
job->driver->stamps[UNIFORM_SHARED_VIEWPORT]++;
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_viewport_for_size (GskGLRenderJob *job,
float width,
float height,
graphene_rect_t *prev_viewport)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
if (prev_viewport)
memcpy (prev_viewport, &job->viewport, sizeof *prev_viewport);
job->viewport.origin.x = 0;
job->viewport.origin.y = 0;
job->viewport.size.width = width;
job->viewport.size.height = height;
job->driver->stamps[UNIFORM_SHARED_VIEWPORT]++;
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_transform_bounds (GskGLRenderJob *job,
const graphene_rect_t *rect,
graphene_rect_t *out_rect)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
GskTransform *transform;
GskTransformCategory category;
g_assert (job != NULL);
g_assert (job->modelview->len > 0);
g_assert (rect != NULL);
g_assert (out_rect != NULL);
transform = job->current_modelview->transform;
category = gsk_transform_get_category (transform);
/* Our most common transform is 2d-affine, so inline it.
* Both identity and 2d-translate are virtually unseen here.
*/
ngl: Fix an oversight We were special-casing 2D affine transforms, but overlooked even simpler transforms.
2021-04-05 03:44:26 +00:00
if G_LIKELY (category >= GSK_TRANSFORM_CATEGORY_2D_AFFINE)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Add a shortcut for a common transform case Add a private getter for the translation part of a transform node, to avoid pulling it out of the transform over and over.
2021-07-24 14:48:20 +00:00
float scale_x = job->current_modelview->scale_x;
float scale_y = job->current_modelview->scale_y;
float dx = job->current_modelview->dx;
float dy = job->current_modelview->dy;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Init directly into out rect */
out_rect->origin.x = ((rect->origin.x + job->offset_x) * scale_x) + dx;
out_rect->origin.y = ((rect->origin.y + job->offset_y) * scale_y) + dy;
out_rect->size.width = rect->size.width * scale_x;
out_rect->size.height = rect->size.height * scale_y;
ngl: Typo fix
2021-03-28 17:27:09 +00:00
/* Normalize in place */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (out_rect->size.width < 0.f)
{
float size = fabsf (out_rect->size.width);
out_rect->origin.x -= size;
out_rect->size.width = size;
}
if (out_rect->size.height < 0.f)
{
float size = fabsf (out_rect->size.height);
out_rect->origin.y -= size;
out_rect->size.height = size;
}
}
else
{
graphene_rect_t r;
r.origin.x = rect->origin.x + job->offset_x;
r.origin.y = rect->origin.y + job->offset_y;
r.size.width = rect->size.width;
r.size.height = rect->size.height;
gsk_transform_transform_bounds (transform, &r, out_rect);
}
}
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
static inline void
gsk_gl_render_job_untransform_bounds (GskGLRenderJob *job,
const graphene_rect_t *rect,
graphene_rect_t *out_rect)
{
GskTransform *transform;
transform = gsk_transform_invert (gsk_transform_ref (job->current_modelview->transform));
gsk_transform_transform_bounds (transform, rect, out_rect);
out_rect->origin.x -= job->offset_x;
out_rect->origin.y -= job->offset_y;
gsk: Plug a memory leak in the gl renderer Found by asan.
2023-06-28 01:52:08 +00:00
gsk_transform_unref (transform);
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_translate_rounded_rect (GskGLRenderJob *job,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
const GskRoundedRect *rect,
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
GskRoundedRect *out_rect)
{
out_rect->bounds.origin.x = job->offset_x + rect->bounds.origin.x;
out_rect->bounds.origin.y = job->offset_y + rect->bounds.origin.y;
out_rect->bounds.size.width = rect->bounds.size.width;
out_rect->bounds.size.height = rect->bounds.size.height;
memcpy (out_rect->corner, rect->corner, sizeof rect->corner);
}
gsk: Transform rounded rects better
2023-04-30 15:19:47 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
static inline void
rounded_rect_scale_corners (const GskRoundedRect *rect,
GskRoundedRect *out_rect,
float scale_x,
float scale_y)
{
for (guint i = 0; i < G_N_ELEMENTS (out_rect->corner); i++)
gsk: Transform rounded rects better
2023-04-30 15:19:47 +00:00
{
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
out_rect->corner[i].width = rect->corner[i].width * fabs (scale_x);
out_rect->corner[i].height = rect->corner[i].height * fabs (scale_y);
gsk: Transform rounded rects better
2023-04-30 15:19:47 +00:00
}
if (scale_x < 0)
{
graphene_size_t p;
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
p = out_rect->corner[GSK_CORNER_TOP_LEFT];
out_rect->corner[GSK_CORNER_TOP_LEFT] = out_rect->corner[GSK_CORNER_TOP_RIGHT];
out_rect->corner[GSK_CORNER_TOP_RIGHT] = p;
p = out_rect->corner[GSK_CORNER_BOTTOM_LEFT];
out_rect->corner[GSK_CORNER_BOTTOM_LEFT] = out_rect->corner[GSK_CORNER_BOTTOM_RIGHT];
out_rect->corner[GSK_CORNER_BOTTOM_RIGHT] = p;
gsk: Transform rounded rects better
2023-04-30 15:19:47 +00:00
}
if (scale_y < 0)
{
graphene_size_t p;
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
p = out_rect->corner[GSK_CORNER_TOP_LEFT];
out_rect->corner[GSK_CORNER_TOP_LEFT] = out_rect->corner[GSK_CORNER_BOTTOM_LEFT];
out_rect->corner[GSK_CORNER_BOTTOM_LEFT] = p;
p = out_rect->corner[GSK_CORNER_TOP_RIGHT];
out_rect->corner[GSK_CORNER_TOP_RIGHT] = out_rect->corner[GSK_CORNER_BOTTOM_RIGHT];
out_rect->corner[GSK_CORNER_BOTTOM_RIGHT] = p;
gsk: Transform rounded rects better
2023-04-30 15:19:47 +00:00
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
static inline void
gsk_gl_render_job_transform_rounded_rect (GskGLRenderJob *job,
const GskRoundedRect *rect,
GskRoundedRect *out_rect)
{
gsk_gl_render_job_transform_bounds (job, &rect->bounds, &out_rect->bounds);
rounded_rect_scale_corners (rect, out_rect, job->scale_x, job->scale_y);
}
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
static inline void
rounded_rect_get_inner (const GskRoundedRect *rect,
graphene_rect_t *inner)
{
float left = MAX (rect->corner[GSK_CORNER_TOP_LEFT].width, rect->corner[GSK_CORNER_BOTTOM_LEFT].width);
float right = MAX (rect->corner[GSK_CORNER_TOP_RIGHT].width, rect->corner[GSK_CORNER_BOTTOM_RIGHT].width);
float top = MAX (rect->corner[GSK_CORNER_TOP_LEFT].height, rect->corner[GSK_CORNER_TOP_RIGHT].height);
float bottom = MAX (rect->corner[GSK_CORNER_BOTTOM_LEFT].height, rect->corner[GSK_CORNER_BOTTOM_RIGHT].height);
inner->origin.x = rect->bounds.origin.x + left;
inner->size.width = rect->bounds.size.width - (left + right);
inner->origin.y = rect->bounds.origin.y + top;
inner->size.height = rect->bounds.size.height - (top + bottom);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline gboolean
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
interval_contains (float p1, float w1,
float p2, float w2)
{
if (p2 < p1)
return FALSE;
if (p2 + w2 > p1 + w1)
return FALSE;
return TRUE;
}
static inline gboolean
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_update_clip (GskGLRenderJob *job,
const graphene_rect_t *bounds,
gboolean *pushed_clip)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
graphene_rect_t transformed_bounds;
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
gboolean no_clip = FALSE;
gboolean rect_clip = FALSE;
*pushed_clip = FALSE;
if (job->current_clip->is_fully_contained)
{
/* Already fully contained - no further checks needed */
return TRUE;
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_transform_bounds (job, bounds, &transformed_bounds);
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
gsk: Add gskrectprivate.h Add a bunch of inline functions for graphene_rectangle_t. We use those quite extensively in tight loops so making them as fast as possible via inlining has massive benefits. The current render-heavy benchmark I am playing (th paris-30k in node-editor) went from 49fps to 85fps on my AMD.
2023-07-20 20:25:20 +00:00
if (!gsk_rect_intersects (&job->current_clip->rect.bounds, &transformed_bounds))
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
{
/* Completely clipped away */
return FALSE;
}
if (job->current_clip->is_rectilinear)
{
gsk: Add gskrectprivate.h Add a bunch of inline functions for graphene_rectangle_t. We use those quite extensively in tight loops so making them as fast as possible via inlining has massive benefits. The current render-heavy benchmark I am playing (th paris-30k in node-editor) went from 49fps to 85fps on my AMD.
2023-07-20 20:25:20 +00:00
if (gsk_rect_contains_rect (&job->current_clip->rect.bounds, &transformed_bounds))
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
no_clip = TRUE;
else
rect_clip = TRUE;
}
else if (gsk_rounded_rect_contains_rect (&job->current_clip->rect, &transformed_bounds))
{
no_clip = TRUE;
}
else
{
graphene_rect_t inner;
rounded_rect_get_inner (&job->current_clip->rect, &inner);
if (interval_contains (inner.origin.x, inner.size.width,
transformed_bounds.origin.x, transformed_bounds.size.width) ||
interval_contains (inner.origin.y, inner.size.height,
transformed_bounds.origin.y, transformed_bounds.size.height))
rect_clip = TRUE;
}
if (no_clip)
{
/* This node is completely contained inside the clip.
* Record this fact on the clip stack, so we don't do more work
* for child nodes.
*/
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_contained_clip (job);
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
*pushed_clip = TRUE;
}
else if (rect_clip && !job->current_clip->is_rectilinear)
{
graphene_rect_t rect;
/* The clip gets simpler for this node */
graphene_rect_intersection (&job->current_clip->rect.bounds, &transformed_bounds, &rect);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (job, &GSK_ROUNDED_RECT_INIT_FROM_RECT (rect));
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
*pushed_clip = TRUE;
}
return TRUE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
static inline void
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
rgba_to_half (const GdkRGBA *rgba,
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
guint16 h[4])
{
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
float_to_half4 ((const float *)rgba, h);
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
}
ngl: Set color as vertex attribute Instead of using uniforms for color used in multiple programs, pass it as vertex attributes. This will let us batch more draw calls, since we don't have to change uniforms so often. In particular, for syntax-highlighted text.
2021-03-12 00:46:15 +00:00
/* fill_vertex_data */
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_coords (GskGLRenderJob *job,
float min_x,
float min_y,
float max_x,
float max_y,
float min_u,
float min_v,
float max_u,
float max_v,
guint16 c[4])
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLDrawVertex *vertices = gsk_gl_command_queue_add_vertices (job->command_queue);
vertices[0] = (GskGLDrawVertex) { .position = { min_x, min_y }, .uv = { min_u, min_v }, .color = { c[0], c[1], c[2], c[3] } };
vertices[1] = (GskGLDrawVertex) { .position = { min_x, max_y }, .uv = { min_u, max_v }, .color = { c[0], c[1], c[2], c[3] } };
vertices[2] = (GskGLDrawVertex) { .position = { max_x, min_y }, .uv = { max_u, min_v }, .color = { c[0], c[1], c[2], c[3] } };
vertices[3] = (GskGLDrawVertex) { .position = { max_x, max_y }, .uv = { max_u, max_v }, .color = { c[0], c[1], c[2], c[3] } };
vertices[4] = (GskGLDrawVertex) { .position = { min_x, max_y }, .uv = { min_u, max_v }, .color = { c[0], c[1], c[2], c[3] } };
vertices[5] = (GskGLDrawVertex) { .position = { max_x, min_y }, .uv = { max_u, min_v }, .color = { c[0], c[1], c[2], c[3] } };
ngl: Set color as vertex attribute Instead of using uniforms for color used in multiple programs, pass it as vertex attributes. This will let us batch more draw calls, since we don't have to change uniforms so often. In particular, for syntax-highlighted text.
2021-03-12 00:46:15 +00:00
}
/* load_vertex_data_with_region */
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_offscreen_with_color (GskGLRenderJob *job,
const graphene_rect_t *bounds,
const GskGLRenderOffscreen *offscreen,
guint16 color[4])
ngl: Set color as vertex attribute Instead of using uniforms for color used in multiple programs, pass it as vertex attributes. This will let us batch more draw calls, since we don't have to change uniforms so often. In particular, for syntax-highlighted text.
2021-03-12 00:46:15 +00:00
{
float min_x = job->offset_x + bounds->origin.x;
float min_y = job->offset_y + bounds->origin.y;
float max_x = min_x + bounds->size.width;
float max_y = min_y + bounds->size.height;
float y1 = offscreen->was_offscreen ? offscreen->area.y2 : offscreen->area.y;
float y2 = offscreen->was_offscreen ? offscreen->area.y : offscreen->area.y2;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_coords (job,
min_x, min_y, max_x, max_y,
offscreen->area.x, y1, offscreen->area.x2, y2,
color);
ngl: Drop gsk_ngl_render_job_set_color Colors are not state that we carry across draw ops, so setting the color on the render job doesn't make much sense. Instead, pass the color to the various draw calls. Add a few new ones for that purpose. Also, shorten the names of some by going from 'load_vertices_from_offscreen' to 'draw_offscreen'.
2021-03-13 19:29:34 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_offscreen (GskGLRenderJob *job,
const graphene_rect_t *bounds,
const GskGLRenderOffscreen *offscreen)
ngl: Drop gsk_ngl_render_job_set_color Colors are not state that we carry across draw ops, so setting the color on the render job doesn't make much sense. Instead, pass the color to the various draw calls. Add a few new ones for that purpose. Also, shorten the names of some by going from 'load_vertices_from_offscreen' to 'draw_offscreen'.
2021-03-13 19:29:34 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_offscreen_with_color (job, bounds, offscreen,
(guint16[]) { FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO });
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
/* load_float_vertex_data */
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_with_color (GskGLRenderJob *job,
float x,
float y,
float width,
float height,
guint16 color[4])
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
float min_x = job->offset_x + x;
float min_y = job->offset_y + y;
float max_x = min_x + width;
float max_y = min_y + height;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_coords (job, min_x, min_y, max_x, max_y, 0, 0, 1, 1, color);
ngl: Drop gsk_ngl_render_job_set_color Colors are not state that we carry across draw ops, so setting the color on the render job doesn't make much sense. Instead, pass the color to the various draw calls. Add a few new ones for that purpose. Also, shorten the names of some by going from 'load_vertices_from_offscreen' to 'draw_offscreen'.
2021-03-13 19:29:34 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw (GskGLRenderJob *job,
float x,
float y,
float width,
float height)
ngl: Drop gsk_ngl_render_job_set_color Colors are not state that we carry across draw ops, so setting the color on the render job doesn't make much sense. Instead, pass the color to the various draw calls. Add a few new ones for that purpose. Also, shorten the names of some by going from 'load_vertices_from_offscreen' to 'draw_offscreen'.
2021-03-13 19:29:34 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_with_color (job, x, y, width, height,
(guint16[]) { FP_ZERO, FP_ZERO, FP_ZERO, FP_ZERO });
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
/* load_vertex_data */
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_rect_with_color (GskGLRenderJob *job,
const graphene_rect_t *bounds,
guint16 color[4])
ngl: Drop gsk_ngl_render_job_set_color Colors are not state that we carry across draw ops, so setting the color on the render job doesn't make much sense. Instead, pass the color to the various draw calls. Add a few new ones for that purpose. Also, shorten the names of some by going from 'load_vertices_from_offscreen' to 'draw_offscreen'.
2021-03-13 19:29:34 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_with_color (job,
bounds->origin.x,
bounds->origin.y,
bounds->size.width,
bounds->size.height,
color);
ngl: Drop gsk_ngl_render_job_set_color Colors are not state that we carry across draw ops, so setting the color on the render job doesn't make much sense. Instead, pass the color to the various draw calls. Add a few new ones for that purpose. Also, shorten the names of some by going from 'load_vertices_from_offscreen' to 'draw_offscreen'.
2021-03-13 19:29:34 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_rect (GskGLRenderJob *job,
const graphene_rect_t *bounds)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw (job,
bounds->origin.x,
bounds->origin.y,
bounds->size.width,
bounds->size.height);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
/* load_offscreen_vertex_data */
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_offscreen_rect (GskGLRenderJob *job,
const graphene_rect_t *bounds)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
float min_x = job->offset_x + bounds->origin.x;
float min_y = job->offset_y + bounds->origin.y;
float max_x = min_x + bounds->size.width;
float max_y = min_y + bounds->size.height;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4] = { FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO };
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_coords (job,
min_x, min_y, max_x, max_y,
0, 1, 1, 0,
color);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
static inline gboolean
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_begin_draw (GskGLRenderJob *job,
GskGLProgram *program)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
ngl: Add infrastructure for clip-variants of shaders Make ngl compile each shader three times. Once with #define NO_CLIP 1 once with #define RECT_CLIP 1 and once as before.
2021-03-07 22:50:19 +00:00
job->current_program = program;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (!gsk_gl_command_queue_begin_draw (job->command_queue,
program->program_info,
job->viewport.size.width,
job->viewport.size.height))
return FALSE;
ngl: move uniform key mapping into uniform state Previously, we translated the uniform key (an enum) into a location within the shader program in GskNglProgram. A number of performance improvements were focused around having low nubers for the uniform locations. Generally this is the case, but some drivers such as old Intel drivers on Windows may use rather large numbers for those. To combat this, we can push the translation of uniform keys into locations at the GskNglUniformState level so that we work with unranslated keys through the process until applying them. Fixes #3780
2021-03-23 19:10:12 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_uniform_state_set4fv (program->uniforms,
ngl: move uniform key mapping into uniform state Previously, we translated the uniform key (an enum) into a location within the shader program in GskNglProgram. A number of performance improvements were focused around having low nubers for the uniform locations. Generally this is the case, but some drivers such as old Intel drivers on Windows may use rather large numbers for those. To combat this, we can push the translation of uniform keys into locations at the GskNglUniformState level so that we work with unranslated keys through the process until applying them. Fixes #3780
2021-03-23 19:10:12 +00:00
program->program_info,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
UNIFORM_SHARED_VIEWPORT,
job->driver->stamps[UNIFORM_SHARED_VIEWPORT],
1,
(const float *)&job->viewport);
gsk_gl_uniform_state_set_matrix (program->uniforms,
program->program_info,
UNIFORM_SHARED_MODELVIEW,
job->driver->stamps[UNIFORM_SHARED_MODELVIEW],
&job->current_modelview->matrix);
gsk_gl_uniform_state_set_matrix (program->uniforms,
program->program_info,
UNIFORM_SHARED_PROJECTION,
job->driver->stamps[UNIFORM_SHARED_PROJECTION],
&job->projection);
gsk_gl_uniform_state_set_rounded_rect (program->uniforms,
program->program_info,
UNIFORM_SHARED_CLIP_RECT,
job->driver->stamps[UNIFORM_SHARED_CLIP_RECT],
&job->current_clip->rect);
gsk_gl_uniform_state_set1f (program->uniforms,
program->program_info,
UNIFORM_SHARED_ALPHA,
job->driver->stamps[UNIFORM_SHARED_ALPHA],
job->alpha);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
return TRUE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
ngl: Add infrastructure for clip-variants of shaders Make ngl compile each shader three times. Once with #define NO_CLIP 1 once with #define RECT_CLIP 1 and once as before.
2021-03-07 22:50:19 +00:00
#define CHOOSE_PROGRAM(job,name) \
(job->current_clip->is_fully_contained \
? job->driver->name ## _no_clip \
: (job->current_clip->is_rectilinear \
? job->driver->name ## _rect_clip \
: job->driver->name))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_split_draw (GskGLRenderJob *job)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_command_queue_split_draw (job->command_queue);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_end_draw (GskGLRenderJob *job)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_command_queue_end_draw (job->command_queue);
ngl: Add infrastructure for clip-variants of shaders Make ngl compile each shader three times. Once with #define NO_CLIP 1 once with #define RECT_CLIP 1 and once as before.
