Previously, we relied on ops to deduce whether a draw would have
hardware coverage modulation as a result mixed samples. This is
problematic because *any* draw can have mixed samples coverage if
there is a multisampled stencil clip. No ops were checking for stencil
clip, and most just said they never used mixed samples.
Now that the only usecase for mixed samples is the stencil buffer,
this CL makes the processorSet automatically deduce mixed samples
coverage from the stencil settings and fsaaType.
Bug: skia:
Change-Id: Ib69b84bc03b12f6efb8e7d6ed721ae1612785315
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/197281
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
This is a reland of dfe5000a5f
Original change's description:
> Stack-allocate pipelines for GrMeshDrawOp
>
> Stack-allocates the pipelines in onExecute. This saves us from having
> to store the pipelines on the heap, as well as delaying the need to
> detach processors until onExecute. The delay is an improvement because
> it allows us to keep visiting proxies after onPrepare. (Previously,
> they were moved out of GrProcessorSet and into a pipeline during
> onPrepare, so visiting proxies was impossible after that point.)
>
> Bug: skia:8731
> Change-Id: Idc05063fb0dfbfed42b434e429fa5a497097bdae
> Reviewed-on: https://skia-review.googlesource.com/c/193368
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
TBR=robertphillips@google.com
Bug: skia:8731
Change-Id: I32def1a35bb0593470fa672691a9e697dc6d9680
Reviewed-on: https://skia-review.googlesource.com/c/195261
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
This reverts commit dfe5000a5f.
Reason for revert: HWAA pipeline flag not getting set for dashing.
Original change's description:
> Stack-allocate pipelines for GrMeshDrawOp
>
> Stack-allocates the pipelines in onExecute. This saves us from having
> to store the pipelines on the heap, as well as delaying the need to
> detach processors until onExecute. The delay is an improvement because
> it allows us to keep visiting proxies after onPrepare. (Previously,
> they were moved out of GrProcessorSet and into a pipeline during
> onPrepare, so visiting proxies was impossible after that point.)
>
> Bug: skia:8731
> Change-Id: Idc05063fb0dfbfed42b434e429fa5a497097bdae
> Reviewed-on: https://skia-review.googlesource.com/c/193368
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
TBR=robertphillips@google.com,csmartdalton@google.com
Change-Id: If706f19423310846de70288f393ac12f17ffeee5
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:8731
Reviewed-on: https://skia-review.googlesource.com/c/195161
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Stack-allocates the pipelines in onExecute. This saves us from having
to store the pipelines on the heap, as well as delaying the need to
detach processors until onExecute. The delay is an improvement because
it allows us to keep visiting proxies after onPrepare. (Previously,
they were moved out of GrProcessorSet and into a pipeline during
onPrepare, so visiting proxies was impossible after that point.)
Bug: skia:8731
Change-Id: Idc05063fb0dfbfed42b434e429fa5a497097bdae
Reviewed-on: https://skia-review.googlesource.com/c/193368
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This is a reland of f5efa78902
Original change's description:
> Add a GpuGM class for GPU-only tests
>
> Bug: skia:8731
> Change-Id: Ic81f21621c0183bb356601eb663fa1ee334bca55
> Reviewed-on: https://skia-review.googlesource.com/c/189134
> Reviewed-by: Brian Osman <brianosman@google.com>
> Reviewed-by: Mike Klein <mtklein@google.com>
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
TBR=mtklein@google.com
Bug: skia:8731
Change-Id: Ia322d7e9280b67ebea1edbefc820cae72cd9c3a6
Reviewed-on: https://skia-review.googlesource.com/c/190148
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Bug: skia:8731
Change-Id: Ic81f21621c0183bb356601eb663fa1ee334bca55
Reviewed-on: https://skia-review.googlesource.com/c/189134
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Mechanical. This makes the priv() accessor the same for all the context types.
Change-Id: I40850eb05a33b8d7cc3eabdd42226d24b2ba58aa
Reviewed-on: https://skia-review.googlesource.com/c/189164
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
The function is inlined - the compiler can do whatever optimizations it
wants at each call-site. In the future, stride and uvOffset won't be
static (with optional wide colors), so this is required anyway.
Also, the pointer being passed in is never const, so stop cleverly
casting away the const-ness via reinterpret into and out of intptr_t.
Bug: skia:
Change-Id: I404c418a01d70ac77667ba89338940b086922dc6
Reviewed-on: https://skia-review.googlesource.com/c/179988
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Adds setVertexAttributes and setInstanceAttributes. These take a pointer
to the first attribute, and a count. The count is the total number of
possible attributes, though some may not be initialized. The base class
computes the number of initialized attributes, pre-computes the strides,
and only allows subsequent access to the initialized attributes.
