OpenSubdiv/examples/dxPtexViewer/dxPtexViewer.cpp

2136 lines
72 KiB
C++
Raw Normal View History

2013-10-05 01:28:00 +00:00
//
// Copyright 2013 Pixar
//
// Licensed under the Apache License, Version 2.0 (the "Apache License")
// with the following modification; you may not use this file except in
// compliance with the Apache License and the following modification to it:
// Section 6. Trademarks. is deleted and replaced with:
//
// 6. Trademarks. This License does not grant permission to use the trade
// names, trademarks, service marks, or product names of the Licensor
// and its affiliates, except as required to comply with Section 4(c) of
// the License and to reproduce the content of the NOTICE file.
//
// You may obtain a copy of the Apache License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the Apache License with the above modification is
// distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, either express or implied. See the Apache License for the specific
// language governing permissions and limitations under the Apache License.
//
#include <D3D11.h>
#include <D3Dcompiler.h>
#include <far/error.h>
2013-10-05 01:28:00 +00:00
#include <osd/cpuD3D11VertexBuffer.h>
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
#include <osd/cpuEvaluator.h>
2013-10-05 01:28:00 +00:00
#ifdef OPENSUBDIV_HAS_OPENMP
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
#include <osd/ompEvaluator.h>
2013-10-05 01:28:00 +00:00
#endif
#ifdef OPENSUBDIV_HAS_TBB
#include <osd/tbbEvaluator.h>
#endif
2013-10-05 01:28:00 +00:00
#ifdef OPENSUBDIV_HAS_OPENCL
#include <osd/clD3D11VertexBuffer.h>
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
#include <osd/clEvaluator.h>
#include "../common/clDeviceContext.h"
CLD3D11DeviceContext g_clDeviceContext;
2013-10-05 01:28:00 +00:00
#endif
#ifdef OPENSUBDIV_HAS_CUDA
#include <osd/cudaD3D11VertexBuffer.h>
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
#include <osd/cudaEvaluator.h>
#include "../common/cudaDeviceContext.h"
CudaDeviceContext g_cudaDeviceContext;
2013-10-05 01:28:00 +00:00
#endif
#include <osd/d3d11VertexBuffer.h>
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
#include <osd/d3d11ComputeEvaluator.h>
2013-10-05 01:28:00 +00:00
#include <osd/d3d11Mesh.h>
OpenSubdiv::Osd::D3D11MeshInterface *g_mesh;
2013-10-05 01:28:00 +00:00
#include "./sky.h"
2013-10-05 01:28:00 +00:00
#include "Ptexture.h"
#include "PtexUtils.h"
#include "../../regression/common/far_utils.h"
2013-10-05 01:28:00 +00:00
#include "../common/stopwatch.h"
#include "../common/simple_math.h"
#include "../common/d3d11Hud.h"
#include "../common/d3d11PtexMipmapTexture.h"
#include "../common/d3d11ShaderCache.h"
#include "../common/hdr_reader.h"
2013-10-05 01:28:00 +00:00
#include <osd/hlslPatchShaderSource.h>
2013-10-05 01:28:00 +00:00
static const char *g_shaderSource =
#include "shader.gen.h"
2013-10-05 01:28:00 +00:00
;
#include <algorithm>
#include <cfloat>
#include <fstream>
#include <string>
#include <iostream>
#include <iterator>
2013-10-05 01:28:00 +00:00