2021-03-07 22:50:19 +00:00
job->current_program = NULL;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_as_fallback (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
float scale_x = job->scale_x;
float scale_y = job->scale_y;
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
int surface_width = ceilf (node->bounds.size.width * fabs (scale_x));
int surface_height = ceilf (node->bounds.size.height * fabs (scale_y));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
GdkTexture *texture;
cairo_surface_t *surface;
cairo_surface_t *rendered_surface;
cairo_t *cr;
int texture_id;
GskTextureKey key;
if (surface_width <= 0 || surface_height <= 0)
return;
key.pointer = node;
key.pointer_is_child = FALSE;
key.scale_x = scale_x;
key.scale_y = scale_y;
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
texture_id = gsk_gl_driver_lookup_texture (job->driver, &key);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
if (texture_id != 0)
goto done;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* We first draw the recording surface on an image surface,
* just because the scaleY(-1) later otherwise screws up the
* rendering... */
{
rendered_surface = cairo_image_surface_create (CAIRO_FORMAT_ARGB32,
surface_width,
surface_height);
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
cairo_surface_set_device_scale (rendered_surface, fabs (scale_x), fabs (scale_y));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
cr = cairo_create (rendered_surface);
cairo_save (cr);
cairo_translate (cr, - floorf (node->bounds.origin.x), - floorf (node->bounds.origin.y));
/* Render nodes don't modify state, so casting away the const is fine here */
gsk_render_node_draw ((GskRenderNode *)node, cr);
cairo_restore (cr);
cairo_destroy (cr);
}
surface = cairo_image_surface_create (CAIRO_FORMAT_ARGB32,
surface_width,
surface_height);
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
cairo_surface_set_device_scale (surface, fabs (scale_x), fabs (scale_y));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
cr = cairo_create (surface);
/* We draw upside down here, so it matches what GL does. */
cairo_save (cr);
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
cairo_scale (cr, scale_x < 0 ? -1 : 1, scale_y < 0 ? 1 : -1);
cairo_translate (cr, scale_x < 0 ? - surface_width / fabs (scale_x) : 0,
scale_y < 0 ? 0 : - surface_height / fabs (scale_y));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
cairo_set_source_surface (cr, rendered_surface, 0, 0);
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
cairo_rectangle (cr, 0, 0, surface_width / fabs (scale_x), surface_height / fabs (scale_y));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
cairo_fill (cr);
cairo_restore (cr);
#ifdef G_ENABLE_DEBUG
if (job->debug_fallback)
{
cairo_move_to (cr, 0, 0);
cairo_rectangle (cr, 0, 0, node->bounds.size.width, node->bounds.size.height);
if (gsk_render_node_get_node_type (node) == GSK_CAIRO_NODE)
cairo_set_source_rgba (cr, 0.3, 0, 1, 0.25);
else
cairo_set_source_rgba (cr, 1, 0, 0, 0.25);
cairo_fill_preserve (cr);
if (gsk_render_node_get_node_type (node) == GSK_CAIRO_NODE)
cairo_set_source_rgba (cr, 0.3, 0, 1, 1);
else
cairo_set_source_rgba (cr, 1, 0, 0, 1);
cairo_stroke (cr);
}
#endif
cairo_destroy (cr);
/* Create texture to upload */
texture = gdk_texture_new_for_surface (surface);
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
texture_id = gsk_gl_driver_load_texture (job->driver, texture, FALSE);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (gdk_gl_context_has_debug (job->command_queue->context))
gdk_gl_context_label_object_printf (job->command_queue->context, GL_TEXTURE, texture_id,
"Fallback %s %d",
g_type_name_from_instance ((GTypeInstance *) node),
texture_id);
g_object_unref (texture);
cairo_surface_destroy (surface);
cairo_surface_destroy (rendered_surface);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_driver_cache_texture (job->driver, &key, texture_id);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
done:
if (scale_x < 0 || scale_y < 0)
{
gsk: Try to fix flipped fallback rendering This patch is pretty much a blind guess, but it seems to fix flipped rendering of fill and stroke nodes.
2023-08-07 21:52:00 +00:00
GskTransform *transform = gsk_transform_translate (gsk_transform_scale (NULL, scale_x < 0 ? -1 : 1, scale_y < 0 ? -1 : 1),
&GRAPHENE_POINT_INIT (scale_x < 0 ? - (node->bounds.size.width + 2 * node->bounds.origin.x) : 0,
scale_y < 0 ? - (node->bounds.size.height + 2 * node->bounds.origin.y) : 0));
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_push_modelview (job, transform);
gsk_transform_unref (transform);
}
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (!gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
goto out;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_offscreen_rect (job, &node->bounds);
gsk_gl_render_job_end_draw (job);
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
out:
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
if (scale_x < 0 || scale_y < 0)
gsk_gl_render_job_pop_modelview (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static guint
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
blur_offscreen (GskGLRenderJob *job,
GskGLRenderOffscreen *offscreen,
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
int texture_to_blur_width,
int texture_to_blur_height,
float blur_radius_x,
float blur_radius_y)
{
const GskRoundedRect new_clip = GSK_ROUNDED_RECT_INIT (0, 0, texture_to_blur_width, texture_to_blur_height);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderTarget *pass1;
GskGLRenderTarget *pass2;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
graphene_matrix_t prev_projection;
graphene_rect_t prev_viewport;
guint prev_fbo;
g_assert (blur_radius_x > 0);
g_assert (blur_radius_y > 0);
g_assert (offscreen->texture_id > 0);
g_assert (offscreen->area.x2 > offscreen->area.x);
g_assert (offscreen->area.y2 > offscreen->area.y);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_driver_create_render_target (job->driver,
MAX (texture_to_blur_width, 1),
MAX (texture_to_blur_height, 1),
job->target_format,
&pass1))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return 0;
if (texture_to_blur_width <= 0 || texture_to_blur_height <= 0)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
return gsk_gl_driver_release_render_target (job->driver, pass1, FALSE);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_driver_create_render_target (job->driver,
texture_to_blur_width,
texture_to_blur_height,
job->target_format,
&pass2))
return gsk_gl_driver_release_render_target (job->driver, pass1, FALSE);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_viewport (job, &new_clip.bounds, &prev_viewport);
gsk_gl_render_job_set_projection_from_rect (job, &new_clip.bounds, &prev_projection);
gsk_gl_render_job_set_modelview (job, NULL);
gsk_gl_render_job_push_clip (job, &new_clip);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Bind new framebuffer and clear it */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
prev_fbo = gsk_gl_command_queue_bind_framebuffer (job->command_queue, pass1->framebuffer_id);
gsk_gl_command_queue_clear (job->command_queue, 0, &job->viewport);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Begin drawing the first horizontal pass, using offscreen as the
* source texture for the program.
*/
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blur)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen->texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform1f (job->current_program,
UNIFORM_BLUR_RADIUS, 0,
blur_radius_x);
gsk_gl_program_set_uniform2f (job->current_program,
UNIFORM_BLUR_SIZE, 0,
texture_to_blur_width,
texture_to_blur_height);
gsk_gl_program_set_uniform2f (job->current_program,
UNIFORM_BLUR_DIR, 0,
1, 0);
gsk_gl_render_job_draw_coords (job,
0, 0, texture_to_blur_width, texture_to_blur_height,
0, 1, 1, 0,
(guint16[]) { FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO });
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Bind second pass framebuffer and clear it */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_command_queue_bind_framebuffer (job->command_queue, pass2->framebuffer_id);
gsk_gl_command_queue_clear (job->command_queue, 0, &job->viewport);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Draw using blur program with first pass as source texture */
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blur)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
pass1->texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform1f (job->current_program,
UNIFORM_BLUR_RADIUS, 0,
blur_radius_y);
gsk_gl_program_set_uniform2f (job->current_program,
UNIFORM_BLUR_SIZE, 0,
texture_to_blur_width,
texture_to_blur_height);
gsk_gl_program_set_uniform2f (job->current_program,
UNIFORM_BLUR_DIR, 0,
0, 1);
gsk_gl_render_job_draw_coords (job,
0, 0, texture_to_blur_width, texture_to_blur_height,
0, 1, 1, 0,
(guint16[]) { FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO });
gsk_gl_render_job_end_draw (job);
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_modelview (job);
gsk_gl_render_job_pop_clip (job);
gsk_gl_render_job_set_viewport (job, &prev_viewport, NULL);
gsk_gl_render_job_set_projection (job, &prev_projection);
gsk_gl_command_queue_bind_framebuffer (job->command_queue, prev_fbo);
gsk_gl_driver_release_render_target (job->driver, pass1, TRUE);
return gsk_gl_driver_release_render_target (job->driver, pass2, FALSE);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
blur_node (GskGLRenderJob *job,
GskGLRenderOffscreen *offscreen,
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
const GskRenderNode *node,
float blur_radius,
float *min_x,
float *max_x,
float *min_y,
float *max_y)
{
const float blur_extra = blur_radius * 2.0; /* 2.0 = shader radius_multiplier */
const float half_blur_extra = (blur_extra / 2.0);
float scale_x = job->scale_x;
float scale_y = job->scale_y;
float texture_width;
float texture_height;
g_assert (blur_radius > 0);
/* Increase texture size for the given blur radius and scale it */
texture_width = ceilf ((node->bounds.size.width + blur_extra));
texture_height = ceilf ((node->bounds.size.height + blur_extra));
/* Only blur this if the out region has no texture id yet */
if (offscreen->texture_id == 0)
{
const graphene_rect_t bounds = GRAPHENE_RECT_INIT (node->bounds.origin.x - half_blur_extra,
node->bounds.origin.y - half_blur_extra,
texture_width, texture_height);
offscreen->bounds = &bounds;
offscreen->reset_clip = TRUE;
offscreen->force_offscreen = TRUE;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, node, offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert_not_reached ();
/* Ensure that we actually got a real texture_id */
g_assert (offscreen->texture_id != 0);
offscreen->texture_id = blur_offscreen (job,
offscreen,
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
texture_width * fabs (scale_x),
texture_height * fabs (scale_y),
blur_radius * fabs (scale_x),
blur_radius * fabs (scale_y));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
init_full_texture_region (offscreen);
}
*min_x = job->offset_x + node->bounds.origin.x - half_blur_extra;
*max_x = job->offset_x + node->bounds.origin.x + node->bounds.size.width + half_blur_extra;
*min_y = job->offset_y + node->bounds.origin.y - half_blur_extra;
*max_y = job->offset_y + node->bounds.origin.y + node->bounds.size.height + half_blur_extra;
}
ngl: Optimize underlines in text Opportunistically use the coloring program for drawing underlines instead of the color program. This avoids program changes in the middle of text. For the Emoji text scrolling benchmark, this reduces the program changes per frame from > 1000 to around 100.
2021-04-09 03:11:06 +00:00
#define ATLAS_SIZE 512
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_color_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
ngl: Do nothing for transparent color nodes No need to send commands to the GPU to render transparency.
2021-07-15 17:16:37 +00:00
const GdkRGBA *rgba;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLProgram *program;
GskGLCommandBatch *batch;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
ngl: Do nothing for transparent color nodes No need to send commands to the GPU to render transparency.
2021-07-15 17:16:37 +00:00
rgba = gsk_color_node_get_color (node);
ngl: Cosmetics Use the macro we already have to check that colors are fully transparent.
2021-07-26 11:40:08 +00:00
if (RGBA_IS_CLEAR (rgba))
ngl: Do nothing for transparent color nodes No need to send commands to the GPU to render transparency.
2021-07-15 17:16:37 +00:00
return;
rgba_to_half (rgba, color);
ngl: Optimize underlines in text Opportunistically use the coloring program for drawing underlines instead of the color program. This avoids program changes in the middle of text. For the Emoji text scrolling benchmark, this reduces the program changes per frame from > 1000 to around 100.
2021-04-09 03:11:06 +00:00
/* Avoid switching away from the coloring program for
* rendering a solid color.
*/
program = CHOOSE_PROGRAM (job, coloring);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
batch = gsk_gl_command_queue_get_batch (job->command_queue);
ngl: Optimize underlines in text Opportunistically use the coloring program for drawing underlines instead of the color program. This avoids program changes in the middle of text. For the Emoji text scrolling benchmark, this reduces the program changes per frame from > 1000 to around 100.
2021-04-09 03:11:06 +00:00
Fix spelling errors These were pointed out by codespell.
2022-11-18 03:24:18 +00:00
/* Limit the size, or we end up with a coordinate overflow somewhere. */
gl renderer: Don't assume an atlas The source uniform may or may not point to a glyph atlas. The optimization we do for color nodes is only possible if it does, so check this. Fixes: #6094
2023-09-13 20:56:47 +00:00
if (job->source_is_glyph_atlas &&
node->bounds.size.width < 300 &&
ngl: Avoid a coordinate overflow This was showing up as big text selections going missing sporadically. Fixes: #4214
2021-08-28 19:42:30 +00:00
node->bounds.size.height < 300 &&
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
batch->any.kind == GSK_GL_COMMAND_KIND_DRAW &&
ngl: Optimize underlines in text Opportunistically use the coloring program for drawing underlines instead of the color program. This avoids program changes in the middle of text. For the Emoji text scrolling benchmark, this reduces the program changes per frame from > 1000 to around 100.
2021-04-09 03:11:06 +00:00
batch->any.program == program->id)
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
ngl: Optimize underlines in text Opportunistically use the coloring program for drawing underlines instead of the color program. This avoids program changes in the middle of text. For the Emoji text scrolling benchmark, this reduces the program changes per frame from > 1000 to around 100.
2021-04-09 03:11:06 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, program))
{
/* The top left few pixels in our atlases are always
* solid white, so we can use it here, without
* having to choose any particular atlas texture.
*/
offscreen.was_offscreen = FALSE;
offscreen.area.x = 1.f / ATLAS_SIZE;
offscreen.area.y = 1.f / ATLAS_SIZE;
offscreen.area.x2 = 2.f / ATLAS_SIZE;
offscreen.area.y2 = 2.f / ATLAS_SIZE;
gsk_gl_render_job_draw_offscreen_with_color (job,
&node->bounds,
&offscreen,
color);
ngl: Optimize underlines in text Opportunistically use the coloring program for drawing underlines instead of the color program. This avoids program changes in the middle of text. For the Emoji text scrolling benchmark, this reduces the program changes per frame from > 1000 to around 100.
2021-04-09 03:11:06 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_end_draw (job);
}
ngl: Optimize underlines in text Opportunistically use the coloring program for drawing underlines instead of the color program. This avoids program changes in the middle of text. For the Emoji text scrolling benchmark, this reduces the program changes per frame from > 1000 to around 100.
2021-04-09 03:11:06 +00:00
}
else
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, color)))
{
gsk_gl_render_job_draw_rect_with_color (job, &node->bounds, color);
gsk_gl_render_job_end_draw (job);
}
ngl: Optimize underlines in text Opportunistically use the coloring program for drawing underlines instead of the color program. This avoids program changes in the middle of text. For the Emoji text scrolling benchmark, this reduces the program changes per frame from > 1000 to around 100.
2021-04-09 03:11:06 +00:00
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_linear_gradient_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskColorStop *stops = gsk_linear_gradient_node_get_color_stops (node, NULL);
const graphene_point_t *start = gsk_linear_gradient_node_get_start (node);
const graphene_point_t *end = gsk_linear_gradient_node_get_end (node);
int n_color_stops = gsk_linear_gradient_node_get_n_color_stops (node);
gboolean repeat = gsk_render_node_get_node_type (node) == GSK_REPEATING_LINEAR_GRADIENT_NODE;
float x1 = job->offset_x + start->x;
float x2 = job->offset_x + end->x;
float y1 = job->offset_y + start->y;
float y2 = job->offset_y + end->y;
g_assert (n_color_stops < MAX_GRADIENT_STOPS);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, linear_gradient)))
{
gsk_gl_program_set_uniform1i (job->current_program,
UNIFORM_LINEAR_GRADIENT_NUM_COLOR_STOPS, 0,
n_color_stops);
gsk_gl_program_set_uniform1fv (job->current_program,
UNIFORM_LINEAR_GRADIENT_COLOR_STOPS, 0,
n_color_stops * 5,
(const float *)stops);
gsk_gl_program_set_uniform4f (job->current_program,
UNIFORM_LINEAR_GRADIENT_POINTS, 0,
x1, y1, x2 - x1, y2 - y1);
gsk_gl_program_set_uniform1i (job->current_program,
UNIFORM_LINEAR_GRADIENT_REPEAT, 0,
repeat);
gsk_gl_render_job_draw_rect (job, &node->bounds);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_conic_gradient_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
static const float scale = 0.5f * M_1_PI;
const GskColorStop *stops = gsk_conic_gradient_node_get_color_stops (node, NULL);
const graphene_point_t *center = gsk_conic_gradient_node_get_center (node);
int n_color_stops = gsk_conic_gradient_node_get_n_color_stops (node);
float angle = gsk_conic_gradient_node_get_angle (node);
float bias = angle * scale + 2.0f;
g_assert (n_color_stops < MAX_GRADIENT_STOPS);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, conic_gradient)))
{
gsk_gl_program_set_uniform1i (job->current_program,
UNIFORM_CONIC_GRADIENT_NUM_COLOR_STOPS, 0,
n_color_stops);
gsk_gl_program_set_uniform1fv (job->current_program,
UNIFORM_CONIC_GRADIENT_COLOR_STOPS, 0,
n_color_stops * 5,
(const float *)stops);
gsk_gl_program_set_uniform4f (job->current_program,
UNIFORM_CONIC_GRADIENT_GEOMETRY, 0,
job->offset_x + center->x,
job->offset_y + center->y,
scale,
bias);
gsk_gl_render_job_draw_rect (job, &node->bounds);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_radial_gradient_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
int n_color_stops = gsk_radial_gradient_node_get_n_color_stops (node);
const GskColorStop *stops = gsk_radial_gradient_node_get_color_stops (node, NULL);
const graphene_point_t *center = gsk_radial_gradient_node_get_center (node);
float start = gsk_radial_gradient_node_get_start (node);
float end = gsk_radial_gradient_node_get_end (node);
float hradius = gsk_radial_gradient_node_get_hradius (node);
float vradius = gsk_radial_gradient_node_get_vradius (node);
gboolean repeat = gsk_render_node_get_node_type (node) == GSK_REPEATING_RADIAL_GRADIENT_NODE;
float scale = 1.0f / (end - start);
float bias = -start * scale;
g_assert (n_color_stops < MAX_GRADIENT_STOPS);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, radial_gradient)))
{
gsk_gl_program_set_uniform1i (job->current_program,
UNIFORM_RADIAL_GRADIENT_NUM_COLOR_STOPS, 0,
n_color_stops);
gsk_gl_program_set_uniform1fv (job->current_program,
UNIFORM_RADIAL_GRADIENT_COLOR_STOPS, 0,
n_color_stops * 5,
(const float *)stops);
gsk_gl_program_set_uniform1i (job->current_program,
UNIFORM_RADIAL_GRADIENT_REPEAT, 0,
repeat);
gsk_gl_program_set_uniform2f (job->current_program,
UNIFORM_RADIAL_GRADIENT_RANGE, 0,
scale, bias);
gsk_gl_program_set_uniform4f (job->current_program,
UNIFORM_RADIAL_GRADIENT_GEOMETRY, 0,
job->offset_x + center->x,
job->offset_y + center->y,
1.0f / (hradius * job->scale_x),
1.0f / (vradius * job->scale_y));
gsk_gl_render_job_draw_rect (job, &node->bounds);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_clipped_child (GskGLRenderJob *job,
const GskRenderNode *child,
const graphene_rect_t *clip)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
graphene_rect_t transformed_clip;
GskRoundedRect intersection;
Improve rounded rect intersection Move the rounded rect/rect intersection code to gskroundedrect.c, make it more precise, and add tests for it. Update the callers in the GL renderer.