The attributes need to be allocated contiguously. Some GPs place them in
an array, though most just place them as consecutive members, and pass
a pointer to the first one.
Indexed access would be possible, but now it makes more sense to iterate
over all attributes, so enable that, and use range-based for everywhere.
Completely remove the per-attribute offset helper (again - possible, but
not real helpful), and make the stride always available. In many ops,
just use the GP's computed stride, rather than re-computing it.
Bug: skia:
Change-Id: Ie4cccb7969a98ee5a10b373e714fbd702e875b3e
Reviewed-on: https://skia-review.googlesource.com/c/169241
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:
Change-Id: I47e4b1e96905e27aed78cc317d0b7f263b8cf25a
Reviewed-on: https://skia-review.googlesource.com/c/166760
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
- Mechanical replacement of GrColor4h (used throughout Ops) with
SkPMColor4f.
- API adaptation (to/FromGrColor -> to/From_BytesRGBA).
- Complete removal of FromFloats (source was already SkPMColor4f),
and toFloats (setting uniforms can directly access .vec()).
Bug: skia:
Change-Id: I37eece1fa7ed2545dc6843e840d4cc3c60f19747
Reviewed-on: https://skia-review.googlesource.com/c/166620
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:
Change-Id: I8486d0631e4723f1a457a607bc21abd2581133d8
Reviewed-on: https://skia-review.googlesource.com/c/166562
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
We still degrade to bytes when creating vertices (tagged TODO4F).
Note: Guarded for Chrome (by making GrColor4h a wrapper around
GrColor).
Bug: skia:
Change-Id: Id8a1d9eec7978d52b059cd9952666bc1217ee073
Reviewed-on: https://skia-review.googlesource.com/c/165527
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
are not recyclable.
Change-Id: I5ff9c48b3d7b37531a3f052bd5188a8afcacf3cb
Reviewed-on: https://skia-review.googlesource.com/c/161860
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Centralize the calculation to SkGlyphCacheCommon. This allows this
function to be used with NO_GPU.
In addition, this was used in the last remaining function in GrTest.
That function was used in a single place. I made the function a static
and remove GrTest.h. This had massive knock-on effects.
Change-Id: I80f874a988f9af4383a83acf7c273d23b8d67c22
Reviewed-on: https://skia-review.googlesource.com/151480
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Herb Derby <herb@google.com>
This reverts commit fdf05f4ff4.
Reason for revert: Android fixed after removing multitexture support from TextureOp.
Original change's description:
> Revert "Move GrGeometryProcessor's textures out of classes and into"
>
> This reverts commit af87483873.
>
> Revert "GrGeometryProcessor derives from GrNonAtomicRef not GrProgramElement."
>
> This reverts commit 607be37e3d.
>
> Revert "Store GrMeshDrawOps' meshes in GrOpFlushState's arena."
>
> This reverts commit b948572c78.
>
> Revert "Remove multitexturing support from GrTextureOp."
>
> This reverts commit 986f64c601.
>
> Revert "Make result of GrOp::combineIfPossible be an enum."
>
> This reverts commit 641ac7daa8.
>
> Bug: b/112244393
> Change-Id: I579491a3f2f2f2093f1e2a6141fa1e4cc7b760a4
> Reviewed-on: https://skia-review.googlesource.com/145646
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
TBR=bsalomon@google.com
Change-Id: I1d41c2ecf7862e31fb025a7a00bb07bae9d83a47
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: b/112244393
Reviewed-on: https://skia-review.googlesource.com/145780
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit af87483873.
Revert "GrGeometryProcessor derives from GrNonAtomicRef not GrProgramElement."
This reverts commit 607be37e3d.
Revert "Store GrMeshDrawOps' meshes in GrOpFlushState's arena."
This reverts commit b948572c78.
Revert "Remove multitexturing support from GrTextureOp."
This reverts commit 986f64c601.
Revert "Make result of GrOp::combineIfPossible be an enum."
This reverts commit 641ac7daa8.
Bug: b/112244393
Change-Id: I579491a3f2f2f2093f1e2a6141fa1e4cc7b760a4
Reviewed-on: https://skia-review.googlesource.com/145646
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This is to prepare for a third value that requests that ops be linked
together so that the first op may do the work for multiple linked ops
without actually merging the GrOp objects.
Change-Id: Ib6e012a89be5edd054aee69d8475bea612331852
Reviewed-on: https://skia-review.googlesource.com/145522
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: Brian Salomon <bsalomon@google.com>
Remove late draw consolidation in GrOpFlushState. Rarely did anything
and doesn't work with new allocation strategy. Ops can use GrMesh arrays
to acheive the same thing. (Each Op that cared to would have to implement
but it isn't applicable to most Ops).