#include <sstream>
#include <vector>
#define SAFE_RELEASE(p) { if(p) { (p)->Release(); (p)=NULL; } }
enum KernelType { kCPU = 0,
kOPENMP = 1,
kTBB = 2,
kCUDA = 3,
kCL = 4,
kDirectCompute = 5 };
2013-10-05 01:28:00 +00:00
enum HudCheckBox { HUD_CB_ADAPTIVE,
HUD_CB_DISPLAY_OCCLUSION,
HUD_CB_DISPLAY_NORMALMAP,
HUD_CB_DISPLAY_SPECULAR,
HUD_CB_ANIMATE_VERTICES,
HUD_CB_VIEW_LOD,
HUD_CB_FRACTIONAL_SPACING,
HUD_CB_PATCH_CULL,
HUD_CB_IBL,
HUD_CB_BLOOM,
2013-10-16 00:47:58 +00:00
HUD_CB_SEAMLESS_MIPMAP,
2013-10-05 01:28:00 +00:00
HUD_CB_FREEZE };
enum HudRadioGroup { HUD_RB_KERNEL,
HUD_RB_LEVEL,
HUD_RB_SCHEME,
HUD_RB_WIRE,
HUD_RB_COLOR,
HUD_RB_DISPLACEMENT,
HUD_RB_NORMAL };
enum DisplayType { DISPLAY_WIRE,
DISPLAY_SHADED,
DISPLAY_WIRE_ON_SHADED };
enum ColorType { COLOR_NONE,
COLOR_PTEX_NEAREST,
COLOR_PTEX_HW_BILINEAR,
COLOR_PTEX_BILINEAR,
COLOR_PTEX_BIQUADRATIC,
COLOR_PATCHTYPE,
COLOR_PATCHCOORD,
COLOR_NORMAL };
enum DisplacementType { DISPLACEMENT_NONE,
DISPLACEMENT_HW_BILINEAR,
DISPLACEMENT_BILINEAR,
DISPLACEMENT_BIQUADRATIC };
enum NormalType { NORMAL_SURFACE,
NORMAL_FACET,
NORMAL_HW_SCREENSPACE,
NORMAL_SCREENSPACE,
NORMAL_BIQUADRATIC,
NORMAL_BIQUADRATIC_WG };
//-----------------------------------------------------------------------------
int g_frame = 0,
g_repeatCount = 0;
// GUI variables
int g_fullscreen = 0,
g_wire = DISPLAY_SHADED,
g_drawNormals = 0,
g_mbutton[3] = {0, 0, 0},
g_level = 2,
g_tessLevel = 2,
g_kernel = kCPU,
g_scheme = 0,
g_running = 1,
g_maxMipmapLevels = 10,
g_color = COLOR_PTEX_BILINEAR,
g_displacement = DISPLACEMENT_NONE,
g_normal = NORMAL_SURFACE;
float g_moveScale = 0.0f,
g_displacementScale = 1.0f,
g_mipmapBias = 0.0;
bool g_adaptive = true,
g_yup = false,
g_patchCull = true,
g_screenSpaceTess = true,
g_fractionalSpacing = true,
g_ibl = false,
g_bloom = false,
g_freeze = false;
// ptex switch
bool g_occlusion = false,
g_specular = false;
2013-10-16 00:47:58 +00:00
bool g_seamless = true;
2013-10-05 01:28:00 +00:00
// camera
float g_rotate[2] = {0, 0},
g_prev_x = 0,
g_prev_y = 0,
g_dolly = 5,
g_pan[2] = {0, 0},
g_center[3] = {0, 0, 0},
g_size = 0;
// viewport
int g_width = 1024,
g_height = 1024;
D3D11hud *g_hud = NULL;
// performance
float g_cpuTime = 0;
float g_gpuTime = 0;
#define NUM_FPS_TIME_SAMPLES 6
float g_fpsTimeSamples[NUM_FPS_TIME_SAMPLES] = {0, 0, 0, 0, 0, 0};
int g_currentFpsTimeSample = 0;
Stopwatch g_fpsTimer;
float g_animTime = 0;
// geometry
std::vector<float> g_positions,
g_normals;
std::vector<std::vector<float> > g_animPositions;
Sky * g_sky=0;
D3D11PtexMipmapTexture * g_osdPTexImage = 0;
D3D11PtexMipmapTexture * g_osdPTexDisplacement = 0;
D3D11PtexMipmapTexture * g_osdPTexOcclusion = 0;
D3D11PtexMipmapTexture * g_osdPTexSpecular = 0;
2013-10-05 01:28:00 +00:00
const char * g_ptexColorFilename;
size_t g_ptexMemoryUsage = 0;
2013-10-05 01:28:00 +00:00
ID3D11Device * g_pd3dDevice = NULL;
ID3D11DeviceContext * g_pd3dDeviceContext = NULL;
IDXGISwapChain * g_pSwapChain = NULL;
ID3D11RenderTargetView * g_pSwapChainRTV = NULL;
ID3D11RasterizerState* g_pRasterizerState = NULL;