2023-05-01 13:52:46 +00:00
GskRoundedRectIntersection result;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_transform_bounds (job, clip, &transformed_clip);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (job->current_clip->is_rectilinear)
{
memset (&intersection.corner, 0, sizeof intersection.corner);
graphene_rect_intersection (&transformed_clip,
&job->current_clip->rect.bounds,
&intersection.bounds);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (job, &intersection);
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_pop_clip (job);
Improve rounded rect intersection Move the rounded rect/rect intersection code to gskroundedrect.c, make it more precise, and add tests for it. Update the callers in the GL renderer.
2023-05-01 13:52:46 +00:00
return;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
Improve rounded rect intersection Move the rounded rect/rect intersection code to gskroundedrect.c, make it more precise, and add tests for it. Update the callers in the GL renderer.
2023-05-01 13:52:46 +00:00
result = gsk_rounded_rect_intersect_with_rect (&job->current_clip->rect,
&transformed_clip,
&intersection);
if (result == GSK_INTERSECTION_EMPTY)
return;
if (result == GSK_INTERSECTION_NONEMPTY)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (job, &intersection);
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_pop_clip (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
else
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
ngl: Avoid huge intermediate textures Instead of rendering the unclipped child to a texture (and risking blowing the texture size limit, and bad downscaling), just render the clipped region, and live with the fact that we can't cache the rendered texture. This avoid bad artifacts when scrolling long textviews in rounded clips.
2021-04-02 23:33:44 +00:00
offscreen.bounds = clip;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
offscreen.force_offscreen = TRUE;
ngl: Avoid huge intermediate textures Instead of rendering the unclipped child to a texture (and risking blowing the texture size limit, and bad downscaling), just render the clipped region, and live with the fact that we can't cache the rendered texture. This avoid bad artifacts when scrolling long textviews in rounded clips.
2021-04-02 23:33:44 +00:00
offscreen.reset_clip = TRUE;
offscreen.do_not_cache = TRUE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node_with_offscreen (job, child, &offscreen);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (offscreen.texture_id);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_offscreen_rect (job, clip);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_clip_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const graphene_rect_t *clip = gsk_clip_node_get_clip (node);
const GskRenderNode *child = gsk_clip_node_get_child (node);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_clipped_child (job, child, clip);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_rounded_clip_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRenderNode *child = gsk_rounded_clip_node_get_child (node);
const GskRoundedRect *clip = gsk_rounded_clip_node_get_clip (node);
GskRoundedRect transformed_clip;
gboolean need_offscreen;
if (node_is_invisible (child))
return;
gsk: Transform rounded rects better
2023-04-30 15:19:47 +00:00
gsk_gl_render_job_transform_rounded_rect (job, clip, &transformed_clip);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (job->current_clip->is_rectilinear)
{
GskRoundedRect intersected_clip;
Improve rounded rect intersection Move the rounded rect/rect intersection code to gskroundedrect.c, make it more precise, and add tests for it. Update the callers in the GL renderer.
2023-05-01 13:52:46 +00:00
GskRoundedRectIntersection result;
result = gsk_rounded_rect_intersect_with_rect (&transformed_clip,
&job->current_clip->rect.bounds,
&intersected_clip);
if (result == GSK_INTERSECTION_EMPTY)
return;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Improve rounded rect intersection Move the rounded rect/rect intersection code to gskroundedrect.c, make it more precise, and add tests for it. Update the callers in the GL renderer.
2023-05-01 13:52:46 +00:00
if (result == GSK_INTERSECTION_NONEMPTY)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (job, &intersected_clip);
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_pop_clip (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
}
}
/* After this point we are really working with a new and a current clip
* which both have rounded corners.
*/
if (job->clip->len <= 1)
need_offscreen = FALSE;
gsk: Avoid some offscreens We don't have to be too conservative when checking if a rect is contained in a rounded rect, we have a precise api for this.
2023-05-05 02:30:12 +00:00
else if (gsk_rounded_rect_contains_rect (&job->current_clip->rect, &transformed_clip.bounds))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
need_offscreen = FALSE;
else
need_offscreen = TRUE;
if (!need_offscreen)
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (job, &transformed_clip);
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_pop_clip (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
else
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
offscreen.bounds = &node->bounds;
offscreen.force_offscreen = TRUE;
ngl: Be explicit about offscreen clipping Explicitly set all the input arguments. This makes it clearer what is going on.
2021-03-31 18:34:16 +00:00
offscreen.reset_clip = FALSE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (job, &transformed_clip);
if (!gsk_gl_render_job_visit_node_with_offscreen (job, child, &offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert_not_reached ();
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_clip (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (offscreen.texture_id);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_offscreen (job, &node->bounds, &offscreen);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_rect_border_node (GskGLRenderJob *job,
const GskRenderNode *node)
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
{
const GdkRGBA *colors = gsk_border_node_get_colors (node);
const float *widths = gsk_border_node_get_widths (node);
const graphene_point_t *origin = &node->bounds.origin;
const graphene_size_t *size = &node->bounds.size;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, color)))
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (widths[0] > 0)
{
rgba_to_half (&colors[0], color);
gsk_gl_render_job_draw_rect_with_color (job,
&GRAPHENE_RECT_INIT (origin->x, origin->y, size->width - widths[1], widths[0]),
color);
}
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (widths[1] > 0)
{
rgba_to_half (&colors[1], color);
gsk_gl_render_job_draw_rect_with_color (job,
&GRAPHENE_RECT_INIT (origin->x + size->width - widths[1], origin->y, widths[1], size->height - widths[2]),
color);
}
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (widths[2] > 0)
{
rgba_to_half (&colors[2], color);
gsk_gl_render_job_draw_rect_with_color (job,
&GRAPHENE_RECT_INIT (origin->x + widths[3], origin->y + size->height - widths[2], size->width - widths[3], widths[2]),
color);
}
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (widths[3] > 0)
{
rgba_to_half (&colors[3], color);
gsk_gl_render_job_draw_rect_with_color (job,
&GRAPHENE_RECT_INIT (origin->x, origin->y + widths[0], widths[3], size->height - widths[0]),
color);
}
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_end_draw (job);
}
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_border_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRoundedRect *rounded_outline = gsk_border_node_get_outline (node);
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
const GdkRGBA *colors = gsk_border_node_get_colors (node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
const float *widths = gsk_border_node_get_widths (node);
struct {
float w;
float h;
} sizes[4];
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
float min_x = job->offset_x + node->bounds.origin.x;
float min_y = job->offset_y + node->bounds.origin.y;
float max_x = min_x + node->bounds.size.width;
float max_y = min_y + node->bounds.size.height;
GskRoundedRect outline;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
memset (sizes, 0, sizeof sizes);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (widths[0] > 0)
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
{
sizes[0].h = MAX (widths[0], rounded_outline->corner[0].height);
sizes[1].h = MAX (widths[0], rounded_outline->corner[1].height);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (widths[1] > 0)
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
{
sizes[1].w = MAX (widths[1], rounded_outline->corner[1].width);
sizes[2].w = MAX (widths[1], rounded_outline->corner[2].width);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
if (widths[2] > 0)
{
sizes[2].h = MAX (widths[2], rounded_outline->corner[2].height);
sizes[3].h = MAX (widths[2], rounded_outline->corner[3].height);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
if (widths[3] > 0)
{
sizes[0].w = MAX (widths[3], rounded_outline->corner[0].width);
sizes[3].w = MAX (widths[3], rounded_outline->corner[3].width);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_translate_rounded_rect (job, rounded_outline, &outline);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, border)))
{
gsk_gl_program_set_uniform4fv (job->current_program,
UNIFORM_BORDER_WIDTHS, 0,
1,
widths);
gsk_gl_program_set_uniform_rounded_rect (job->current_program,
UNIFORM_BORDER_OUTLINE_RECT, 0,
&outline);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (widths[0] > 0)
{
GskGLDrawVertex *vertices = gsk_gl_command_queue_add_vertices (job->command_queue);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
rgba_to_half (&colors[0], color);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[0] = (GskGLDrawVertex) { .position = { min_x, min_y }, .uv = { 0, 1 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[1] = (GskGLDrawVertex) { .position = { min_x + sizes[0].w, min_y + sizes[0].h }, .uv = { 0, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[2] = (GskGLDrawVertex) { .position = { max_x, min_y }, .uv = { 1, 1 }, .color = { color[0], color[1], color[2], color[3] } };
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[3] = (GskGLDrawVertex) { .position = { max_x - sizes[1].w, min_y + sizes[1].h }, .uv = { 1, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[4] = (GskGLDrawVertex) { .position = { min_x + sizes[0].w, min_y + sizes[0].h }, .uv = { 0, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[5] = (GskGLDrawVertex) { .position = { max_x, min_y }, .uv = { 1, 1 }, .color = { color[0], color[1], color[2], color[3] } };
}
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (widths[1] > 0)
{
GskGLDrawVertex *vertices = gsk_gl_command_queue_add_vertices (job->command_queue);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
rgba_to_half (&colors[1], color);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[0] = (GskGLDrawVertex) { .position = { max_x - sizes[1].w, min_y + sizes[1].h }, .uv = { 0, 1 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[1] = (GskGLDrawVertex) { .position = { max_x - sizes[2].w, max_y - sizes[2].h }, .uv = { 0, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[2] = (GskGLDrawVertex) { .position = { max_x, min_y }, .uv = { 1, 1 }, .color = { color[0], color[1], color[2], color[3] } };
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[3] = (GskGLDrawVertex) { .position = { max_x, max_y }, .uv = { 1, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[4] = (GskGLDrawVertex) { .position = { max_x - sizes[2].w, max_y - sizes[2].h }, .uv = { 0, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[5] = (GskGLDrawVertex) { .position = { max_x, min_y }, .uv = { 1, 1 }, .color = { color[0], color[1], color[2], color[3] } };
}
ngl: Add infrastructure for clip-variants of shaders Make ngl compile each shader three times. Once with #define NO_CLIP 1 once with #define RECT_CLIP 1 and once as before.
2021-03-07 22:50:19 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (widths[2] > 0)
{
GskGLDrawVertex *vertices = gsk_gl_command_queue_add_vertices (job->command_queue);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
rgba_to_half (&colors[2], color);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[0] = (GskGLDrawVertex) { .position = { min_x + sizes[3].w, max_y - sizes[3].h }, .uv = { 0, 1 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[1] = (GskGLDrawVertex) { .position = { min_x, max_y }, .uv = { 0, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[2] = (GskGLDrawVertex) { .position = { max_x - sizes[2].w, max_y - sizes[2].h }, .uv = { 1, 1 }, .color = { color[0], color[1], color[2], color[3] } };
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[3] = (GskGLDrawVertex) { .position = { max_x, max_y }, .uv = { 1, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[4] = (GskGLDrawVertex) { .position = { min_x , max_y }, .uv = { 0, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[5] = (GskGLDrawVertex) { .position = { max_x - sizes[2].w, max_y - sizes[2].h }, .uv = { 1, 1 }, .color = { color[0], color[1], color[2], color[3] } };
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (widths[3] > 0)
{
GskGLDrawVertex *vertices = gsk_gl_command_queue_add_vertices (job->command_queue);
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
rgba_to_half (&colors[3], color);
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[0] = (GskGLDrawVertex) { .position = { min_x, min_y }, .uv = { 0, 1 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[1] = (GskGLDrawVertex) { .position = { min_x, max_y }, .uv = { 0, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[2] = (GskGLDrawVertex) { .position = { min_x + sizes[0].w, min_y + sizes[0].h }, .uv = { 1, 1 }, .color = { color[0], color[1], color[2], color[3] } };
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[3] = (GskGLDrawVertex) { .position = { min_x + sizes[3].w, max_y - sizes[3].h }, .uv = { 1, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[4] = (GskGLDrawVertex) { .position = { min_x, max_y }, .uv = { 0, 0 }, .color = { color[0], color[1], color[2], color[3] } };
vertices[5] = (GskGLDrawVertex) { .position = { min_x + sizes[0].w, min_y + sizes[0].h }, .uv = { 1, 1 }, .color = { color[0], color[1], color[2], color[3] } };
}
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_end_draw (job);
ngl: Refactor gsk_ngl_render_job_visit_border_node With color sorting out of the picture, this can be much simpler.
2021-03-13 18:40:04 +00:00
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
/* A special case for a pattern that occurs frequently with CSS
* backgrounds: two sibling nodes, the first of which is a rounded
* clip node with a color node as child, and the second one is a
* border node, with the same outline as the clip node. We render
* this using the filled_border shader.
*/
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_css_background (GskGLRenderJob *job,
const GskRenderNode *node,
const GskRenderNode *node2)
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
{
const GskRenderNode *child = gsk_rounded_clip_node_get_child (node);
const GskRoundedRect *rounded_outline = gsk_border_node_get_outline (node2);
const float *widths = gsk_border_node_get_widths (node2);
float min_x = job->offset_x + node2->bounds.origin.x;
float min_y = job->offset_y + node2->bounds.origin.y;
float max_x = min_x + node2->bounds.size.width;
float max_y = min_y + node2->bounds.size.height;
GskRoundedRect outline;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLDrawVertex *vertices;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
guint16 color2[4];
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
if (node_is_invisible (node2))
return;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
rgba_to_half (&gsk_border_node_get_colors (node2)[0], color);
rgba_to_half (gsk_color_node_get_color (child), color2);
ngl: Use fp16 for colors This reduces the size of our Vertex struct from 48 to 32 bytes. It would be nicer if we could store the colors in fp16 format in the rendernodes, and avoid conversion here. But this is still good.
2021-04-07 01:55:50 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_translate_rounded_rect (job, rounded_outline, &outline);
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, filled_border)))
{
gsk_gl_program_set_uniform4fv (job->current_program,
UNIFORM_FILLED_BORDER_WIDTHS, 0,
1,
widths);
gsk_gl_program_set_uniform_rounded_rect (job->current_program,
UNIFORM_FILLED_BORDER_OUTLINE_RECT, 0,
&outline);
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices = gsk_gl_command_queue_add_vertices (job->command_queue);
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
vertices[0] = (GskGLDrawVertex) { .position = { min_x, min_y }, .color = { color[0], color[1], color[2], color[3] }, .color2 = { color2[0], color2[1], color2[2], color2[3] } };
vertices[1] = (GskGLDrawVertex) { .position = { min_x, max_y }, .color = { color[0], color[1], color[2], color[3] }, .color2 = { color2[0], color2[1], color2[2], color2[3] } };
vertices[2] = (GskGLDrawVertex) { .position = { max_x, min_y }, .color = { color[0], color[1], color[2], color[3] }, .color2 = { color2[0], color2[1], color2[2], color2[3] } };
vertices[3] = (GskGLDrawVertex) { .position = { max_x, max_y }, .color = { color[0], color[1], color[2], color[3] }, .color2 = { color2[0], color2[1], color2[2], color2[3] } };
vertices[4] = (GskGLDrawVertex) { .position = { min_x, max_y }, .color = { color[0], color[1], color[2], color[3] }, .color2 = { color2[0], color2[1], color2[2], color2[3] } };
vertices[5] = (GskGLDrawVertex) { .position = { max_x, min_y }, .color = { color[0], color[1], color[2], color[3] }, .color2 = { color2[0], color2[1], color2[2], color2[3] } };
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_end_draw (job);
}
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Returns TRUE if applying @transform to @bounds
* yields an axis-aligned rectangle
*/
static gboolean
result_is_axis_aligned (GskTransform *transform,
const graphene_rect_t *bounds)
{
graphene_matrix_t m;
graphene_quad_t q;
graphene_rect_t b;
graphene_point_t b1, b2;
const graphene_point_t *p;
gsk_transform_to_matrix (transform, &m);
gsk_matrix_transform_rect (&m, bounds, &q);
graphene_quad_bounds (&q, &b);
graphene_rect_get_top_left (&b, &b1);
graphene_rect_get_bottom_right (&b, &b2);
for (guint i = 0; i < 4; i++)
{
p = graphene_quad_get_point (&q, i);
if (fabs (p->x - b1.x) > FLT_EPSILON && fabs (p->x - b2.x) > FLT_EPSILON)
return FALSE;
if (fabs (p->y - b1.y) > FLT_EPSILON && fabs (p->y - b2.y) > FLT_EPSILON)
return FALSE;
}
return TRUE;
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_transform_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
GskTransform *transform = gsk_transform_node_get_transform (node);
const GskTransformCategory category = gsk_transform_get_category (transform);
const GskRenderNode *child = gsk_transform_node_get_child (node);
switch (category)
{
case GSK_TRANSFORM_CATEGORY_IDENTITY:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, child);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_TRANSFORM_CATEGORY_2D_TRANSLATE:
{
float dx, dy;
gsk: Add a shortcut for a common transform case Add a private getter for the translation part of a transform node, to avoid pulling it out of the transform over and over.
2021-07-24 14:48:20 +00:00
gsk_transform_node_get_translate (node, &dx, &dy);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_offset (job, dx, dy);
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_offset (job, -dx, -dy);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
break;
case GSK_TRANSFORM_CATEGORY_2D_AFFINE:
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_modelview (job, transform);
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_pop_modelview (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
break;
case GSK_TRANSFORM_CATEGORY_2D:
gl: Linear gradients don't support 3d transforms Add another helper similar to the one for transforms, but that only works on 2d transforms. Fixes #4501
2021-12-10 19:07:39 +00:00
if (node_supports_2d_transform (child))
{
gsk_gl_render_job_push_modelview (job, transform);
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_pop_modelview (job);
return;
}
G_GNUC_FALLTHROUGH;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
case GSK_TRANSFORM_CATEGORY_3D:
case GSK_TRANSFORM_CATEGORY_ANY:
case GSK_TRANSFORM_CATEGORY_UNKNOWN:
if (node_supports_transform (child))
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_modelview (job, transform);
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_pop_modelview (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
else
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
Improve transformed offscreen rendering Preserve the scale for 2D transforms to avoid a pixellated appearance.
2021-05-13 21:32:52 +00:00
float sx = 1, sy = 1;
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
gboolean linear_filter = FALSE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
offscreen.bounds = &child->bounds;
ngl: Plug a memory leak This was introduced in f9457af128012fd3935a69fdcefb.
2021-04-02 22:41:53 +00:00
offscreen.force_offscreen = FALSE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
offscreen.reset_clip = TRUE;
if (!result_is_axis_aligned (transform, &child->bounds))
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
linear_filter = TRUE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Improve transformed offscreen rendering Preserve the scale for 2D transforms to avoid a pixellated appearance.
2021-05-13 21:32:52 +00:00
if (category == GSK_TRANSFORM_CATEGORY_2D)
{
graphene_matrix_t m;
double a, b, c, d, tx, ty;
g_assert (transform != NULL);
gsk_transform_to_matrix (transform, &m);
if (graphene_matrix_to_2d (&m, &a, &b, &c, &d, &tx, &ty))
{
sx = sqrt (a * a + b * b);
sy = sqrt (c * c + d * d);
}
else
sx = sy = 1;
if (sx != 1 || sy != 1)
{
GskTransform *scale;
scale = gsk_transform_translate (gsk_transform_scale (NULL, sx, sy), &GRAPHENE_POINT_INIT (tx, ty));
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_modelview (job, scale);
Improve transformed offscreen rendering Preserve the scale for 2D transforms to avoid a pixellated appearance.
2021-05-13 21:32:52 +00:00
transform = gsk_transform_transform (gsk_transform_invert (scale), transform);
}
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (gsk_gl_render_job_visit_node_with_offscreen (job, child, &offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
/* For non-trivial transforms, we draw everything on a texture and then
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
* draw the texture transformed.
*/
Improve transformed offscreen rendering Preserve the scale for 2D transforms to avoid a pixellated appearance.
2021-05-13 21:32:52 +00:00
if (transform)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_modelview (job, transform);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture_with_filter (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id,
linear_filter ? GL_LINEAR : GL_NEAREST,
linear_filter ? GL_LINEAR : GL_NEAREST);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_offscreen (job, &child->bounds, &offscreen);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Improve transformed offscreen rendering Preserve the scale for 2D transforms to avoid a pixellated appearance.
2021-05-13 21:32:52 +00:00
if (transform)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_modelview (job);
Improve transformed offscreen rendering Preserve the scale for 2D transforms to avoid a pixellated appearance.