Modify GrMeshDrawOp::Target::draw() to take array of meshes, with single
mesh as a special case.
Change-Id: I552677de47b9ffd2fcaf55af85f70f290e5aa9c7
Reviewed-on: https://skia-review.googlesource.com/145426
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
It probably doesn't need to be ref counted at all and should be stored
in GrOpFlushState's arena but that's a larger change for another day.
Change-Id: I5f593fb426b8e7794f2ca81194f5a8d9e0f0a072
Reviewed-on: https://skia-review.googlesource.com/145332
Auto-Submit: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Pins out-of-range solutions for cubic/quad max curvature instead of
throwing them out. Code that wants to know the endpoint(s) closest to
max curvature now has the information. Code not interested in these
values can just ignore 0 and 1.
Bug: skia:
Change-Id: I8e7e2ef236b4ab963865dc049ac3e09d5396757d
Reviewed-on: https://skia-review.googlesource.com/143041
Reviewed-by: Cary Clark <caryclark@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
The most common use-case for F16 is to draw things that are explicitly
outside of [0,1]. We can't prevent out-of-range values in general, users
could just draw out-of-range content in src mode. We've decided that
it's best to avoid surprising behavior, and trust the user in these
situations.
This effectively reverts https://skia-review.googlesource.com/c/skia/+/48183
Bug: skia:
Change-Id: I71c2dd925c567e2862d96ab70c752f19245a69c4
Reviewed-on: https://skia-review.googlesource.com/140500
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Remove scissor rect from GrPipeline.
Draws can specify "fixed dynamic state" which doesn't use the dynamism at
all or can specify dynamic state arrays with an entry per GrMesh.
When we state other than scissor rects this will allow the caller to
use a mix of truly dynamic and fixed dynamic state. So a caller that
only has dynamic scissor rects doesn't need to store its remaining
unvarying state in an array.
Change-Id: I8fcc07eb600c72a26cc712b185755c2116021a8a
Reviewed-on: https://skia-review.googlesource.com/137223
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit 19c1233c44.
Reason for revert: want to make sure Google3 can roll
Original change's description:
> Change how vertex/instance attributes are handled in geometry processors.
>
> * No longer register vertex/instance attributes on base class, just counts
>
> * Separate instance and vertex attributes and remove InputRate and offset
>
> * Make attributes constexpr where possible
>
> Change-Id: I1f1d5e772fa177a96d2aeb805aab7b69f35bfae6
> Reviewed-on: https://skia-review.googlesource.com/132405
> Commit-Queue: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Chris Dalton <csmartdalton@google.com>
TBR=egdaniel@google.com,bsalomon@google.com,csmartdalton@google.com
Change-Id: I4800632515e14fbf54af52826928ac915657b59f
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/135661
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
* No longer register vertex/instance attributes on base class, just counts
* Separate instance and vertex attributes and remove InputRate and offset
* Make attributes constexpr where possible
Change-Id: I1f1d5e772fa177a96d2aeb805aab7b69f35bfae6
Reviewed-on: https://skia-review.googlesource.com/132405
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Chris Dalton <csmartdalton@google.com>
There is no good way to split the monster CL up. This breaks out
the GrContext plumbing but doesn't use it.
Change-Id: I90856d428d372bcec3f8821e6364667b367927d4
Reviewed-on: https://skia-review.googlesource.com/133382
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Most of this is (obviously) not necessary to do, but once
I started, I figured I'd just get it all. Tools (nanobench,
DM, skiaserve), all GMs, benches, and unit tests, plus support
code (command line parsing and config stuff).
This is almost entirely mechanical.
Bug: skia:
Change-Id: I209500f8df8c5bd43f8298ff26440d1c4d7425fb
Reviewed-on: https://skia-review.googlesource.com/131153
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Moves getCaps() from GrContext to GrContextPriv and removes unused refCaps().
Change-Id: Ic6a8951b656c0d1b2773eae73bff8e88af819866
Reviewed-on: https://skia-review.googlesource.com/127389
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Otherwise the cubic math is too unstable.
Bug: skia:
Change-Id: I2c8d8d9abcae517d42dd3eddb5bf455c7c7487dd
Reviewed-on: https://skia-review.googlesource.com/121709
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Moved method are not used by chromium, google3,
or android.
SkPoint::setRectIFan isn't used or tested at all.
SkPoint::setRectFan and SkPoint::setRectTriStrip
are only used internally.
These routines pretend that a SkPoint is part
of an array of points. Since that's kind of an
odd contract to make public, and because they
aren't used outside of Skia, relegate them to
a priv file.