ID3D11InputLayout * g_pInputLayout = NULL;
ID3D11DepthStencilState * g_pDepthStencilState = NULL;
2013-10-05 01:28:00 +00:00
ID3D11Texture2D * g_pDepthStencilBuffer = NULL;
ID3D11Buffer * g_pcbPerFrame = NULL;
ID3D11Buffer * g_pcbTessellation = NULL;
ID3D11Buffer * g_pcbLighting = NULL;
ID3D11Buffer * g_pcbConfig = NULL;
ID3D11DepthStencilView * g_pDepthStencilView = NULL;
ID3D11Texture2D * g_diffuseEnvironmentMap = NULL;
ID3D11ShaderResourceView * g_diffuseEnvironmentSRV = NULL;
ID3D11Texture2D * g_specularEnvironmentMap = NULL;
ID3D11ShaderResourceView * g_specularEnvironmentSRV = NULL;
ID3D11SamplerState * g_iblSampler = NULL;
2013-10-05 01:28:00 +00:00
ID3D11Query * g_pipelineStatsQuery = NULL;
2013-10-05 01:28:00 +00:00
bool g_bDone = false;
//------------------------------------------------------------------------------
static ID3D11Texture2D *
createHDRTexture(ID3D11Device * device, char const * hdrFile) {
HdrInfo info;
unsigned char * image = loadHdr(hdrFile, &info, /*convertToFloat=*/true);
if (image) {
D3D11_TEXTURE2D_DESC texDesc;
ZeroMemory(&texDesc, sizeof(texDesc));
texDesc.Width = info.width;
texDesc.Height = info.height;
texDesc.MipLevels = 1;
texDesc.ArraySize = 1;
texDesc.Format = DXGI_FORMAT_R32G32B32A32_FLOAT;
texDesc.SampleDesc.Count = 1;
texDesc.SampleDesc.Quality = 0;
texDesc.Usage = D3D11_USAGE_DEFAULT;
texDesc.BindFlags = D3D11_BIND_UNORDERED_ACCESS | D3D11_BIND_SHADER_RESOURCE;
texDesc.CPUAccessFlags = 0;
texDesc.MiscFlags = 0;
D3D11_SUBRESOURCE_DATA subData;
subData.pSysMem = (void *)image;
subData.SysMemPitch = info.width * 4 * sizeof(float);
ID3D11Texture2D * texture=0;
device->CreateTexture2D(&texDesc, &subData, &texture);
free(image);
return texture;
}
return 0;
}
2013-10-05 01:28:00 +00:00
//------------------------------------------------------------------------------
static void
calcNormals(OpenSubdiv::Far::TopologyRefiner * refiner,
std::vector<float> const & pos, std::vector<float> & result ) {
typedef OpenSubdiv::Far::ConstIndexArray IndexArray;
2013-10-05 01:28:00 +00:00
// calc normal vectors
OpenSubdiv::Far::TopologyLevel const & refBaseLevel = refiner->GetLevel(0);
int nverts = refBaseLevel.GetNumVertices(),
nfaces = refBaseLevel.GetNumFaces();
for (int face = 0; face < nfaces; ++face) {
2013-10-05 01:28:00 +00:00
IndexArray fverts = refBaseLevel.GetFaceVertices(face);
2013-10-05 01:28:00 +00:00
float const * p0 = &pos[fverts[0]*3],
* p1 = &pos[fverts[1]*3],
* p2 = &pos[fverts[2]*3];
2013-10-05 01:28:00 +00:00
float n[3];
cross(n, p0, p1, p2);
for (int vert = 0; vert < fverts.size(); ++vert) {
int idx = fverts[vert] * 3;
2013-10-05 01:28:00 +00:00
result[idx ] += n[0];
result[idx+1] += n[1];
result[idx+2] += n[2];
}
}
for (int i = 0; i < nverts; ++i)
normalize(&result[i*3]);
}
//------------------------------------------------------------------------------
static void
updateGeom() {
int nverts = (int)g_positions.size() / 3;
if (g_moveScale && g_adaptive && !g_animPositions.empty()) {
2013-10-05 01:28:00 +00:00
// baked animation only works with adaptive for now