2021-05-13 21:32:52 +00:00
}
if (category == GSK_TRANSFORM_CATEGORY_2D)
{
if (sx != 1 || sy != 1)
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_modelview (job);
Improve transformed offscreen rendering Preserve the scale for 2D transforms to avoid a pixellated appearance.
2021-05-13 21:32:52 +00:00
gsk_transform_unref (transform);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
break;
default:
g_assert_not_reached ();
}
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_unblurred_inset_shadow_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRoundedRect *outline = gsk_inset_shadow_node_get_outline (node);
GskRoundedRect transformed_outline;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_translate_rounded_rect (job, outline, &transformed_outline);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, inset_shadow)))
{
gsk_gl_program_set_uniform_rounded_rect (job->current_program,
UNIFORM_INSET_SHADOW_OUTLINE_RECT, 0,
&transformed_outline);
gsk_gl_program_set_uniform1f (job->current_program,
UNIFORM_INSET_SHADOW_SPREAD, 0,
gsk_inset_shadow_node_get_spread (node));
gsk_gl_program_set_uniform2f (job->current_program,
UNIFORM_INSET_SHADOW_OFFSET, 0,
gsk_inset_shadow_node_get_dx (node),
gsk_inset_shadow_node_get_dy (node));
rgba_to_half (gsk_inset_shadow_node_get_color (node), color);
gsk_gl_render_job_draw_rect_with_color (job, &node->bounds, color);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_blurred_inset_shadow_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRoundedRect *node_outline = gsk_inset_shadow_node_get_outline (node);
float blur_radius = gsk_inset_shadow_node_get_blur_radius (node);
float offset_x = gsk_inset_shadow_node_get_dx (node);
float offset_y = gsk_inset_shadow_node_get_dy (node);
float scale_x = job->scale_x;
float scale_y = job->scale_y;
float blur_extra = blur_radius * 2.0; /* 2.0 = shader radius_multiplier */
float half_blur_extra = blur_radius;
float texture_width;
float texture_height;
int blurred_texture_id;
GskTextureKey key;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (blur_radius > 0);
texture_width = ceilf ((node_outline->bounds.size.width + blur_extra) * scale_x);
texture_height = ceilf ((node_outline->bounds.size.height + blur_extra) * scale_y);
key.pointer = node;
key.pointer_is_child = FALSE;
key.scale_x = scale_x;
key.scale_y = scale_y;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
blurred_texture_id = gsk_gl_driver_lookup_texture (job->driver, &key);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (blurred_texture_id == 0)
{
float spread = gsk_inset_shadow_node_get_spread (node) + half_blur_extra;
GskRoundedRect transformed_outline;
GskRoundedRect outline_to_blur;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderTarget *render_target;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
graphene_matrix_t prev_projection;
graphene_rect_t prev_viewport;
guint prev_fbo;
/* TODO: In the following code, we have to be careful about where we apply the scale.
* We're manually scaling stuff (e.g. the outline) so we can later use texture_width
* and texture_height (which are already scaled) as the geometry and keep the modelview
* at a scale of 1. That's kinda complicated though... */
/* Outline of what we actually want to blur later.
* Spread grows inside, so we don't need to account for that. But the blur will need
* to read outside of the inset shadow, so we need to draw some color in there. */
outline_to_blur = *node_outline;
gsk_rounded_rect_shrink (&outline_to_blur,
-half_blur_extra,
-half_blur_extra,
-half_blur_extra,
-half_blur_extra);
/* Fit to our texture */
outline_to_blur.bounds.origin.x = 0;
outline_to_blur.bounds.origin.y = 0;
outline_to_blur.bounds.size.width *= scale_x;
outline_to_blur.bounds.size.height *= scale_y;
for (guint i = 0; i < 4; i ++)
{
outline_to_blur.corner[i].width *= scale_x;
outline_to_blur.corner[i].height *= scale_y;
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_driver_create_render_target (job->driver,
texture_width, texture_height,
get_target_format (job, node),
&render_target))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert_not_reached ();
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_viewport_for_size (job, texture_width, texture_height, &prev_viewport);
gsk_gl_render_job_set_projection_for_size (job, texture_width, texture_height, &prev_projection);
gsk_gl_render_job_set_modelview (job, NULL);
gsk_gl_render_job_push_clip (job, &GSK_ROUNDED_RECT_INIT (0, 0, texture_width, texture_height));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
prev_fbo = gsk_gl_command_queue_bind_framebuffer (job->command_queue, render_target->framebuffer_id);
gsk_gl_command_queue_clear (job->command_queue, 0, &job->viewport);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_translate_rounded_rect (job, &outline_to_blur, &transformed_outline);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Actual inset shadow outline drawing */
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, inset_shadow)))
{
gsk_gl_program_set_uniform_rounded_rect (job->current_program,
UNIFORM_INSET_SHADOW_OUTLINE_RECT, 0,
&transformed_outline);
gsk_gl_program_set_uniform1f (job->current_program,
UNIFORM_INSET_SHADOW_SPREAD, 0,
spread * MAX (scale_x, scale_y));
gsk_gl_program_set_uniform2f (job->current_program,
UNIFORM_INSET_SHADOW_OFFSET, 0,
offset_x * scale_x,
offset_y * scale_y);
rgba_to_half (gsk_inset_shadow_node_get_color (node), color);
gsk_gl_render_job_draw_with_color (job,
0, 0, texture_width, texture_height,
color);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_modelview (job);
gsk_gl_render_job_pop_clip (job);
gsk_gl_render_job_set_projection (job, &prev_projection);
gsk_gl_render_job_set_viewport (job, &prev_viewport, NULL);
gsk_gl_command_queue_bind_framebuffer (job->command_queue, prev_fbo);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
offscreen.texture_id = render_target->texture_id;
init_full_texture_region (&offscreen);
blurred_texture_id = blur_offscreen (job,
&offscreen,
texture_width,
texture_height,
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
blur_radius * fabs (scale_x),
blur_radius * fabs (scale_y));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_driver_release_render_target (job->driver, render_target, TRUE);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_driver_cache_texture (job->driver, &key, blurred_texture_id);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
g_assert (blurred_texture_id != 0);
/* Blur the rendered unblurred inset shadow */
/* Use a clip to cut away the unwanted parts outside of the original outline */
{
const gboolean needs_clip = !gsk_rounded_rect_is_rectilinear (node_outline);
const float tx1 = half_blur_extra * scale_x / texture_width;
const float tx2 = 1.0 - tx1;
const float ty1 = half_blur_extra * scale_y / texture_height;
const float ty2 = 1.0 - ty1;
if (needs_clip)
{
GskRoundedRect node_clip;
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_translate_rounded_rect (job, node_outline, &node_clip);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (job, &node_clip);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
offscreen.was_offscreen = TRUE;
offscreen.area.x = tx1;
offscreen.area.y = ty1;
offscreen.area.x2 = tx2;
offscreen.area.y2 = ty2;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
blurred_texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_offscreen (job, &node->bounds, &offscreen);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (needs_clip)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_clip (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_unblurred_outset_shadow_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRoundedRect *outline = gsk_outset_shadow_node_get_outline (node);
GskRoundedRect transformed_outline;
float x = node->bounds.origin.x;
float y = node->bounds.origin.y;
float w = node->bounds.size.width;
float h = node->bounds.size.height;
float spread = gsk_outset_shadow_node_get_spread (node);
float dx = gsk_outset_shadow_node_get_dx (node);
float dy = gsk_outset_shadow_node_get_dy (node);
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
const float edge_sizes[] = { // Top, right, bottom, left
spread - dy, spread + dx, spread + dy, spread - dx
};
const float corner_sizes[][2] = { // top left, top right, bottom right, bottom left
{ outline->corner[0].width + spread - dx, outline->corner[0].height + spread - dy },
{ outline->corner[1].width + spread + dx, outline->corner[1].height + spread - dy },
{ outline->corner[2].width + spread + dx, outline->corner[2].height + spread + dy },
{ outline->corner[3].width + spread - dx, outline->corner[3].height + spread + dy },
};
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
rgba_to_half (gsk_outset_shadow_node_get_color (node), color);
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_translate_rounded_rect (job, outline, &transformed_outline);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, unblurred_outset_shadow)))
{
gsk_gl_program_set_uniform_rounded_rect (job->current_program,
UNIFORM_UNBLURRED_OUTSET_SHADOW_OUTLINE_RECT, 0,
&transformed_outline);
gsk_gl_program_set_uniform1f (job->current_program,
UNIFORM_UNBLURRED_OUTSET_SHADOW_SPREAD, 0,
spread);
gsk_gl_program_set_uniform2f (job->current_program,
UNIFORM_UNBLURRED_OUTSET_SHADOW_OFFSET, 0,
dx, dy);
/* Corners... */
if (corner_sizes[0][0] > 0 && corner_sizes[0][1] > 0) /* Top left */
gsk_gl_render_job_draw_with_color (job,
x, y, corner_sizes[0][0], corner_sizes[0][1],
color);
if (corner_sizes[1][0] > 0 && corner_sizes[1][1] > 0) /* Top right */
gsk_gl_render_job_draw_with_color (job,
x + w - corner_sizes[1][0], y,
corner_sizes[1][0], corner_sizes[1][1],
color);
if (corner_sizes[2][0] > 0 && corner_sizes[2][1] > 0) /* Bottom right */
gsk_gl_render_job_draw_with_color (job,
x + w - corner_sizes[2][0], y + h - corner_sizes[2][1],
corner_sizes[2][0], corner_sizes[2][1],
color);
if (corner_sizes[3][0] > 0 && corner_sizes[3][1] > 0) /* Bottom left */
gsk_gl_render_job_draw_with_color (job,
x, y + h - corner_sizes[3][1],
corner_sizes[3][0], corner_sizes[3][1],
color);
/* Edges... */;
if (edge_sizes[0] > 0) /* Top */
gsk_gl_render_job_draw_with_color (job,
x + corner_sizes[0][0], y,
w - corner_sizes[0][0] - corner_sizes[1][0], edge_sizes[0],
color);
if (edge_sizes[1] > 0) /* Right */
gsk_gl_render_job_draw_with_color (job,
x + w - edge_sizes[1], y + corner_sizes[1][1],
edge_sizes[1], h - corner_sizes[1][1] - corner_sizes[2][1],
color);
if (edge_sizes[2] > 0) /* Bottom */
gsk_gl_render_job_draw_with_color (job,
x + corner_sizes[3][0], y + h - edge_sizes[2],
w - corner_sizes[3][0] - corner_sizes[2][0], edge_sizes[2],
color);
if (edge_sizes[3] > 0) /* Left */
gsk_gl_render_job_draw_with_color (job,
x, y + corner_sizes[0][1],
edge_sizes[3], h - corner_sizes[0][1] - corner_sizes[3][1],
color);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_blurred_outset_shadow_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRoundedRect *outline = gsk_outset_shadow_node_get_outline (node);
float scale_x = job->scale_x;
float scale_y = job->scale_y;
float blur_radius = gsk_outset_shadow_node_get_blur_radius (node);
float blur_extra = blur_radius * 2.0f; /* 2.0 = shader radius_multiplier */
float half_blur_extra = blur_extra / 2.0f;
ngl: Fix unevenly scaled shadows Ensure that we don't cut them off at the edges.
2021-04-12 02:26:36 +00:00
int extra_blur_pixels_x = ceilf (half_blur_extra * scale_x);
int extra_blur_pixels_y = ceilf (half_blur_extra * scale_y);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
float spread = gsk_outset_shadow_node_get_spread (node);
float dx = gsk_outset_shadow_node_get_dx (node);
float dy = gsk_outset_shadow_node_get_dy (node);
GskRoundedRect scaled_outline;
GskRoundedRect transformed_outline;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
int texture_width, texture_height;
int blurred_texture_id;
int cached_tid;
gboolean do_slicing;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
ngl: Don't slice unsliceable shadows When the corners are too big to make slicing work, don't do it, since it leads to broken results.
2021-04-12 02:09:45 +00:00
float half_width = outline->bounds.size.width / 2;
float half_height = outline->bounds.size.height / 2;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
rgba_to_half (gsk_outset_shadow_node_get_color (node), color);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* scaled_outline is the minimal outline we need to draw the given drop shadow,
* enlarged by the spread and offset by the blur radius. */
scaled_outline = *outline;
if (outline->bounds.size.width < blur_extra ||
ngl: Don't slice unsliceable shadows When the corners are too big to make slicing work, don't do it, since it leads to broken results.
2021-04-12 02:09:45 +00:00
outline->bounds.size.height < blur_extra ||
outline->corner[0].width >= half_width ||
outline->corner[1].width >= half_width ||
outline->corner[2].width >= half_width ||
outline->corner[3].width >= half_width ||
outline->corner[0].height >= half_height ||
outline->corner[1].height >= half_height ||
outline->corner[2].height >= half_height ||
outline->corner[3].height >= half_height)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
do_slicing = FALSE;
gsk_rounded_rect_shrink (&scaled_outline, -spread, -spread, -spread, -spread);
}
else
{
/* Shrink our outline to the minimum size that can still hold all the border radii */
gsk_rounded_rect_shrink_to_minimum (&scaled_outline);
/* Increase by the spread */
gsk_rounded_rect_shrink (&scaled_outline, -spread, -spread, -spread, -spread);
/* Grow bounds but don't grow corners */
graphene_rect_inset (&scaled_outline.bounds, - blur_extra / 2.0, - blur_extra / 2.0);
/* For the center part, we add a few pixels */
scaled_outline.bounds.size.width += SHADOW_EXTRA_SIZE;
scaled_outline.bounds.size.height += SHADOW_EXTRA_SIZE;
do_slicing = TRUE;
}
texture_width = (int)ceil ((scaled_outline.bounds.size.width + blur_extra) * scale_x);
texture_height = (int)ceil ((scaled_outline.bounds.size.height + blur_extra) * scale_y);
ngl: Fix unevenly scaled shadows Ensure that we don't cut them off at the edges.
2021-04-12 02:26:36 +00:00
scaled_outline.bounds.origin.x = extra_blur_pixels_x;
scaled_outline.bounds.origin.y = extra_blur_pixels_y;
scaled_outline.bounds.size.width = texture_width - (extra_blur_pixels_x * 2);
scaled_outline.bounds.size.height = texture_height - (extra_blur_pixels_y * 2);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
for (guint i = 0; i < G_N_ELEMENTS (scaled_outline.corner); i++)
{
scaled_outline.corner[i].width *= scale_x;
scaled_outline.corner[i].height *= scale_y;
}
gsk/gl: use consistent library naming
2022-03-18 21:55:13 +00:00
cached_tid = gsk_gl_shadow_library_lookup (job->driver->shadows_library,
&scaled_outline,
blur_radius);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (cached_tid == 0)
{
GdkGLContext *context = job->command_queue->context;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderTarget *render_target;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
graphene_matrix_t prev_projection;
graphene_rect_t prev_viewport;
guint prev_fbo;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_driver_create_render_target (job->driver,
texture_width, texture_height,
get_target_format (job, node),
&render_target);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (gdk_gl_context_has_debug (context))
{
gdk_gl_context_label_object_printf (context,
GL_TEXTURE,
render_target->texture_id,
"Outset Shadow Temp %d",
render_target->texture_id);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gdk_gl_context_label_object_printf (context,
GL_FRAMEBUFFER,
render_target->framebuffer_id,
"Outset Shadow FB Temp %d",
render_target->framebuffer_id);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
/* Change state for offscreen */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_projection_for_size (job, texture_width, texture_height, &prev_projection);
gsk_gl_render_job_set_viewport_for_size (job, texture_width, texture_height, &prev_viewport);
gsk_gl_render_job_set_modelview (job, NULL);
gsk_gl_render_job_push_clip (job, &scaled_outline);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Bind render target and clear it */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
prev_fbo = gsk_gl_command_queue_bind_framebuffer (job->command_queue, render_target->framebuffer_id);
gsk_gl_command_queue_clear (job->command_queue, 0, &job->viewport);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Draw the outline using color program */
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, color)))
{
gsk_gl_render_job_draw_with_color (job, 0, 0, texture_width, texture_height,
(guint16[]){ FP16_ONE, FP16_ONE, FP16_ONE, FP16_ONE });
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Reset state from offscreen */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_clip (job);
gsk_gl_render_job_pop_modelview (job);
gsk_gl_render_job_set_viewport (job, &prev_viewport, NULL);
gsk_gl_render_job_set_projection (job, &prev_projection);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Now blur the outline */
init_full_texture_region (&offscreen);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
offscreen.texture_id = gsk_gl_driver_release_render_target (job->driver, render_target, FALSE);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
blurred_texture_id = blur_offscreen (job,
&offscreen,
texture_width,
texture_height,
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
blur_radius * fabs (scale_x),
blur_radius * fabs (scale_y));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk/gl: use consistent library naming
2022-03-18 21:55:13 +00:00
gsk_gl_shadow_library_insert (job->driver->shadows_library,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
&scaled_outline,
blur_radius,
blurred_texture_id);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_command_queue_bind_framebuffer (job->command_queue, prev_fbo);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
else
{
blurred_texture_id = cached_tid;
}
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gsk_gl_render_job_translate_rounded_rect (job, outline, &transformed_outline);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (!do_slicing)
{
float min_x = floorf (outline->bounds.origin.x - spread - half_blur_extra + dx);
float min_y = floorf (outline->bounds.origin.y - spread - half_blur_extra + dy);
ngl: Fix a case of flipped shadow In the non-sliced case, we were rendering the shadow upside down.
2021-04-12 02:02:44 +00:00
offscreen.was_offscreen = TRUE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
offscreen.texture_id = blurred_texture_id;
init_full_texture_region (&offscreen);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, outset_shadow)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
blurred_texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform_rounded_rect (job->current_program,
UNIFORM_OUTSET_SHADOW_OUTLINE_RECT, 0,
&transformed_outline);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (min_x,
min_y,
texture_width / scale_x,
texture_height / scale_y),
&offscreen,
color);
gsk_gl_render_job_end_draw (job);
}
return;
}
/* slicing */
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, outset_shadow)))
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
blurred_texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_program_set_uniform_rounded_rect (job->current_program,
UNIFORM_OUTSET_SHADOW_OUTLINE_RECT, 0,
&transformed_outline);
ngl: Small optimization for shadows When 9-slicing shadows, omit the center tile when it is entirely contained in the outline (that is not always the case, depending on corners and offsets).
2021-04-10 15:23:06 +00:00
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
float min_x = floorf (outline->bounds.origin.x - spread - half_blur_extra + dx);
float min_y = floorf (outline->bounds.origin.y - spread - half_blur_extra + dy);
float max_x = ceilf (outline->bounds.origin.x + outline->bounds.size.width +
half_blur_extra + dx + spread);
float max_y = ceilf (outline->bounds.origin.y + outline->bounds.size.height +
half_blur_extra + dy + spread);
const GskGLTextureNineSlice *slices;
float left_width, center_width, right_width;
float top_height, center_height, bottom_height;
GskGLTexture *texture;
texture = gsk_gl_driver_get_texture_by_id (job->driver, blurred_texture_id);
slices = gsk_gl_texture_get_nine_slice (texture, &scaled_outline, extra_blur_pixels_x, extra_blur_pixels_y);
offscreen.was_offscreen = TRUE;
/* Our texture coordinates MUST be scaled, while the actual vertex coords
* MUST NOT be scaled.