R=bsalomon@google.com,reed@google.com
Bug: skia: 6898
Change-Id: I5ec2eb47799f6fd4b2994da962b1fa69ce659931
Reviewed-on: https://skia-review.googlesource.com/68121
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>
SkMatrix::mapHomogeneousPoints takes an array of SkScalar,
but expects essentially SkPoint3, so make it so.
R: robertphillips@google.com, reed@google.com
Bug: skia:6898
Change-Id: Ibaf8b05c08b7df16c67d6a77d914667ace9a70da
Reviewed-on: https://skia-review.googlesource.com/58380
Commit-Queue: Cary Clark <caryclark@skia.org>
Commit-Queue: Cary Clark <caryclark@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This reverts commit 62563deb6b.
Reason for revert: change that affected similar set of GMs reverted, relanding now that this is more easily triagable.
Original change's description:
> Revert "Make GPU backend triangulate rects such that they are rendered as tri strips rather than tri fans."
>
> This reverts commit fa2d604a7d.
>
> Reason for revert: <INSERT REASONING HERE>
>
> Original change's description:
> > Make GPU backend triangulate rects such that they are rendered as tri strips rather than tri fans.
> >
> > Right now when we turn rects into quads we use a vertex order compatible with a tri fan rather than a tri strip.
> >
> > I wanted it to be the case that the same code could be used to generate a non-indexed mesh for a single rect or indexed using the quad index buffer when batching. Triangle fanning is not available in all APIS (e.g. is emulated in ANGLE and not supported in Metal) so it seems better to use a triangle strip over a fan in the single rect case.
> >
> >
> > Change-Id: I31eebd794e7328f4b39e3ec3377bf2ec556360ca
> > Reviewed-on: https://skia-review.googlesource.com/60081
> > Commit-Queue: Brian Salomon <bsalomon@google.com>
> > Reviewed-by: Robert Phillips <robertphillips@google.com>
>
> TBR=bsalomon@google.com,robertphillips@google.com
>
> Change-Id: I7c4c23aa418da09c9708b28cce64ab58e282dd3a
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://skia-review.googlesource.com/60683
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
TBR=bsalomon@google.com,robertphillips@google.com
Change-Id: Iefcd16676a7617d32e89fc84206cd4e88e9a06e1
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/61160
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit fa2d604a7d.
Reason for revert: <INSERT REASONING HERE>
Original change's description:
> Make GPU backend triangulate rects such that they are rendered as tri strips rather than tri fans.
>
> Right now when we turn rects into quads we use a vertex order compatible with a tri fan rather than a tri strip.
>
> I wanted it to be the case that the same code could be used to generate a non-indexed mesh for a single rect or indexed using the quad index buffer when batching. Triangle fanning is not available in all APIS (e.g. is emulated in ANGLE and not supported in Metal) so it seems better to use a triangle strip over a fan in the single rect case.
>
>
> Change-Id: I31eebd794e7328f4b39e3ec3377bf2ec556360ca
> Reviewed-on: https://skia-review.googlesource.com/60081
> Commit-Queue: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
TBR=bsalomon@google.com,robertphillips@google.com
Change-Id: I7c4c23aa418da09c9708b28cce64ab58e282dd3a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/60683
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Right now when we turn rects into quads we use a vertex order compatible with a tri fan rather than a tri strip.
I wanted it to be the case that the same code could be used to generate a non-indexed mesh for a single rect or indexed using the quad index buffer when batching. Triangle fanning is not available in all APIS (e.g. is emulated in ANGLE and not supported in Metal) so it seems better to use a triangle strip over a fan in the single rect case.
Change-Id: I31eebd794e7328f4b39e3ec3377bf2ec556360ca
Reviewed-on: https://skia-review.googlesource.com/60081
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Bug: skia:6173
Change-Id: I21042d484d9a7b3eee04aa3301d9793d00ad6908
Reviewed-on: https://skia-review.googlesource.com/48183
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
There is a perf regression (mainly on the Nexus5) for the https://skia-review.googlesource.com/c/skia/+/46200 (Add method to iterate over a GrOp's GrSurfaceProxies)
This is one candidate.
Change-Id: I995d3a88bad2a914f24b49512abbf01aeaf579c8
Reviewed-on: https://skia-review.googlesource.com/46586
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
The extra generality of having a std::function is for MDB reordering. In the current MDB reordering world there is one pass through the surfaceProxies at creation time and a second pass after flush to create the usage intervals.
Change-Id: I3f548417eddc1dad7503d919241301e404255ffe
Reviewed-on: https://skia-review.googlesource.com/46200
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>