// (since non-adaptive requires normals)
int nkeys = (int)g_animPositions.size();
const float fps = 24.0f;
2013-10-05 01:28:00 +00:00
float p = fmodf(g_animTime * fps, (float)nkeys);
int key = (int)p;
float b = p - key;
std::vector<float> vertex;
vertex.reserve(nverts*3);
for (int vert = 0; vert<nverts; ++vert) {
float const * p0 = &g_animPositions[key][vert*3],
* p1 = &g_animPositions[(key+1)%nkeys][vert*3];
for (int i=0; i<3; ++i, ++p0, ++p1) {
vertex.push_back(*p0*(1.0f-b) + *p1*b);
}
// adaptive patches don't need normals, but we do not have an
// an easy shader variant with positions only
vertex.push_back(0.0f);
vertex.push_back(0.0f);
vertex.push_back(0.0f);
2013-10-05 01:28:00 +00:00
}
g_mesh->UpdateVertexBuffer(&vertex[0], 0, nverts);
} else {
std::vector<float> vertex;
vertex.reserve(nverts*6);
const float *p = &g_positions[0];
const float *n = &g_normals[0];
for (int i = 0; i < nverts; ++i) {
float move = g_size*0.005f*cosf(p[0]*100/g_size+g_frame*0.01f);
vertex.push_back(p[0]);
vertex.push_back(p[1]+g_moveScale*move);
vertex.push_back(p[2]);
p += 3;
// if (g_adaptive == false)
{
vertex.push_back(n[0]);
vertex.push_back(n[1]);
vertex.push_back(n[2]);
n += 3;
}
}
g_mesh->UpdateVertexBuffer(&vertex[0], 0, nverts);
}
2013-10-05 01:28:00 +00:00
Stopwatch s;
s.Start();
g_mesh->Refine();
s.Stop();
g_cpuTime = float(s.GetElapsed() * 1000.0f);
s.Start();
g_mesh->Synchronize();
s.Stop();
g_gpuTime = float(s.GetElapsed() * 1000.0f);
}
//-------------------------------------------------------------------------------
static void
fitFrame() {
2013-10-05 01:28:00 +00:00
g_pan[0] = g_pan[1] = 0;
g_dolly = g_size;
}
//-------------------------------------------------------------------------------
Shape *
createPTexGeo(PtexTexture * r) {
2013-10-05 01:28:00 +00:00
PtexMetaData* meta = r->getMetaData();
if (meta->numKeys() < 3) {
return NULL;
}
float const * vp;
int const *vi, *vc;
2013-10-05 01:28:00 +00:00
int nvp, nvi, nvc;
meta->getValue("PtexFaceVertCounts", vc, nvc);
if (nvc == 0) {
2013-10-05 01:28:00 +00:00
return NULL;
}
2013-10-05 01:28:00 +00:00
meta->getValue("PtexVertPositions", vp, nvp);
if (nvp == 0) {
2013-10-05 01:28:00 +00:00
return NULL;
}
2013-10-05 01:28:00 +00:00
meta->getValue("PtexFaceVertIndices", vi, nvi);
if (nvi == 0) {
2013-10-05 01:28:00 +00:00
return NULL;
}
Shape * shape = new Shape;
2013-10-05 01:28:00 +00:00
shape->scheme = kCatmark;
2013-10-05 01:28:00 +00:00
shape->verts.resize(nvp);
for (int i=0; i<nvp; ++i) {
shape->verts[i] = vp[i];
}
shape->nvertsPerFace.resize(nvc);
for (int i=0; i<nvc; ++i) {
shape->nvertsPerFace[i] = vc[i];
}
shape->faceverts.resize(nvi);
for (int i=0; i<nvi; ++i) {
shape->faceverts[i] = vi[i];
}
2013-10-05 01:28:00 +00:00
// compute model bounding
float min[3] = {vp[0], vp[1], vp[2]};
float max[3] = {vp[0], vp[1], vp[2]};
for (int i = 0; i < nvp/3; ++i) {
for (int j = 0; j < 3; ++j) {
float v = vp[i*3+j];
min[j] = std::min(min[j], v);
max[j] = std::max(max[j], v);
}
}
g_size=0.0f;
2013-10-05 01:28:00 +00:00
for (int j = 0; j < 3; ++j) {