*/
left_width = slices[NINE_SLICE_TOP_LEFT].rect.width / scale_x;
right_width = slices[NINE_SLICE_TOP_RIGHT].rect.width / scale_x;
center_width = (max_x - min_x) - (left_width + right_width);
top_height = slices[NINE_SLICE_TOP_LEFT].rect.height / scale_y;
bottom_height = slices[NINE_SLICE_BOTTOM_LEFT].rect.height / scale_y;
center_height = (max_y - min_y) - (top_height + bottom_height);
/* Top left */
if (nine_slice_is_visible (&slices[NINE_SLICE_TOP_LEFT]))
{
memcpy (&offscreen.area, &slices[NINE_SLICE_TOP_LEFT].area, sizeof offscreen.area);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (min_x,
min_y,
left_width,
top_height),
&offscreen,
color);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* Top center */
if (nine_slice_is_visible (&slices[NINE_SLICE_TOP_CENTER]))
{
memcpy (&offscreen.area, &slices[NINE_SLICE_TOP_CENTER].area, sizeof offscreen.area);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (min_x + left_width,
min_y,
center_width,
top_height),
&offscreen,
color);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* Top right */
if (nine_slice_is_visible (&slices[NINE_SLICE_TOP_RIGHT]))
{
memcpy (&offscreen.area, &slices[NINE_SLICE_TOP_RIGHT].area, sizeof offscreen.area);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (max_x - right_width,
min_y,
right_width,
top_height),
&offscreen,
color);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* Bottom right */
if (nine_slice_is_visible (&slices[NINE_SLICE_BOTTOM_RIGHT]))
{
memcpy (&offscreen.area, &slices[NINE_SLICE_BOTTOM_RIGHT].area, sizeof offscreen.area);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (max_x - right_width,
max_y - bottom_height,
right_width,
bottom_height),
&offscreen,
color);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* Bottom left */
if (nine_slice_is_visible (&slices[NINE_SLICE_BOTTOM_LEFT]))
{
memcpy (&offscreen.area, &slices[NINE_SLICE_BOTTOM_LEFT].area, sizeof offscreen.area);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (min_x,
max_y - bottom_height,
left_width,
bottom_height),
&offscreen,
color);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* Left side */
if (nine_slice_is_visible (&slices[NINE_SLICE_LEFT_CENTER]))
{
memcpy (&offscreen.area, &slices[NINE_SLICE_LEFT_CENTER].area, sizeof offscreen.area);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (min_x,
min_y + top_height,
left_width,
center_height),
&offscreen,
color);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* Right side */
if (nine_slice_is_visible (&slices[NINE_SLICE_RIGHT_CENTER]))
{
memcpy (&offscreen.area, &slices[NINE_SLICE_RIGHT_CENTER].area, sizeof offscreen.area);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (max_x - right_width,
min_y + top_height,
right_width,
center_height),
&offscreen,
color);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* Bottom side */
if (nine_slice_is_visible (&slices[NINE_SLICE_BOTTOM_CENTER]))
ngl: Small optimization for shadows When 9-slicing shadows, omit the center tile when it is entirely contained in the outline (that is not always the case, depending on corners and offsets).
2021-04-10 15:23:06 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
memcpy (&offscreen.area, &slices[NINE_SLICE_BOTTOM_CENTER].area, sizeof offscreen.area);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (min_x + left_width,
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
max_y - bottom_height,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
center_width,
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
bottom_height),
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
&offscreen,
color);
ngl: Small optimization for shadows When 9-slicing shadows, omit the center tile when it is entirely contained in the outline (that is not always the case, depending on corners and offsets).
2021-04-10 15:23:06 +00:00
}
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* Middle */
if (nine_slice_is_visible (&slices[NINE_SLICE_CENTER]))
{
if (!gsk_rounded_rect_contains_rect (outline, &GRAPHENE_RECT_INIT (min_x + left_width,
min_y + top_height,
center_width,
center_height)))
{
memcpy (&offscreen.area, &slices[NINE_SLICE_CENTER].area, sizeof offscreen.area);
gsk_gl_render_job_draw_offscreen_with_color (job,
&GRAPHENE_RECT_INIT (min_x + left_width,
min_y + top_height,
center_width,
center_height),
&offscreen,
color);
}
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline gboolean G_GNUC_PURE
equal_texture_nodes (const GskRenderNode *node1,
const GskRenderNode *node2)
{
if (gsk_render_node_get_node_type (node1) != GSK_TEXTURE_NODE ||
gsk_render_node_get_node_type (node2) != GSK_TEXTURE_NODE)
return FALSE;
if (gsk_texture_node_get_texture (node1) !=
gsk_texture_node_get_texture (node2))
return FALSE;
return graphene_rect_equal (&node1->bounds, &node2->bounds);
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_cross_fade_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRenderNode *start_node = gsk_cross_fade_node_get_start_child (node);
const GskRenderNode *end_node = gsk_cross_fade_node_get_end_child (node);
float progress = gsk_cross_fade_node_get_progress (node);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen_start = {0};
GskGLRenderOffscreen offscreen_end = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (progress > 0.0);
g_assert (progress < 1.0);
offscreen_start.force_offscreen = TRUE;
offscreen_start.reset_clip = TRUE;
offscreen_start.bounds = &node->bounds;
offscreen_end.force_offscreen = TRUE;
offscreen_end.reset_clip = TRUE;
offscreen_end.bounds = &node->bounds;
gsk: Better handling of offscreens Take the scale into account for sizing the offscreens for masks, blends and cross-fades.
2023-05-05 15:35:35 +00:00
gsk_gl_render_job_set_modelview (job, gsk_transform_scale (NULL, fabs (job->scale_x), fabs (job->scale_y)));
gsk: Fix cross-fade transform handling We must reset the modelview transform when rendering offscreen, or else we end up applying it twice.
2023-05-01 00:21:47 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, start_node, &offscreen_start))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Fix cross-fade transform handling We must reset the modelview transform when rendering offscreen, or else we end up applying it twice.
2023-05-01 00:21:47 +00:00
gsk_gl_render_job_pop_modelview (job);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, end_node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
}
g_assert (offscreen_start.texture_id);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, end_node, &offscreen_end))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Fix cross-fade transform handling We must reset the modelview transform when rendering offscreen, or else we end up applying it twice.
2023-05-01 00:21:47 +00:00
float prev_alpha;
gsk_gl_render_job_pop_modelview (job);
prev_alpha = gsk_gl_render_job_set_alpha (job, job->alpha * progress);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, start_node);
gsk_gl_render_job_set_alpha (job, prev_alpha);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
}
gsk: Fix cross-fade transform handling We must reset the modelview transform when rendering offscreen, or else we end up applying it twice.
2023-05-01 00:21:47 +00:00
gsk_gl_render_job_pop_modelview (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (offscreen_end.texture_id);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, cross_fade)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen_start.texture_id);
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_CROSS_FADE_SOURCE2, 0,
GL_TEXTURE_2D,
GL_TEXTURE1,
offscreen_end.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform1f (job->current_program,
UNIFORM_CROSS_FADE_PROGRESS, 0,
progress);
gsk_gl_render_job_draw_offscreen (job, &node->bounds, &offscreen_end);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_opacity_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRenderNode *child = gsk_opacity_node_get_child (node);
float opacity = gsk_opacity_node_get_opacity (node);
float new_alpha = job->alpha * opacity;
if (!ALPHA_IS_CLEAR (new_alpha))
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
float prev_alpha = gsk_gl_render_job_set_alpha (job, new_alpha);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk/gl: Use pre-collected opacity information We now collect this information during node construction, so use it here. The concrete change here is that we now avoid offscreens for container nodes with multiple children, as long as they don't overlap. In particular, this avoid offscreens for ellipsized dim labels.
2022-04-03 23:37:42 +00:00
if (!gsk_render_node_use_offscreen_for_opacity (child))
ngl: Fix opacity handling We need to use an offscreen whenever there is overlapping children somewhere in the tree below, just checking the direct child of the opacity node is not enough. Fixes: #4261
2021-09-15 12:26:54 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, child);
gsk_gl_render_job_set_alpha (job, prev_alpha);
ngl: Fix opacity handling We need to use an offscreen whenever there is overlapping children somewhere in the tree below, just checking the direct child of the opacity node is not enough. Fixes: #4261
2021-09-15 12:26:54 +00:00
}
else
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
offscreen.bounds = &child->bounds;
offscreen.force_offscreen = TRUE;
offscreen.reset_clip = TRUE;
ngl: Fix opacity handling We need to use an offscreen whenever there is overlapping children somewhere in the tree below, just checking the direct child of the opacity node is not enough. Fixes: #4261
2021-09-15 12:26:54 +00:00
/* Note: offscreen rendering resets alpha to 1.0 */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, child, &offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
g_assert (offscreen.texture_id);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_offscreen (job, &node->bounds, &offscreen);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_alpha (job, prev_alpha);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
ngl: Speed up glyph loop Move some work out of the loop in visit_text_node. This takes advantage of the fact that the yoffset of most glyphs is zero, so yphase generally does not change in a line of text.
2021-03-29 00:20:50 +00:00
static inline int
compute_phase_and_pos (float value, float *pos)
{
float v;
*pos = floorf (value);
v = value - *pos;
if (v < 0.125)
return 0;
else if (v < 0.375)
return 1;
else if (v < 0.625)
return 2;
else if (v < 0.875)
return 3;
else
{
*pos += 1;
return 0;
}
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_text_node (GskGLRenderJob *job,
const GskRenderNode *node,
const GdkRGBA *color,
gboolean force_color)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const PangoFont *font = gsk_text_node_get_font (node);
const PangoGlyphInfo *glyphs = gsk_text_node_get_glyphs (node, NULL);
const graphene_point_t *offset = gsk_text_node_get_offset (node);
gsk: Avoid a crash with negative scales Don't crash when both scale_x and scale_y are -1. A test is included.
2022-11-20 03:24:20 +00:00
float text_scale = MAX (fabs (job->scale_x), fabs (job->scale_y)); /* TODO: Fix for uneven scales? */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
guint num_glyphs = gsk_text_node_get_num_glyphs (node);
float x = offset->x + job->offset_x;
float y = offset->y + job->offset_y;
gsk/gl: rename glyphs to glyphs_library This naming style is less likely to collide with shader naming and makes it clear where it is consumed what it is.
2022-03-17 23:48:54 +00:00
GskGLGlyphLibrary *library = job->driver->glyphs_library;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLCommandBatch *batch;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
int x_position = 0;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLGlyphKey lookup;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
guint last_texture = 0;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLDrawVertex *vertices;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
guint used = 0;
ngl: Use per-glyph color information Decide per-glyph whether we need color nor not. Fixes: #4141
2021-07-31 15:41:43 +00:00
guint16 nc[4] = { FP16_MINUS_ONE, FP16_MINUS_ONE, FP16_MINUS_ONE, FP16_MINUS_ONE };
guint16 cc[4];
const guint16 *c;
ngl: Cosmetics Remove a local variable in gsk_render_job_visit_text_node.
2021-03-28 20:41:59 +00:00
const PangoGlyphInfo *gi;
guint i;
ngl: Speed up glyph loop Move some work out of the loop in visit_text_node. This takes advantage of the fact that the yoffset of most glyphs is zero, so yphase generally does not change in a line of text.
2021-03-29 00:20:50 +00:00
int yshift;
float ypos;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (num_glyphs == 0)
return;
ngl: Use per-glyph color information Decide per-glyph whether we need color nor not. Fixes: #4141
2021-07-31 15:41:43 +00:00
if ((force_color || !gsk_text_node_has_color_glyphs (node)) &&
RGBA_IS_CLEAR (color))
return;
ngl: Do nothing for transparent text nodes Like the previous commit - a transparent text node will not produce any visible pixels, so bail out early.
2021-07-15 17:17:17 +00:00
ngl: Use per-glyph color information Decide per-glyph whether we need color nor not. Fixes: #4141
2021-07-31 15:41:43 +00:00
rgba_to_half (color, cc);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
lookup.font = (PangoFont *)font;
lookup.scale = (guint) (text_scale * 1024);
ngl: Speed up glyph loop Move some work out of the loop in visit_text_node. This takes advantage of the fact that the yoffset of most glyphs is zero, so yphase generally does not change in a line of text.
2021-03-29 00:20:50 +00:00
yshift = compute_phase_and_pos (y, &ypos);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, coloring)))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
batch = gsk_gl_command_queue_get_batch (job->command_queue);
vertices = gsk_gl_command_queue_add_n_vertices (job->command_queue, num_glyphs);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
/* We use one quad per character */
for (i = 0, gi = glyphs; i < num_glyphs; i++, gi++)
ngl: Speed up glyph loop Move some work out of the loop in visit_text_node. This takes advantage of the fact that the yoffset of most glyphs is zero, so yphase generally does not change in a line of text.
2021-03-29 00:20:50 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
const GskGLGlyphValue *glyph;
float glyph_x, glyph_y, glyph_x2, glyph_y2;
float tx, ty, tx2, ty2;
float cx;
float cy;
guint texture_id;
lookup.glyph = gi->glyph;
/* If the glyph has color, we don't need to recolor anything.
* We tell the shader by setting the color to vec4(-1).
*/
if (!force_color && gi->attr.is_color)
c = nc;
else
c = cc;
cx = (float)(x_position + gi->geometry.x_offset) / PANGO_SCALE;
lookup.xshift = compute_phase_and_pos (x + cx, &cx);
if G_UNLIKELY (gi->geometry.y_offset != 0)
{
cy = (float)(gi->geometry.y_offset) / PANGO_SCALE;
lookup.yshift = compute_phase_and_pos (y + cy, &cy);
}
else
{
lookup.yshift = yshift;
cy = ypos;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
x_position += gi->geometry.width;
ngl: Avoid a goto Reshuffle the code slightly, so we can use a continue instead.
2021-03-29 00:53:46 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
texture_id = gsk_gl_glyph_library_lookup_or_add (library, &lookup, &glyph);
if G_UNLIKELY (texture_id == 0)
continue;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if G_UNLIKELY (last_texture != texture_id || batch->draw.vbo_count + GSK_GL_N_VERTICES > 0xffff)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if G_LIKELY (last_texture != 0)
{
guint vbo_offset = batch->draw.vbo_offset + batch->draw.vbo_count;
/* Since we have batched added our VBO vertices to avoid repeated
* calls to the buffer, we need to manually tweak the vbo offset
* of the new batch as otherwise it will point at the end of our
* vbo array.
*/
gsk_gl_render_job_split_draw (job);
batch = gsk_gl_command_queue_get_batch (job->command_queue);
batch->draw.vbo_offset = vbo_offset;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = TRUE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
last_texture = texture_id;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
tx = glyph->entry.area.x;
ty = glyph->entry.area.y;
tx2 = glyph->entry.area.x2;
ty2 = glyph->entry.area.y2;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
glyph_x = cx + glyph->ink_rect.x;
glyph_y = cy + glyph->ink_rect.y;
glyph_x2 = glyph_x + glyph->ink_rect.width;
glyph_y2 = glyph_y + glyph->ink_rect.height;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
*(vertices++) = (GskGLDrawVertex) { .position = { glyph_x, glyph_y }, .uv = { tx, ty }, .color = { c[0], c[1], c[2], c[3] } };
*(vertices++) = (GskGLDrawVertex) { .position = { glyph_x, glyph_y2 }, .uv = { tx, ty2 }, .color = { c[0], c[1], c[2], c[3] } };
*(vertices++) = (GskGLDrawVertex) { .position = { glyph_x2, glyph_y }, .uv = { tx2, ty }, .color = { c[0], c[1], c[2], c[3] } };
ngl: Rewrite gsk_ngl_render_job_visit_text_node Make this more compact, and thus easier to read.
2021-03-13 19:36:03 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
*(vertices++) = (GskGLDrawVertex) { .position = { glyph_x2, glyph_y2 }, .uv = { tx2, ty2 }, .color = { c[0], c[1], c[2], c[3] } };
*(vertices++) = (GskGLDrawVertex) { .position = { glyph_x, glyph_y2 }, .uv = { tx, ty2 }, .color = { c[0], c[1], c[2], c[3] } };
*(vertices++) = (GskGLDrawVertex) { .position = { glyph_x2, glyph_y }, .uv = { tx2, ty }, .color = { c[0], c[1], c[2], c[3] } };
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
batch->draw.vbo_count += GSK_GL_N_VERTICES;
used++;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (used != num_glyphs)
gsk_gl_command_queue_retract_n_vertices (job->command_queue, num_glyphs - used);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_shadow_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const gsize n_shadows = gsk_shadow_node_get_n_shadows (node);
const GskRenderNode *original_child = gsk_shadow_node_get_child (node);
const GskRenderNode *shadow_child = original_child;
/* Shadow nodes recolor every pixel of the source texture, but leave the alpha in tact.
* If the child is a color matrix node that doesn't touch the alpha, we can throw that away. */
if (gsk_render_node_get_node_type (shadow_child) == GSK_COLOR_MATRIX_NODE &&
!color_matrix_modifies_alpha (shadow_child))
shadow_child = gsk_color_matrix_node_get_child (shadow_child);
for (guint i = 0; i < n_shadows; i++)
{
const GskShadow *shadow = gsk_shadow_node_get_shadow (node, i);
const float dx = shadow->dx;
const float dy = shadow->dy;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
graphene_rect_t bounds;
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 color[4];
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (RGBA_IS_CLEAR (&shadow->color))
continue;
if (node_is_invisible (shadow_child))
continue;
ngl: Improve shortcuts for shadow nodes For shadow nodes, we try pretty hard to avoid rendering shadows, and and we have a shortcut that just renders text offset, but we can try harder to do nothing - if the text is offset by zero, we don't need to draw it at all.
2021-09-16 12:04:09 +00:00
if (shadow->radius == 0 &&
gsk_render_node_get_node_type (shadow_child) == GSK_TEXT_NODE)
{
if (dx != 0 || dy != 0)
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_offset (job, dx, dy);
gsk_gl_render_job_visit_text_node (job, shadow_child, &shadow->color, TRUE);
gsk_gl_render_job_offset (job, -dx, -dy);
ngl: Improve shortcuts for shadow nodes For shadow nodes, we try pretty hard to avoid rendering shadows, and and we have a shortcut that just renders text offset, but we can try harder to do nothing - if the text is offset by zero, we don't need to draw it at all.
2021-09-16 12:04:09 +00:00
}
continue;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (shadow->radius > 0)
{
float min_x;
float min_y;
float max_x;
float max_y;
offscreen.do_not_cache = TRUE;
blur_node (job,
&offscreen,
shadow_child,
shadow->radius,
&min_x, &max_x,
&min_y, &max_y);
bounds.origin.x = min_x - job->offset_x;
bounds.origin.y = min_y - job->offset_y;
bounds.size.width = max_x - min_x;
bounds.size.height = max_y - min_y;
offscreen.was_offscreen = TRUE;
}
else
{
offscreen.bounds = &shadow_child->bounds;
offscreen.reset_clip = TRUE;
offscreen.do_not_cache = TRUE;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, shadow_child, &offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert_not_reached ();
bounds = shadow_child->bounds;
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_offset (job, dx, dy);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, coloring)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
rgba_to_half (&shadow->color, color);
gsk_gl_render_job_draw_offscreen_with_color (job, &bounds, &offscreen, color);
gsk_gl_render_job_end_draw (job);
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_offset (job, -dx, -dy);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
/* Now draw the child normally */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, original_child);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_blur_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRenderNode *child = gsk_blur_node_get_child (node);
float blur_radius = gsk_blur_node_get_radius (node);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
GskTextureKey key;
gboolean cache_texture;
float min_x;
float max_x;
float min_y;
float max_y;
g_assert (blur_radius > 0);
Revert "ngl: Remove a few redundant checks" This reverts commit 4375f8e382ef035d970dd255f56d1a7470c1a78c. These were not redundant after all, as pointed out by Timm.
2021-03-08 13:10:25 +00:00
if (node_is_invisible (child))
return;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
key.pointer = node;
key.pointer_is_child = FALSE;
key.scale_x = job->scale_x;
key.scale_y = job->scale_y;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
offscreen.texture_id = gsk_gl_driver_lookup_texture (job->driver, &key);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
cache_texture = offscreen.texture_id == 0;
blur_node (job,
&offscreen,
child,
blur_radius,
&min_x, &max_x, &min_y, &max_y);
g_assert (offscreen.texture_id != 0);
if (cache_texture)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_driver_cache_texture (job->driver, &key, offscreen.texture_id);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_coords (job,
min_x, min_y, max_x, max_y,
0, 1, 1, 0,
(guint16[]) { FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO } );
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_blend_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRenderNode *top_child = gsk_blend_node_get_top_child (node);
const GskRenderNode *bottom_child = gsk_blend_node_get_bottom_child (node);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen top_offscreen = {0};
GskGLRenderOffscreen bottom_offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
top_offscreen.bounds = &node->bounds;
top_offscreen.force_offscreen = TRUE;
top_offscreen.reset_clip = TRUE;
bottom_offscreen.bounds = &node->bounds;
bottom_offscreen.force_offscreen = TRUE;
bottom_offscreen.reset_clip = TRUE;
gsk: Better handling of offscreens Take the scale into account for sizing the offscreens for masks, blends and cross-fades.
2023-05-05 15:35:35 +00:00
gsk_gl_render_job_set_modelview (job, gsk_transform_scale (NULL, fabs (job->scale_x), fabs (job->scale_y)));
gsk: Fix blend transform handling We must reset the modelview transform when rendering offscreen, or else we end up applying it twice.
2023-05-01 00:27:38 +00:00
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* TODO: We create 2 textures here as big as the blend node, but both the
* start and the end node might be a lot smaller than that. */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, bottom_child, &bottom_offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Fix blend transform handling We must reset the modelview transform when rendering offscreen, or else we end up applying it twice.
2023-05-01 00:27:38 +00:00
gsk_gl_render_job_pop_modelview (job);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, top_child);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
}
g_assert (bottom_offscreen.was_offscreen);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, top_child, &top_offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Fix blend transform handling We must reset the modelview transform when rendering offscreen, or else we end up applying it twice.
2023-05-01 00:27:38 +00:00
gsk_gl_render_job_pop_modelview (job);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
bottom_offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_offscreen (job, &node->bounds, &bottom_offscreen);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
}
g_assert (top_offscreen.was_offscreen);
gsk: Fix blend transform handling We must reset the modelview transform when rendering offscreen, or else we end up applying it twice.
2023-05-01 00:27:38 +00:00
gsk_gl_render_job_pop_modelview (job);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blend)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
bottom_offscreen.texture_id);
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_BLEND_SOURCE2, 0,
GL_TEXTURE_2D,
GL_TEXTURE1,
top_offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform1i (job->current_program,
UNIFORM_BLEND_MODE, 0,
gsk_blend_node_get_blend_mode (node));
gsk_gl_render_job_draw_offscreen_rect (job, &node->bounds);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
gl: Support mask nodes Add a shader for masking.
2023-02-11 15:38:56 +00:00
static inline void
Make mask nodes more versatile Add a GskMaskMode enumeration and implement it in the GL and cairo renderers.
2023-02-14 18:44:39 +00:00
gsk_gl_render_job_visit_mask_node (GskGLRenderJob *job,
const GskRenderNode *node)
gl: Support mask nodes Add a shader for masking.
2023-02-11 15:38:56 +00:00
{
const GskRenderNode *source = gsk_mask_node_get_source (node);
const GskRenderNode *mask = gsk_mask_node_get_mask (node);
GskGLRenderOffscreen source_offscreen = {0};
GskGLRenderOffscreen mask_offscreen = {0};
source_offscreen.bounds = &node->bounds;
source_offscreen.force_offscreen = TRUE;
source_offscreen.reset_clip = TRUE;
mask_offscreen.bounds = &node->bounds;
mask_offscreen.force_offscreen = TRUE;
mask_offscreen.reset_clip = TRUE;
Make mask nodes more versatile Add a GskMaskMode enumeration and implement it in the GL and cairo renderers.
2023-02-14 18:44:39 +00:00
mask_offscreen.do_not_cache = TRUE;
gl: Support mask nodes Add a shader for masking.
2023-02-11 15:38:56 +00:00
gsk: Better handling of offscreens Take the scale into account for sizing the offscreens for masks, blends and cross-fades.
2023-05-05 15:35:35 +00:00
gsk_gl_render_job_set_modelview (job, gsk_transform_scale (NULL, fabs (job->scale_x), fabs (job->scale_y)));
gsk: Fix mask transform handling We must reset the modelview transform when offscreen, or else we end up applying it twice.
2023-05-01 00:40:00 +00:00
gl: Support mask nodes Add a shader for masking.
2023-02-11 15:38:56 +00:00
/* TODO: We create 2 textures here as big as the mask node, but both
* nodes might be a lot smaller than that.
*/
if (!gsk_gl_render_job_visit_node_with_offscreen (job, source, &source_offscreen))
{
gsk: Fix mask transform handling We must reset the modelview transform when offscreen, or else we end up applying it twice.
2023-05-01 00:40:00 +00:00
gsk_gl_render_job_pop_modelview (job);
gl: Support mask nodes Add a shader for masking.
2023-02-11 15:38:56 +00:00
gsk_gl_render_job_visit_node (job, source);
return;
}
g_assert (source_offscreen.was_offscreen);
if (!gsk_gl_render_job_visit_node_with_offscreen (job, mask, &mask_offscreen))
{
gsk: Fix mask transform handling We must reset the modelview transform when offscreen, or else we end up applying it twice.
2023-05-01 00:40:00 +00:00
gsk_gl_render_job_pop_modelview (job);
gl: Support mask nodes Add a shader for masking.
2023-02-11 15:38:56 +00:00
return;
}
g_assert (mask_offscreen.was_offscreen);
gsk: Fix mask transform handling We must reset the modelview transform when offscreen, or else we end up applying it twice.
2023-05-01 00:40:00 +00:00
gsk_gl_render_job_pop_modelview (job);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, mask)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
source_offscreen.texture_id);
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_MASK_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE1,
mask_offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform1i (job->current_program,
UNIFORM_MASK_MODE, 0,
gsk_mask_node_get_mask_mode (node));
gsk_gl_render_job_draw_offscreen_rect (job, &node->bounds);
gsk_gl_render_job_end_draw (job);
}
gl: Support mask nodes Add a shader for masking.
2023-02-11 15:38:56 +00:00
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_color_matrix_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRenderNode *child = gsk_color_matrix_node_get_child (node);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
float offset[4];
Revert "ngl: Remove a few redundant checks" This reverts commit 4375f8e382ef035d970dd255f56d1a7470c1a78c. These were not redundant after all, as pointed out by Timm.
2021-03-08 13:10:25 +00:00
if (node_is_invisible (child))
return;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
offscreen.bounds = &node->bounds;
offscreen.reset_clip = TRUE;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, child, &offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert_not_reached ();
g_assert (offscreen.texture_id > 0);
graphene_vec4_to_float (gsk_color_matrix_node_get_color_offset (node), offset);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, color_matrix)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform_matrix (job->current_program,
UNIFORM_COLOR_MATRIX_COLOR_MATRIX, 0,
gsk_color_matrix_node_get_color_matrix (node));
gsk_gl_program_set_uniform4fv (job->current_program,
UNIFORM_COLOR_MATRIX_COLOR_OFFSET, 0,
1,
offset);
gsk_gl_render_job_draw_offscreen (job, &node->bounds, &offscreen);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_gl_shader_node_fallback (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Move color conversion up Doing the color conversion early lets us reuse the results whenever we do more than one draw per node.
2021-04-07 03:52:33 +00:00
guint16 pink[4] = { 15360, 13975, 14758, 15360 }; /* 255 105 180 */
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, color)))
{
gsk_gl_render_job_draw_rect_with_color (job, &node->bounds, pink);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_gl_shader_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
GError *error = NULL;
GskGLShader *shader;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLProgram *program;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
int n_children;
shader = gsk_gl_shader_node_get_shader (node);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
program = gsk_gl_driver_lookup_shader (job->driver, shader, &error);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
n_children = gsk_gl_shader_node_get_n_children (node);
if G_UNLIKELY (program == NULL)
{
if (g_object_get_data (G_OBJECT (shader), "gsk-did-warn") == NULL)
{
g_object_set_data (G_OBJECT (shader), "gsk-did-warn", GUINT_TO_POINTER (1));
g_warning ("Failed to compile gl shader: %s", error->message);
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_gl_shader_node_fallback (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_clear_error (&error);
}
else
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreens[4] = {{0}};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
const GskGLUniform *uniforms;
const guint8 *base;
GBytes *args;
int n_uniforms;
g_assert (n_children < G_N_ELEMENTS (offscreens));
for (guint i = 0; i < n_children; i++)
{
const GskRenderNode *child = gsk_gl_shader_node_get_child (node, i);
offscreens[i].bounds = &node->bounds;
offscreens[i].force_offscreen = TRUE;
offscreens[i].reset_clip = TRUE;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, child, &offscreens[i]))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
}
args = gsk_gl_shader_node_get_args (node);
base = g_bytes_get_data (args, NULL);
uniforms = gsk_gl_shader_get_uniforms (shader, &n_uniforms);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, program))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
for (guint i = 0; i < n_children; i++)
gsk_gl_program_set_uniform_texture (program,
UNIFORM_CUSTOM_TEXTURE1 + i, 0,
GL_TEXTURE_2D,
GL_TEXTURE0 + i,
offscreens[i].texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform2f (program,
UNIFORM_CUSTOM_SIZE, 0,
node->bounds.size.width,
node->bounds.size.height);
for (guint i = 0; i < n_uniforms; i++)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
const GskGLUniform *u = &uniforms[i];
const guint8 *data = base + u->offset;
switch (u->type)
{
default:
case GSK_GL_UNIFORM_TYPE_NONE:
break;
case GSK_GL_UNIFORM_TYPE_FLOAT:
gsk_gl_uniform_state_set1fv (job->command_queue->uniforms,
program->program_info,
UNIFORM_CUSTOM_ARG0 + i,
0, 1, (const float *)data);
break;
case GSK_GL_UNIFORM_TYPE_INT:
gsk_gl_uniform_state_set1i (job->command_queue->uniforms,
program->program_info,
UNIFORM_CUSTOM_ARG0 + i,
0, *(const gint32 *)data);
break;
case GSK_GL_UNIFORM_TYPE_UINT:
case GSK_GL_UNIFORM_TYPE_BOOL:
gsk_gl_uniform_state_set1ui (job->command_queue->uniforms,
program->program_info,
UNIFORM_CUSTOM_ARG0 + i,
0, *(const guint32 *)data);
break;
case GSK_GL_UNIFORM_TYPE_VEC2:
gsk_gl_uniform_state_set2fv (job->command_queue->uniforms,
program->program_info,
UNIFORM_CUSTOM_ARG0 + i,
0, 1, (const float *)data);
break;
case GSK_GL_UNIFORM_TYPE_VEC3:
gsk_gl_uniform_state_set3fv (job->command_queue->uniforms,
program->program_info,
UNIFORM_CUSTOM_ARG0 + i,
0, 1, (const float *)data);
break;
case GSK_GL_UNIFORM_TYPE_VEC4:
gsk_gl_uniform_state_set4fv (job->command_queue->uniforms,
program->program_info,
UNIFORM_CUSTOM_ARG0 + i,
0, 1, (const float *)data);
break;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_offscreen_rect (job, &node->bounds);
gsk_gl_render_job_end_draw (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
}
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_upload_texture (GskGLRenderJob *job,
GdkTexture *texture,
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
gboolean ensure_mipmap,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen *offscreen)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
GdkGLTexture *gl_texture = NULL;
if (GDK_IS_GL_TEXTURE (texture))
gl_texture = GDK_GL_TEXTURE (texture);
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
if (!ensure_mipmap &&
gl: Pass filters down from visit_texture_node For now, we always pass GL_LINEAR.
2023-02-08 00:29:07 +00:00
gsk_gl_texture_library_can_cache ((GskGLTextureLibrary *)job->driver->icons_library,
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
texture->width,
texture->height) &&
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
!gl_texture)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
const GskGLIconData *icon_data;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk/gl: use consistent library naming
2022-03-18 21:55:13 +00:00
gsk_gl_icon_library_lookup_or_add (job->driver->icons_library, texture, &icon_data);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
offscreen->texture_id = GSK_GL_TEXTURE_ATLAS_ENTRY_TEXTURE (icon_data);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
memcpy (&offscreen->area, &icon_data->entry.area, sizeof offscreen->area);
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
offscreen->has_mipmap = FALSE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
else
{
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
/* Only generate a mipmap if it does not make use reupload
* a GL texture which we could otherwise use directly.
*/
if (gl_texture &&
gdk_gl_context_is_shared (gdk_gl_texture_get_context (gl_texture), job->command_queue->context))
ensure_mipmap = gdk_gl_texture_has_mipmap (gl_texture);
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
offscreen->texture_id = gsk_gl_driver_load_texture (job->driver, texture, ensure_mipmap);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
init_full_texture_region (offscreen);
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
offscreen->has_mipmap = ensure_mipmap;
gsk: Synchronize when using textures Pass the GLsync object from texture into our command queue, and when executing the queue, wait on the sync object the first time we use its associated texture.
2023-01-28 20:20:26 +00:00
if (gl_texture && offscreen->texture_id == gdk_gl_texture_get_id (gl_texture))
offscreen->sync = gdk_gl_texture_get_sync (gl_texture);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
static inline void
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
gsk_gl_render_job_visit_texture (GskGLRenderJob *job,
GdkTexture *texture,
const graphene_rect_t *bounds)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
int max_texture_size = job->command_queue->max_texture_size;
gsk: Cosmetics
2023-03-20 10:57:56 +00:00
float scale_x = bounds->size.width / texture->width;
float scale_y = bounds->size.height / texture->height;
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
gboolean use_mipmap;
gsk: Use mipmaps when it is beneficial If a texture node ends up getting scaled down a lot, it helps to use trilinear filtering.
2023-03-20 01:44:54 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
use_mipmap = (scale_x * fabs (job->scale_x)) < 0.5 ||
(scale_y * fabs (job->scale_y)) < 0.5;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if G_LIKELY (texture->width <= max_texture_size &&
texture->height <= max_texture_size)
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
gsk_gl_render_job_upload_texture (job, texture, use_mipmap, &offscreen);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (offscreen.texture_id);
g_assert (offscreen.was_offscreen == FALSE);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture_with_sync (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id,
offscreen.has_mipmap ? GL_LINEAR_MIPMAP_LINEAR : GL_LINEAR,
GL_LINEAR,
offscreen.sync);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_offscreen (job, bounds, &offscreen);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
else
{
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
float min_x = job->offset_x + bounds->origin.x;
float min_y = job->offset_y + bounds->origin.y;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLTextureSlice *slices = NULL;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
guint n_slices = 0;
gsk: Avoid copying GL texture in more cases When the GL texture already has a mipmap, we don't have to download and reupload it to generate one. We differentiate the handling for texture scale nodes, where we do want to force the mipmap creation even if it requires us to reupload the GL texture, and plain texture nodes, where we just take advantage of a preexisting mipmap to allow trilinear filtering for downscaling, or create one if we have to upload the texture anyway.
2023-03-23 12:47:54 +00:00
gsk_gl_driver_slice_texture (job->driver, texture, use_mipmap, &slices, &n_slices);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (slices != NULL);
g_assert (n_slices > 0);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
for (unsigned int i = 0; i < n_slices; i++)
{
const GskGLTextureSlice *slice = &slices[i];
float x1, x2, y1, y2;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
x1 = min_x + (scale_x * slice->rect.x);
x2 = x1 + (slice->rect.width * scale_x);
y1 = min_y + (scale_y * slice->rect.y);
y2 = y1 + (slice->rect.height * scale_y);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (i > 0)
gsk_gl_render_job_split_draw (job);
gsk_gl_program_set_uniform_texture_with_filter (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
slice->texture_id,
use_mipmap ? GL_LINEAR_MIPMAP_LINEAR : GL_LINEAR,
GL_LINEAR);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_coords (job,
x1, y1, x2, y2,
slice->area.x, slice->area.y,
slice->area.x2, slice->area.y2,
(guint16[]) { FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO });
}
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
static inline void
gsk_gl_render_job_visit_texture_node (GskGLRenderJob *job,
const GskRenderNode *node)
{
GdkTexture *texture = gsk_texture_node_get_texture (node);
const graphene_rect_t *bounds = &node->bounds;
gsk_gl_render_job_visit_texture (job, texture, bounds);
}
static inline void
gsk_gl_render_job_visit_texture_scale_node (GskGLRenderJob *job,
const GskRenderNode *node)
{
GdkTexture *texture = gsk_texture_scale_node_get_texture (node);
const graphene_rect_t *bounds = &node->bounds;
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
GskScalingFilter filter = gsk_texture_scale_node_get_filter (node);
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
int min_filters[] = { GL_LINEAR, GL_NEAREST, GL_LINEAR_MIPMAP_LINEAR };
int mag_filters[] = { GL_LINEAR, GL_NEAREST, GL_LINEAR };
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
int min_filter = min_filters[filter];
int mag_filter = mag_filters[filter];
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
int max_texture_size = job->command_queue->max_texture_size;
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
graphene_rect_t clip_rect;
GskGLRenderTarget *render_target;
graphene_rect_t viewport;
graphene_rect_t prev_viewport;
graphene_matrix_t prev_projection;
float prev_alpha;
guint prev_fbo;
float u0, u1, v0, v1;
gsk: Cache scaled texture Keep the texture we create for texture scale nodes. Fixes: #5642
2023-03-09 11:35:07 +00:00
GskTextureKey key;
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
guint texture_id;
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
if (filter == GSK_SCALING_FILTER_LINEAR)
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
{
gsk_gl_render_job_visit_texture (job, texture, bounds);
return;
}
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
gsk_gl_render_job_untransform_bounds (job, &job->current_clip->rect.bounds, &clip_rect);
if (!graphene_rect_intersection (bounds, &clip_rect, &clip_rect))
return;
gsk: Cache scaled texture Keep the texture we create for texture scale nodes. Fixes: #5642
2023-03-09 11:35:07 +00:00
key.pointer = node;
key.pointer_is_child = TRUE;
key.parent_rect = clip_rect;
key.scale_x = 1.;
key.scale_y = 1.;
texture_id = gsk_gl_driver_lookup_texture (job->driver, &key);
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
gsk: Cache scaled texture Keep the texture we create for texture scale nodes. Fixes: #5642
2023-03-09 11:35:07 +00:00
if (texture_id != 0)
goto render_texture;
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
viewport = GRAPHENE_RECT_INIT (0, 0,
clip_rect.size.width,
clip_rect.size.height);
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
if (!gsk_gl_driver_create_render_target (job->driver,
(int) ceilf (clip_rect.size.width),
(int) ceilf (clip_rect.size.height),
get_target_format (job, node),
&render_target))
gskglrenderjob: Fix compilation with clang
2023-02-14 09:29:25 +00:00
{
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
gsk_gl_render_job_visit_texture (job, texture, bounds);
return;
}
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
gsk_gl_render_job_set_viewport (job, &viewport, &prev_viewport);
gsk_gl_render_job_set_projection_from_rect (job, &viewport, &prev_projection);
gsk_gl_render_job_set_modelview (job, NULL);
prev_alpha = gsk_gl_render_job_set_alpha (job, 1.0f);
gsk_gl_render_job_push_clip (job, &GSK_ROUNDED_RECT_INIT_FROM_RECT (viewport));
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
prev_fbo = gsk_gl_command_queue_bind_framebuffer (job->command_queue, render_target->framebuffer_id);
gsk_gl_command_queue_clear (job->command_queue, 0, &viewport);
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
if G_LIKELY (texture->width <= max_texture_size &&
texture->height <= max_texture_size)
{
gsk: Synchronize when using textures Pass the GLsync object from texture into our command queue, and when executing the queue, wait on the sync object the first time we use its associated texture.
2023-01-28 20:20:26 +00:00
gpointer sync;
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
texture_id = gsk_gl_driver_load_texture (job->driver, texture, filter == GSK_SCALING_FILTER_TRILINEAR);
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
gsk: Synchronize when using textures Pass the GLsync object from texture into our command queue, and when executing the queue, wait on the sync object the first time we use its associated texture.