g_center[j] = (min[j] + max[j]) * 0.5f;
g_size += (max[j]-min[j])*(max[j]-min[j]);
}
g_size = sqrtf(g_size);
return shape;
2013-10-05 01:28:00 +00:00
}
//------------------------------------------------------------------------------
static const char *
getKernelName(int kernel) {
if (kernel == kCPU)
2013-10-05 01:28:00 +00:00
return "CPU";
else if (kernel == kOPENMP)
return "OpenMP";
else if (kernel == kTBB)
return "TBB";
2013-10-05 01:28:00 +00:00
else if (kernel == kCUDA)
return "Cuda";
else if (kernel == kCL)
return "OpenCL";
else if (kernel == kDirectCompute)
return "DirectCompute";
return "Unknown";
}
//------------------------------------------------------------------------------
union Effect {
struct {
unsigned int wire:2;
unsigned int color:3;
unsigned int displacement:2;
unsigned int normal:3;
int occlusion:1;
int specular:1;
int patchCull:1;
int screenSpaceTess:1;
int fractionalSpacing:1;
int ibl:1;
2013-10-16 00:47:58 +00:00
int seamless:1;
2013-10-05 01:28:00 +00:00
};
int value;
bool operator < (const Effect &e) const {
return value < e.value;
}
};
struct EffectDesc {
EffectDesc(OpenSubdiv::Far::PatchDescriptor desc,
Effect effect) : desc(desc), effect(effect),
maxValence(0), numElements(0) { }
OpenSubdiv::Far::PatchDescriptor desc;
Effect effect;
int maxValence;
int numElements;
bool operator < (const EffectDesc &e) const {
return desc < e.desc || (desc == e.desc &&
(maxValence < e.maxValence || ((maxValence == e.maxValence) &&
(effect < e.effect))));
}
};
2013-10-05 01:28:00 +00:00
class ShaderCache : public D3D11ShaderCache<EffectDesc> {
public:
virtual D3D11DrawConfig *CreateDrawConfig(EffectDesc const &effectDesc) {
using namespace OpenSubdiv;
2013-10-05 01:28:00 +00:00
D3D11DrawConfig *config = new D3D11DrawConfig();
2013-10-05 01:28:00 +00:00
Far::PatchDescriptor::Type type = effectDesc.desc.GetType();
2013-10-05 01:28:00 +00:00
// common defines
std::stringstream ss;
if (type == Far::PatchDescriptor::QUADS) {
ss << "#define PRIM_QUAD\n";
} else {
ss << "#define PRIM_TRI\n";
}
// OSD tessellation controls
if (effectDesc.effect.screenSpaceTess) {
ss << "#define OSD_ENABLE_SCREENSPACE_TESSELLATION\n";
}
if (effectDesc.effect.fractionalSpacing) {
ss << "#define OSD_FRACTIONAL_ODD_SPACING\n";
}
if (effectDesc.effect.patchCull) {
ss << "#define OSD_ENABLE_PATCH_CULL\n";
2013-10-05 01:28:00 +00:00
}
// add ptex functions
ss << D3D11PtexMipmapTexture::GetShaderSource();
2013-10-05 01:28:00 +00:00
// -------------------------------------------------------------
// display styles
// -------------------------------------------------------------
2013-10-05 01:28:00 +00:00
// mipmap
if (effectDesc.effect.seamless) {
ss << "#define SEAMLESS_MIPMAP\n";
}
2013-10-05 01:28:00 +00:00
// wire
if (effectDesc.effect.wire == 0) {
ss << "#define GEOMETRY_OUT_WIRE\n";
} else if (effectDesc.effect.wire == 1) {
ss << "#define GEOMETRY_OUT_FILL\n";
} else if (effectDesc.effect.wire == 2) {
ss << "#define GEOMETRY_OUT_LINE\n";
}
2013-10-05 01:28:00 +00:00
// color
switch(effectDesc.effect.color) {