2023-01-28 20:20:26 +00:00
if (GDK_IS_GL_TEXTURE (texture) && texture_id == gdk_gl_texture_get_id (GDK_GL_TEXTURE (texture)))
sync = gdk_gl_texture_get_sync (GDK_GL_TEXTURE (texture));
else
sync = NULL;
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
u0 = (clip_rect.origin.x - bounds->origin.x) / bounds->size.width;
v0 = (clip_rect.origin.y - bounds->origin.y) / bounds->size.height;
u1 = (clip_rect.origin.x + clip_rect.size.width - bounds->origin.x) / bounds->size.width;
v1 = (clip_rect.origin.y + clip_rect.size.height - bounds->origin.y) / bounds->size.height;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture_with_sync (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
texture_id,
min_filter,
mag_filter,
sync);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_coords (job,
0, 0, clip_rect.size.width, clip_rect.size.height,
u0, v0, u1, v1,
(guint16[]) { FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO });
gsk_gl_render_job_end_draw (job);
}
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
}
else
{
float scale_x = bounds->size.width / texture->width;
float scale_y = bounds->size.height / texture->height;
GskGLTextureSlice *slices = NULL;
guint n_slices = 0;
gsk: Drop some unused arguments All callers of the slice apis pass 0 for min_rows and min_cols. So just drop these.
2023-03-21 02:02:17 +00:00
gsk_gl_driver_slice_texture (job->driver, texture, filter == GSK_SCALING_FILTER_TRILINEAR, &slices, &n_slices);
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
{
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
for (guint i = 0; i < n_slices; i++)
{
const GskGLTextureSlice *slice = &slices[i];
graphene_rect_t slice_bounds;
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
slice_bounds.origin.x = bounds->origin.x - clip_rect.origin.x + slice->rect.x * scale_x;
slice_bounds.origin.y = bounds->origin.y - clip_rect.origin.y + slice->rect.y * scale_y;
slice_bounds.size.width = slice->rect.width * scale_x;
slice_bounds.size.height = slice->rect.height * scale_y;
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
gsk: Add gskrectprivate.h Add a bunch of inline functions for graphene_rectangle_t. We use those quite extensively in tight loops so making them as fast as possible via inlining has massive benefits. The current render-heavy benchmark I am playing (th paris-30k in node-editor) went from 49fps to 85fps on my AMD.
2023-07-20 20:25:20 +00:00
if (!gsk_rect_intersects (&slice_bounds, &viewport))
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
continue;
gsk: Drop texture subsetting We don't do this for unfiltered textures either, and with working slicing, it isn't needed.
2023-03-17 15:25:38 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (i > 0)
gsk_gl_render_job_split_draw (job);
gsk: Drop texture subsetting We don't do this for unfiltered textures either, and with working slicing, it isn't needed.
2023-03-17 15:25:38 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform_texture_with_filter (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
slice->texture_id,
min_filter,
mag_filter);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_coords (job,
slice_bounds.origin.x,
slice_bounds.origin.y,
slice_bounds.origin.x + slice_bounds.size.width,
slice_bounds.origin.y + slice_bounds.size.height,
slice->area.x, slice->area.y,
slice->area.x2, slice->area.y2,
(guint16[]){ FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO } );
}
gsk: Drop texture subsetting We don't do this for unfiltered textures either, and with working slicing, it isn't needed.
2023-03-17 15:25:38 +00:00
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_end_draw (job);
}
gsk: Add slicing to texture scale nodes This makes it possible to use large textures with filtering. Fixes: #5664
2023-03-17 04:50:27 +00:00
}
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
gsk_gl_render_job_pop_clip (job);
gsk_gl_render_job_pop_modelview (job);
gsk_gl_render_job_set_viewport (job, &prev_viewport, NULL);
gsk_gl_render_job_set_projection (job, &prev_projection);
gsk_gl_render_job_set_alpha (job, prev_alpha);
gsk_gl_command_queue_bind_framebuffer (job->command_queue, prev_fbo);
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
gl: Respect clip wehn drawing scale nodes Use the same approach and only create an offscreen that is big enough for the clipped part of the scaled texture. If the clipped part is still too large for a single texture, we give up and just render the texture without filters (using the regular texture rendering code path which supports slicing). The following commit will add the texture-scale-magnify-10000x test which fails without this fix.
2023-02-16 01:10:58 +00:00
texture_id = gsk_gl_driver_release_render_target (job->driver, render_target, FALSE);
gsk: Cache scaled texture Keep the texture we create for texture scale nodes. Fixes: #5642
2023-03-09 11:35:07 +00:00
gsk_gl_driver_cache_texture (job->driver, &key, texture_id);
render_texture:
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_coords (job,
job->offset_x + clip_rect.origin.x,
job->offset_y + clip_rect.origin.y,
job->offset_x + clip_rect.origin.x + clip_rect.size.width,
job->offset_y + clip_rect.origin.y + clip_rect.size.height,
0, clip_rect.size.width / ceilf (clip_rect.size.width),
clip_rect.size.height / ceilf (clip_rect.size.height), 0,
(guint16[]){ FP16_ZERO, FP16_ZERO, FP16_ZERO, FP16_ZERO } );
gsk_gl_render_job_end_draw (job);
}
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
static inline void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_repeat_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const GskRenderNode *child = gsk_repeat_node_get_child (node);
const graphene_rect_t *child_bounds = gsk_repeat_node_get_child_bounds (node);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderOffscreen offscreen = {0};
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (node_is_invisible (child))
return;
if (!graphene_rect_equal (child_bounds, &child->bounds))
{
/* TODO: implement these repeat nodes. */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_as_fallback (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
}
/* If the size of the repeat node is smaller than the size of the
* child node, we don't repeat at all and can just draw that part
* of the child texture... */
gsk: Add gskrectprivate.h Add a bunch of inline functions for graphene_rectangle_t. We use those quite extensively in tight loops so making them as fast as possible via inlining has massive benefits. The current render-heavy benchmark I am playing (th paris-30k in node-editor) went from 49fps to 85fps on my AMD.
2023-07-20 20:25:20 +00:00
if (gsk_rect_contains_rect (child_bounds, &node->bounds))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_clipped_child (job, child, &node->bounds);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
}
offscreen.bounds = &child->bounds;
offscreen.reset_clip = TRUE;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_visit_node_with_offscreen (job, child, &offscreen))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert_not_reached ();
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, repeat)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
offscreen.texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_program_set_uniform4f (job->current_program,
UNIFORM_REPEAT_CHILD_BOUNDS, 0,
(node->bounds.origin.x - child_bounds->origin.x) / child_bounds->size.width,
(node->bounds.origin.y - child_bounds->origin.y) / child_bounds->size.height,
node->bounds.size.width / child_bounds->size.width,
node->bounds.size.height / child_bounds->size.height);
gsk_gl_program_set_uniform4f (job->current_program,
UNIFORM_REPEAT_TEXTURE_RECT, 0,
offscreen.area.x,
offscreen.was_offscreen ? offscreen.area.y2 : offscreen.area.y,
offscreen.area.x2,
offscreen.was_offscreen ? offscreen.area.y : offscreen.area.y2);
gsk_gl_render_job_draw_offscreen (job, &node->bounds, &offscreen);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (GskGLRenderJob *job,
const GskRenderNode *node)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
gboolean has_clip;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert (job != NULL);
g_assert (node != NULL);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
g_assert (GSK_IS_GL_DRIVER (job->driver));
g_assert (GSK_IS_GL_COMMAND_QUEUE (job->command_queue));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
if (node_is_invisible (node))
return;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_render_job_update_clip (job, &node->bounds, &has_clip))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return;
switch (gsk_render_node_get_node_type (node))
{
case GSK_BLEND_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_blend_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_BLUR_NODE:
if (gsk_blur_node_get_radius (node) > 0)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_blur_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
else
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, gsk_blur_node_get_child (node));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_BORDER_NODE:
ngl: Render simple borders simply Use 4 color commands for uniform, rectilinear borders. Most of our borders are rounded with Adwaita, but there are still some of these.
2021-03-09 02:37:28 +00:00
if (gsk_border_node_get_uniform_color (node) &&
gsk_rounded_rect_is_rectilinear (gsk_border_node_get_outline (node)))
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_rect_border_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
else
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_border_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_CLIP_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_clip_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_COLOR_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_color_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_COLOR_MATRIX_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_color_matrix_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_CONIC_GRADIENT_NODE:
if (gsk_conic_gradient_node_get_n_color_stops (node) < MAX_GRADIENT_STOPS)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_conic_gradient_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
else
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_as_fallback (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_CONTAINER_NODE:
{
gsk: Add a private getter for container children No need to get them one-by-one, in our inner loop.
2021-07-24 13:49:09 +00:00
GskRenderNode **children;
guint n_children;
children = gsk_container_node_get_children (node, &n_children);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
for (guint i = 0; i < n_children; i++)
{
gsk: Add a private getter for container children No need to get them one-by-one, in our inner loop.
2021-07-24 13:49:09 +00:00
const GskRenderNode *child = children[i];
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
if (i + 1 < n_children &&
job->current_clip->is_fully_contained &&
gsk_render_node_get_node_type (child) == GSK_ROUNDED_CLIP_NODE)
{
const GskRenderNode *grandchild = gsk_rounded_clip_node_get_child (child);
gsk: Add a private getter for container children No need to get them one-by-one, in our inner loop.
2021-07-24 13:49:09 +00:00
const GskRenderNode *child2 = children[i + 1];
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
if (gsk_render_node_get_node_type (grandchild) == GSK_COLOR_NODE &&
gsk_render_node_get_node_type (child2) == GSK_BORDER_NODE &&
gsk_border_node_get_uniform_color (child2) &&
ngl: Small optimization We really want equality here.
2021-04-11 00:26:45 +00:00
rounded_rect_equal (gsk_rounded_clip_node_get_clip (child),
gsk_border_node_get_outline (child2)))
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_css_background (job, child, child2);
ngl: Special-case css backgrounds Recognize a common pattern: A rounded clip with a color node, followed by a border node, with the same outline. This is what CSS backgrounds frequently produce, and we can render it more efficiently with a combined shader.
2021-03-14 04:30:18 +00:00
i++; /* skip the border node */
continue;
}
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, child);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
}
break;
case GSK_CROSS_FADE_NODE:
{
const GskRenderNode *start_node = gsk_cross_fade_node_get_start_child (node);
const GskRenderNode *end_node = gsk_cross_fade_node_get_end_child (node);
float progress = gsk_cross_fade_node_get_progress (node);
if (progress <= 0.0f)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, gsk_cross_fade_node_get_start_child (node));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
else if (progress >= 1.0f || equal_texture_nodes (start_node, end_node))
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, gsk_cross_fade_node_get_end_child (node));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
else
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_cross_fade_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
break;
case GSK_DEBUG_NODE:
/* Debug nodes are ignored because draws get reordered anyway */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, gsk_debug_node_get_child (node));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_GL_SHADER_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_gl_shader_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_INSET_SHADOW_NODE:
if (gsk_inset_shadow_node_get_blur_radius (node) > 0)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_blurred_inset_shadow_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
else
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_unblurred_inset_shadow_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_LINEAR_GRADIENT_NODE:
case GSK_REPEATING_LINEAR_GRADIENT_NODE:
if (gsk_linear_gradient_node_get_n_color_stops (node) < MAX_GRADIENT_STOPS)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_linear_gradient_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
else
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_as_fallback (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
gsk: Introduce mask nodes Add GskMaskNode, and support it in the render node parser, in the inspector and in GtkSnapshot. The rendering is just fallback for now. Based on old work by Timm Bäder.
2023-02-11 13:59:06 +00:00
case GSK_MASK_NODE:
gl: Support mask nodes Add a shader for masking.
2023-02-11 15:38:56 +00:00
gsk_gl_render_job_visit_mask_node (job, node);
gsk: Introduce mask nodes Add GskMaskNode, and support it in the render node parser, in the inspector and in GtkSnapshot. The rendering is just fallback for now. Based on old work by Timm Bäder.
2023-02-11 13:59:06 +00:00
break;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
case GSK_OPACITY_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_opacity_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_OUTSET_SHADOW_NODE:
if (gsk_outset_shadow_node_get_blur_radius (node) > 0)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_blurred_outset_shadow_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
else
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_unblurred_outset_shadow_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_RADIAL_GRADIENT_NODE:
case GSK_REPEATING_RADIAL_GRADIENT_NODE:
ngl: Treat radial gradients like others Handle colorstop overflow for radial gradients the same way we do for linear and conic gradients.
2021-03-07 15:17:34 +00:00
if (gsk_radial_gradient_node_get_n_color_stops (node) < MAX_GRADIENT_STOPS)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_radial_gradient_node (job, node);
ngl: Treat radial gradients like others Handle colorstop overflow for radial gradients the same way we do for linear and conic gradients.
2021-03-07 15:17:34 +00:00
else
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_as_fallback (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_REPEAT_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_repeat_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_ROUNDED_CLIP_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_rounded_clip_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_SHADOW_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_shadow_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_TEXT_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_text_node (job,
node,
gsk_text_node_get_color (node),
FALSE);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_TEXTURE_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_texture_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
gsk: Introduce GskTextureScaleNode
2023-02-08 03:52:06 +00:00
case GSK_TEXTURE_SCALE_NODE:
gl: Handle GskTextureScaleNodes
2023-02-08 03:58:47 +00:00
gsk_gl_render_job_visit_texture_scale_node (job, node);
gsk: Introduce GskTextureScaleNode
2023-02-08 03:52:06 +00:00
break;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
case GSK_TRANSFORM_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_transform_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
case GSK_CAIRO_NODE:
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_as_fallback (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
break;
gsk: Add GskFillNode Take a rendernode as source and a GskPath and fill the region inside the path with the source, just like cairo_fill() would.
2020-11-10 23:44:32 +00:00
case GSK_FILL_NODE:
gsk_gl_render_job_visit_as_fallback (job, node);
break;
gsk: Add GskStrokeNode Take a rendernode as source and a GskPath and GskStroke, and fill the area that is covered when stroking the path with the given stroke parameters, like cairo_stroke() would.
2020-11-11 06:36:04 +00:00
case GSK_STROKE_NODE:
gsk_gl_render_job_visit_as_fallback (job, node);
break;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
case GSK_NOT_A_RENDER_NODE:
default:
g_assert_not_reached ();
break;
}
ngl: Classify clips Maintain the 'fully contained' information in the clip stack. This will let us pick simpler shaders in the future.
2021-03-07 15:02:46 +00:00
if (has_clip)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_clip (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
static gboolean
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node_with_offscreen (GskGLRenderJob *job,
const GskRenderNode *node,
GskGLRenderOffscreen *offscreen)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
GskTextureKey key;
guint cached_id;
g_assert (job != NULL);
g_assert (node != NULL);
g_assert (offscreen != NULL);
g_assert (offscreen->texture_id == 0);
g_assert (offscreen->bounds != NULL);
if (node_is_invisible (node))
{
/* Just to be safe. */
offscreen->texture_id = 0;
init_full_texture_region (offscreen);
offscreen->was_offscreen = FALSE;
return FALSE;
}
if (gsk_render_node_get_node_type (node) == GSK_TEXTURE_NODE &&
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
!offscreen->force_offscreen)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
GdkTexture *texture = gsk_texture_node_get_texture (node);
glrenderer: Handle filters differently Instead of uploading a texture once per filter, ensure textures are uploaded as little as possible and use samplers instead to switch different filters. Sometimes we have to reupload a texture unfortunately, when it is an external one and we want to create mipmaps.
2023-03-17 05:14:53 +00:00
gsk_gl_render_job_upload_texture (job, texture, FALSE, offscreen);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return TRUE;
}
key.pointer = node;
key.pointer_is_child = TRUE; /* Don't conflict with the child using the cache too */
key.parent_rect = *offscreen->bounds;
key.scale_x = job->scale_x;
key.scale_y = job->scale_y;
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
float offset_x = job->offset_x;
float offset_y = job->offset_y;
gboolean flipped_x = job->scale_x < 0;
gboolean flipped_y = job->scale_y < 0;
graphene_rect_t viewport;
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
gboolean reset_clip = FALSE;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
if (flipped_x || flipped_y)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
GskTransform *transform = gsk_transform_scale (NULL,
flipped_x ? -1 : 1,
flipped_y ? -1 : 1);
gsk_gl_render_job_push_modelview (job, transform);
gsk: Plug a few memory leaks
2023-04-30 22:21:23 +00:00
gsk_transform_unref (transform);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
gsk_gl_render_job_transform_bounds (job, offscreen->bounds, &viewport);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
float aligned_x = floorf (viewport.origin.x);
float padding_left = viewport.origin.x - aligned_x;
float aligned_width = ceilf (viewport.size.width + padding_left);
float padding_right = aligned_width - viewport.size.width - padding_left;
float aligned_y = floorf (viewport.origin.y);
float padding_top = viewport.origin.y - aligned_y;
float aligned_height = ceilf (viewport.size.height + padding_top);
float padding_bottom = aligned_height - viewport.size.height - padding_top;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Tweak the scale factor so that the required texture doesn't
* exceed the max texture limit. This will render with a lower
* resolution, but this is better than clipping.
*/
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
g_assert (job->command_queue->max_texture_size > 0);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
float downscale_x = 1;
float downscale_y = 1;
int texture_width;
int texture_height;
int max_texture_size = job->command_queue->max_texture_size;
if (aligned_width > max_texture_size)
downscale_x = (float)max_texture_size / viewport.size.width;
if (aligned_height > max_texture_size)
downscale_y = (float)max_texture_size / viewport.size.height;
if (downscale_x != 1 || downscale_y != 1)
{
GskTransform *transform = gsk_transform_scale (NULL, downscale_x, downscale_y);
gsk_gl_render_job_push_modelview (job, transform);
gsk: Plug a few memory leaks
2023-04-30 22:21:23 +00:00
gsk_transform_unref (transform);
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
gsk_gl_render_job_transform_bounds (job, offscreen->bounds, &viewport);
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
graphene_rect_scale (&viewport, downscale_x, downscale_y, &viewport);
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
}
if (downscale_x == 1)
{
viewport.origin.x = aligned_x;
viewport.size.width = aligned_width;
offscreen->area.x = padding_left / aligned_width;
offscreen->area.x2 = 1.0f - (padding_right / aligned_width);
texture_width = aligned_width;
}
else
{
offscreen->area.x = 0;
offscreen->area.x2 = 1;
texture_width = max_texture_size;
}
if (downscale_y == 1)
{
viewport.origin.y = aligned_y;
viewport.size.height = aligned_height;
offscreen->area.y = padding_bottom / aligned_height;
offscreen->area.y2 = 1.0f - padding_top / aligned_height;
texture_height = aligned_height;
}
else
{
offscreen->area.y = 0;
offscreen->area.y2 = 1;
texture_height = max_texture_size;
}
/* Check if we've already cached the drawn texture. */
cached_id = gsk_gl_driver_lookup_texture (job->driver, &key);
if (cached_id != 0)
{
if (downscale_x != 1 || downscale_y != 1)
gsk_gl_render_job_pop_modelview (job);
if (flipped_x || flipped_y)
gsk_gl_render_job_pop_modelview (job);
offscreen->texture_id = cached_id;
/* We didn't render it offscreen, but hand out an offscreen texture id */
offscreen->was_offscreen = TRUE;
return TRUE;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderTarget *render_target;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
graphene_matrix_t prev_projection;
graphene_rect_t prev_viewport;
float prev_alpha;
guint prev_fbo;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_driver_create_render_target (job->driver,
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
texture_width, texture_height,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
get_target_format (job, node),
&render_target))
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_assert_not_reached ();
if (gdk_gl_context_has_debug (job->command_queue->context))
{
gdk_gl_context_label_object_printf (job->command_queue->context,
GL_TEXTURE,
render_target->texture_id,
"Offscreen<%s> %d",
g_type_name_from_instance ((GTypeInstance *) node),
render_target->texture_id);
gdk_gl_context_label_object_printf (job->command_queue->context,
GL_FRAMEBUFFER,
render_target->framebuffer_id,
"Offscreen<%s> FB %d",
g_type_name_from_instance ((GTypeInstance *) node),
render_target->framebuffer_id);
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_viewport (job, &viewport, &prev_viewport);
gsk_gl_render_job_set_projection_from_rect (job, &job->viewport, &prev_projection);
prev_alpha = gsk_gl_render_job_set_alpha (job, 1.0f);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
prev_fbo = gsk_gl_command_queue_bind_framebuffer (job->command_queue, render_target->framebuffer_id);
gsk_gl_command_queue_clear (job->command_queue, 0, &job->viewport);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (offscreen->reset_clip)
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
{
gsk_gl_render_job_push_clip (job, &GSK_ROUNDED_RECT_INIT_FROM_RECT (job->viewport));
reset_clip = TRUE;
}
else if (flipped_x || flipped_y || downscale_x != 1 || downscale_y != 1)
{
GskRoundedRect new_clip;
float scale_x = flipped_x ? - downscale_x : downscale_x;
float scale_y = flipped_y ? - downscale_y : downscale_y;
graphene_rect_scale (&job->current_clip->rect.bounds, scale_x, scale_y, &new_clip.bounds);
rounded_rect_scale_corners (&job->current_clip->rect, &new_clip, scale_x, scale_y);
gsk_gl_render_job_push_clip (job, &new_clip);
reset_clip = TRUE;
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, node);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: More fixes Work towards handling negative scales and denormalized rects everywhere.
2023-05-01 17:39:23 +00:00
if (reset_clip)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_clip (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
ngl: Fix offscreen rendering with transforms There was confusion here about the handling of the modelview transform. The modelview transform we are getting is already set up for rendering the node we are given, so keep it - except for possible adding an extra scale on top when the texture would otherwise be too big.
2021-03-31 18:36:24 +00:00
if (downscale_x != 1 || downscale_y != 1)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_pop_modelview (job);
gsk/gl: Always align offscreen rendering with the pixel grid This fixes two issues with the offscreen rendering code for nodes with bounds not aligned with the pixel grid: 1.) When drawing to an offscreen buffer the size of the offscreen buffer was rounded up, but then later when used as texture the vertices correspond to the original bounds with the unrounded size. This could then result in the offscreen texture being drawn onscreen at a slightly smaller size, which then lead to it being visually shifted and blurry. This is fixed by adjusting the u/v coordinates to ignore the padding region in the offscreen texture that got added by the size increase from rounding. 2.) The viewport used when rendering to the offscreen buffer was not aligned with the pixel grid for nodes at coordinates not aligned with the pixel grid. Then because the content of the offscreen buffer is not aligned with the pixel grid and later when used as textures sampling from it will result in interpolated values for an onscreen pixel. This could also result in shifting and blurriness, especially for nested offscreen rendering at different offsets. This is fixed by adding similar padding at the beginning of the texture and also adjusting the u/v coordinates to ignore this region. Fixes: https://gitlab.gnome.org/GNOME/gtk/-/issues/3833
2022-03-31 14:44:34 +00:00
if (flipped_x || flipped_y)
gsk_gl_render_job_pop_modelview (job);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_viewport (job, &prev_viewport, NULL);
gsk_gl_render_job_set_projection (job, &prev_projection);
gsk_gl_render_job_set_alpha (job, prev_alpha);
gsk_gl_command_queue_bind_framebuffer (job->command_queue, prev_fbo);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->offset_x = offset_x;
job->offset_y = offset_y;
offscreen->was_offscreen = TRUE;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
offscreen->texture_id = gsk_gl_driver_release_render_target (job->driver,
render_target,
FALSE);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
if (!offscreen->do_not_cache)
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_driver_cache_texture (job->driver, &key, offscreen->texture_id);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return TRUE;
}
void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_render_flipped (GskGLRenderJob *job,
GskRenderNode *root)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
graphene_matrix_t proj;
guint framebuffer_id;
guint texture_id;
guint surface_height;
g_return_if_fail (job != NULL);
g_return_if_fail (root != NULL);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
g_return_if_fail (GSK_IS_GL_DRIVER (job->driver));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
surface_height = job->viewport.size.height;
graphene_matrix_init_ortho (&proj,
job->viewport.origin.x,
job->viewport.origin.x + job->viewport.size.width,
job->viewport.origin.y,
job->viewport.origin.y + job->viewport.size.height,
ORTHO_NEAR_PLANE,
ORTHO_FAR_PLANE);
graphene_matrix_scale (&proj, 1, -1, 1);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
if (!gsk_gl_command_queue_create_render_target (job->command_queue,
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
MAX (1, job->viewport.size.width),
MAX (1, job->viewport.size.height),
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
job->target_format,
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
&framebuffer_id, &texture_id))
return;
/* Setup drawing to our offscreen texture/framebuffer which is flipped */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_command_queue_bind_framebuffer (job->command_queue, framebuffer_id);
gsk_gl_command_queue_clear (job->command_queue, 0, &job->viewport);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Visit all nodes creating batches */
gdk_gl_context_push_debug_group (job->command_queue->context, "Building command queue");
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, root);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gdk_gl_context_pop_debug_group (job->command_queue->context);
/* Now draw to our real destination, but flipped */
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_alpha (job, 1.0f);
gsk_gl_command_queue_bind_framebuffer (job->command_queue, job->framebuffer);
gsk_gl_command_queue_clear (job->command_queue, 0, &job->viewport);
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
if (gsk_gl_render_job_begin_draw (job, CHOOSE_PROGRAM (job, blit)))
{
gsk_gl_program_set_uniform_texture (job->current_program,
UNIFORM_SHARED_SOURCE, 0,
GL_TEXTURE_2D,
GL_TEXTURE0,
texture_id);
gl renderer: Keep track of source We have an optimization that depends on having the source be a glyph atlas, so keep track of that information in the render job.
2023-09-13 20:55:01 +00:00
job->source_is_glyph_atlas = FALSE;
gsk: Stop setting uniforms when we are out of batches When the command queue is out of batches, there is no point in doing further work like allocating uniforms. This helps us avoid assertions in the uniform code that we would hit when we run out of uniform space too.
2023-05-17 00:03:19 +00:00
gsk_gl_render_job_draw_rect (job, &job->viewport);
gsk_gl_render_job_end_draw (job);
}
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gdk_gl_context_push_debug_group (job->command_queue->context, "Executing command queue");
gsk/gl: support non-standard default framebuffer There are situations where our "default framebuffer" is not actually zero, yet we still want to apply a scissor rect. Generally, 0 is the default framebuffer. But on platforms where we need to bind a platform-specific feature to a GL_FRAMEBUFFER, we might have a default that is not 0. For example, on macOS we bind an IOSurfaceRef to a GL_TEXTURE_RECTANGLE which then is assigned as the backing store for a framebuffer. This is different than using gsk_gl_renderer_render_texture() in that we don't want to incur an extra copy to the destination surface nor do we even have a way to pass a texture_id into render_texture().
2022-02-11 05:51:15 +00:00
gsk_gl_command_queue_execute (job->command_queue, surface_height, 1, NULL, job->default_framebuffer);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gdk_gl_context_pop_debug_group (job->command_queue->context);
glDeleteFramebuffers (1, &framebuffer_id);
glDeleteTextures (1, &texture_id);
}
void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_render (GskGLRenderJob *job,
GskRenderNode *root)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
G_GNUC_UNUSED gint64 start_time;
gsk: Pass scale as float to the command queue
2023-04-02 02:48:38 +00:00
float scale;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
guint surface_height;
g_return_if_fail (job != NULL);
g_return_if_fail (root != NULL);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
g_return_if_fail (GSK_IS_GL_DRIVER (job->driver));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk: Pass scale as float to the command queue
2023-04-02 02:48:38 +00:00
scale = MAX (job->scale_x, job->scale_y);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
surface_height = job->viewport.size.height;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_command_queue_make_current (job->command_queue);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Build the command queue using the shared GL context for all renderers
* on the same display.
*/
start_time = GDK_PROFILER_CURRENT_TIME;
gdk_gl_context_push_debug_group (job->command_queue->context, "Building command queue");
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_command_queue_bind_framebuffer (job->command_queue, job->framebuffer);
gsk/gl: avoid clearing opaque regions If the rendering operation is over an opaque region, we can potentially avoid clearing a large section of the framebuffer destination. Some cases you do want to clear, such as when clearing the whole contents as some drivers have fast paths for that to avoid bringing data back into the framebuffer.
2022-02-22 07:27:42 +00:00
if (job->clear_framebuffer)
gsk_gl_command_queue_clear (job->command_queue, 0, &job->viewport);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_visit_node (job, root);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gdk_gl_context_pop_debug_group (job->command_queue->context);
gdk_profiler_add_mark (start_time, GDK_PROFILER_CURRENT_TIME-start_time, "Build GL command queue", "");
#if 0
/* At this point the atlases have uploaded content while we processed
* nodes but have not necessarily been used by the commands in the queue.
*/
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_driver_save_atlases_to_png (job->driver, NULL);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
#endif
/* But now for executing the command queue, we want to use the context
* that was provided to us when creating the render job as framebuffer 0
* is bound to that context.
*/
start_time = GDK_PROFILER_CURRENT_TIME;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_command_queue_make_current (job->command_queue);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gdk_gl_context_push_debug_group (job->command_queue->context, "Executing command queue");
gsk: Pass scale as float to the command queue
2023-04-02 02:48:38 +00:00
gsk_gl_command_queue_execute (job->command_queue, surface_height, scale, job->region, job->default_framebuffer);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gdk_gl_context_pop_debug_group (job->command_queue->context);
gdk_profiler_add_mark (start_time, GDK_PROFILER_CURRENT_TIME-start_time, "Execute GL command queue", "");
}
void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_debug_fallback (GskGLRenderJob *job,
gboolean debug_fallback)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
g_return_if_fail (job != NULL);
job->debug_fallback = !!debug_fallback;
}
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
static int
gles: Be picky abpout glGetFramebufferAttachmentParameter() This function does not at all work like in OpenGL - if it works at all. So make it behave accordingly. Fixes #4571
2021-12-25 14:06:12 +00:00
get_framebuffer_format (GdkGLContext *context,
guint framebuffer)
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
{
int size;
glcontext: Use GdkGLVersion elsewhere ... and add a convenience API to generate GL versions from strings to make the gdk_gl_context_check() API nicer.
2023-04-25 17:32:12 +00:00
if (!gdk_gl_context_check_version (context, NULL, "3.0"))
gles: Be picky abpout glGetFramebufferAttachmentParameter() This function does not at all work like in OpenGL - if it works at all. So make it behave accordingly. Fixes #4571
2021-12-25 14:06:12 +00:00
return GL_RGBA8;
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
glBindFramebuffer (GL_FRAMEBUFFER, framebuffer);
ngl: fbo 0 has no color attachment Query the backbuffer instead.
2021-10-08 00:28:49 +00:00
glGetFramebufferAttachmentParameteriv (GL_FRAMEBUFFER,
framebuffer ? GL_COLOR_ATTACHMENT0
gles: Be picky abpout glGetFramebufferAttachmentParameter() This function does not at all work like in OpenGL - if it works at all. So make it behave accordingly. Fixes #4571
2021-12-25 14:06:12 +00:00
: gdk_gl_context_get_use_es (context) ? GL_BACK
: GL_BACK_LEFT,
ngl: fbo 0 has no color attachment Query the backbuffer instead.
2021-10-08 00:28:49 +00:00
GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE, &size);
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
ngl: Compute correct renderformat We want to round up, not down.
2021-10-08 00:31:25 +00:00
if (size > 16)
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
return GL_RGBA32F;
ngl: Compute correct renderformat We want to round up, not down.
2021-10-08 00:31:25 +00:00
else if (size > 8)
ngl: Determine intermediate formats Look at the framebuffer and the rendernode to determine what format to use for intermediate textures. Our preference here is to use fp16, if we have it and it makes sense for the framebuffer we're given.
2021-10-01 12:48:43 +00:00
return GL_RGBA16F;
else
return GL_RGBA8;
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderJob *
gsk_gl_render_job_new (GskGLDriver *driver,
const graphene_rect_t *viewport,
gsk: Cosmetics Rename scale_factor to scale in various places, now that it is no longer an int but a float.
2023-04-02 02:54:19 +00:00
float scale,
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
const cairo_region_t *region,
gsk/gl: avoid clearing opaque regions If the rendering operation is over an opaque region, we can potentially avoid clearing a large section of the framebuffer destination. Some cases you do want to clear, such as when clearing the whole contents as some drivers have fast paths for that to avoid bringing data back into the framebuffer.
2022-02-22 07:27:42 +00:00
guint framebuffer,
gboolean clear_framebuffer)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
const graphene_rect_t *clip_rect = viewport;
graphene_rect_t transformed_extents;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderJob *job;
gsk/gl: support non-standard default framebuffer There are situations where our "default framebuffer" is not actually zero, yet we still want to apply a scissor rect. Generally, 0 is the default framebuffer. But on platforms where we need to bind a platform-specific feature to a GL_FRAMEBUFFER, we might have a default that is not 0. For example, on macOS we bind an IOSurfaceRef to a GL_TEXTURE_RECTANGLE which then is assigned as the backing store for a framebuffer. This is different than using gsk_gl_renderer_render_texture() in that we don't want to incur an extra copy to the destination surface nor do we even have a way to pass a texture_id into render_texture().
2022-02-11 05:51:15 +00:00
GdkGLContext *context;
GLint default_framebuffer = 0;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
g_return_val_if_fail (GSK_IS_GL_DRIVER (driver), NULL);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_return_val_if_fail (viewport != NULL, NULL);
gsk: Cosmetics Rename scale_factor to scale in various places, now that it is no longer an int but a float.
2023-04-02 02:54:19 +00:00
g_return_val_if_fail (scale > 0, NULL);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
gsk/gl: support non-standard default framebuffer There are situations where our "default framebuffer" is not actually zero, yet we still want to apply a scissor rect. Generally, 0 is the default framebuffer. But on platforms where we need to bind a platform-specific feature to a GL_FRAMEBUFFER, we might have a default that is not 0. For example, on macOS we bind an IOSurfaceRef to a GL_TEXTURE_RECTANGLE which then is assigned as the backing store for a framebuffer. This is different than using gsk_gl_renderer_render_texture() in that we don't want to incur an extra copy to the destination surface nor do we even have a way to pass a texture_id into render_texture().
2022-02-11 05:51:15 +00:00
/* Check for non-standard framebuffer binding as we might not be using
* the default framebuffer on systems like macOS where we've bound an
* IOSurface to a GL_TEXTURE_RECTANGLE. Otherwise, no scissor clip will
* be applied in the command queue causing overdrawing.
*/
context = driver->command_queue->context;
default_framebuffer = GDK_GL_CONTEXT_GET_CLASS (context)->get_default_framebuffer (context);
if (framebuffer == 0 && default_framebuffer != 0)
framebuffer = default_framebuffer;
gsk: Stop using g_slice
2023-03-03 11:48:49 +00:00
job = g_new0 (GskGLRenderJob, 1);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->driver = g_object_ref (driver);
job->command_queue = job->driver->command_queue;
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
job->clip = g_array_sized_new (FALSE, FALSE, sizeof (GskGLRenderClip), 16);
job->modelview = g_array_sized_new (FALSE, FALSE, sizeof (GskGLRenderModelview), 16);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->framebuffer = framebuffer;
gsk/gl: avoid clearing opaque regions If the rendering operation is over an opaque region, we can potentially avoid clearing a large section of the framebuffer destination. Some cases you do want to clear, such as when clearing the whole contents as some drivers have fast paths for that to avoid bringing data back into the framebuffer.
2022-02-22 07:27:42 +00:00
job->clear_framebuffer = !!clear_framebuffer;
gsk/gl: support non-standard default framebuffer There are situations where our "default framebuffer" is not actually zero, yet we still want to apply a scissor rect. Generally, 0 is the default framebuffer. But on platforms where we need to bind a platform-specific feature to a GL_FRAMEBUFFER, we might have a default that is not 0. For example, on macOS we bind an IOSurfaceRef to a GL_TEXTURE_RECTANGLE which then is assigned as the backing store for a framebuffer. This is different than using gsk_gl_renderer_render_texture() in that we don't want to incur an extra copy to the destination surface nor do we even have a way to pass a texture_id into render_texture().
2022-02-11 05:51:15 +00:00
job->default_framebuffer = default_framebuffer;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->offset_x = 0;
job->offset_y = 0;
gsk: Cosmetics Rename scale_factor to scale in various places, now that it is no longer an int but a float.
2023-04-02 02:54:19 +00:00
job->scale_x = scale;
job->scale_y = scale;
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
job->viewport = *viewport;
gles: Be picky abpout glGetFramebufferAttachmentParameter() This function does not at all work like in OpenGL - if it works at all. So make it behave accordingly. Fixes #4571
2021-12-25 14:06:12 +00:00
job->target_format = get_framebuffer_format (job->command_queue->context, framebuffer);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_set_alpha (job, 1.0f);
gsk_gl_render_job_set_projection_from_rect (job, viewport, NULL);
gsk: Cosmetics Rename scale_factor to scale in various places, now that it is no longer an int but a float.
2023-04-02 02:54:19 +00:00
gsk_gl_render_job_set_modelview (job, gsk_transform_scale (NULL, scale, scale));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
/* Setup our initial clip. If region is NULL then we are drawing the
* whole viewport. Otherwise, we need to convert the region to a
* bounding box and clip based on that.
*/
if (region != NULL)
{
cairo_rectangle_int_t extents;
cairo_region_get_extents (region, &extents);
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_transform_bounds (job,
&GRAPHENE_RECT_INIT (extents.x,
extents.y,
extents.width,
extents.height),
&transformed_extents);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
clip_rect = &transformed_extents;
job->region = cairo_region_create_rectangle (&extents);
}
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_push_clip (job,
&GSK_ROUNDED_RECT_INIT (clip_rect->origin.x,
clip_rect->origin.y,
clip_rect->size.width,
clip_rect->size.height));
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
return job;
}
void
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
gsk_gl_render_job_free (GskGLRenderJob *job)
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
{
job->current_modelview = NULL;
job->current_clip = NULL;
while (job->modelview->len > 0)
{
Rename ngl to gl We have only one gl renderer now, and it is a bit odd for it not be called gl.
2021-10-07 03:15:25 +00:00
GskGLRenderModelview *modelview = &g_array_index (job->modelview, GskGLRenderModelview, job->modelview->len-1);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
g_clear_pointer (&modelview->transform, gsk_transform_unref);
job->modelview->len--;
}
g_clear_object (&job->driver);
g_clear_pointer (&job->region, cairo_region_destroy);
g_clear_pointer (&job->modelview, g_array_unref);
g_clear_pointer (&job->clip, g_array_unref);
gsk: Stop using g_slice
2023-03-03 11:48:49 +00:00
g_free (job);
gsk: add OpenGL based GskNglRenderer The primary goal here was to cleanup the current GL renderer to make maintenance easier going forward. Furthermore, it tracks state to allow us to implement more advanced renderer features going forward. Reordering This renderer will reorder batches by render target to reduce the number of times render targets are changed. In the future, we could also reorder by program within the render target if we can determine that vertices do not overlap. Uniform Snapshots To allow for reordering of batches all uniforms need to be tracked for the programs. This allows us to create the full uniform state when the batch has been moved into a new position. Some care was taken as it can be performance sensitive. Attachment Snapshots Similar to uniform snapshots, we need to know all of the texture attachments so that we can rebind them when necessary. Render Jobs To help isolate the process of creating GL commands from the renderer abstraction a render job abstraction was added. This could be extended in the future if we decided to do tiling. Command Queue Render jobs create batches using the command queue. The command queue will snapshot uniform and attachment state so that it can reorder batches right before executing them. Currently, the only reordering done is to ensure that we only visit each render target once. We could extend this by tracking vertices, attachments, and others. This code currently uses an inline array helper to reduce overhead from GArray which was showing up on profiles. It could be changed to use GdkArray without too much work, but had roughly double the instructions. Cycle counts have not yet been determined. GLSL Programs This was simplified to use XMACROS so that we can just extend one file (gskglprograms.defs) instead of multiple places. The programs are added as fields in the driver for easy access. Driver The driver manages textures, render targets, access to atlases, programs, and more. There is one driver per display, by using the shared GL context. Some work could be done here to batch uploads so that we make fewer calls to upload when sending icon theme data to the GPU. We'd need to keep a copy of the atlas data for such purposes.
2020-12-19 01:36:59 +00:00
}
Reference in New Issue Copy Permalink