case COLOR_NONE:
break;
case COLOR_PTEX_NEAREST:
ss << "#define COLOR_PTEX_NEAREST\n";
break;
case COLOR_PTEX_HW_BILINEAR:
ss << "#define COLOR_PTEX_HW_BILINEAR\n";
break;
case COLOR_PTEX_BILINEAR:
ss << "#define COLOR_PTEX_BILINEAR\n";
break;
case COLOR_PTEX_BIQUADRATIC:
ss << "#define COLOR_PTEX_BIQUADRATIC\n";
break;
case COLOR_PATCHTYPE:
ss << "#define COLOR_PATCHTYPE\n";
break;
case COLOR_PATCHCOORD:
ss << "#define COLOR_PATCHCOORD\n";
break;
case COLOR_NORMAL:
ss << "#define COLOR_NORMAL\n";
break;
}
2013-10-16 00:47:58 +00:00
// displacement
switch (effectDesc.effect.displacement) {
case DISPLACEMENT_NONE:
break;
case DISPLACEMENT_HW_BILINEAR:
ss << "#define DISPLACEMENT_HW_BILINEAR\n";
break;
case DISPLACEMENT_BILINEAR:
ss << "#define DISPLACEMENT_BILINEAR\n";
break;
case DISPLACEMENT_BIQUADRATIC:
ss << "#define DISPLACEMENT_BIQUADRATIC\n";
break;
}
2013-10-05 01:28:00 +00:00
// normal
switch (effectDesc.effect.normal) {
case NORMAL_FACET:
ss << "#define NORMAL_FACET\n";
break;
case NORMAL_HW_SCREENSPACE:
ss << "#define NORMAL_HW_SCREENSPACE\n";
break;
case NORMAL_SCREENSPACE:
ss << "#define NORMAL_SCREENSPACE\n";
break;
case NORMAL_BIQUADRATIC:
ss << "#define NORMAL_BIQUADRATIC\n";
break;
case NORMAL_BIQUADRATIC_WG:
ss << "#define OSD_COMPUTE_NORMAL_DERIVATIVES\n";
ss << "#define NORMAL_BIQUADRATIC_WG\n";
break;
}
2013-10-05 01:28:00 +00:00
// occlusion
if (effectDesc.effect.occlusion)
ss << "#define USE_PTEX_OCCLUSION\n";
// specular
if (effectDesc.effect.specular)
ss << "#define USE_PTEX_SPECULAR\n";
// IBL
if (effectDesc.effect.ibl)
ss << "#define USE_IBL\n";
// need for patch color-coding : we need these defines in the fragment shader
if (type == Far::PatchDescriptor::GREGORY) {
ss << "#define OSD_PATCH_GREGORY\n";
} else if (type == Far::PatchDescriptor::GREGORY_BOUNDARY) {
ss << "#define OSD_PATCH_GREGORY_BOUNDARY\n";
} else if (type == Far::PatchDescriptor::GREGORY_BASIS) {
ss << "#define OSD_PATCH_GREGORY_BASIS\n";
}
// include osd PatchCommon
ss << Osd::HLSLPatchShaderSource::GetCommonShaderSource();
std::string common = ss.str();
ss.str("");
2013-10-05 01:28:00 +00:00
// input layout
const D3D11_INPUT_ELEMENT_DESC hInElementDesc[] = {
{ "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 },
{ "NORMAL", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 4*3, D3D11_INPUT_PER_VERTEX_DATA, 0 }
};
// vertex shader
ss << common
<< g_shaderSource
<< Osd::HLSLPatchShaderSource::GetVertexShaderSource(type);
if (effectDesc.desc.IsAdaptive()) {
config->CompileVertexShader("vs_5_0", "vs_main_patches", ss.str(),
&g_pInputLayout,
hInElementDesc,
ARRAYSIZE(hInElementDesc),
g_pd3dDevice);
} else {
config->CompileVertexShader("vs_5_0", "vs_main",
ss.str(),
&g_pInputLayout,
hInElementDesc,
ARRAYSIZE(hInElementDesc),
g_pd3dDevice);
}
ss.str("");
if (effectDesc.desc.IsAdaptive()) {
// hull shader
ss << common
<< g_shaderSource
<< Osd::HLSLPatchShaderSource::GetHullShaderSource(type);
config->CompileHullShader("hs_5_0", "hs_main_patches", ss.str(),
g_pd3dDevice);
ss.str("");
// domain shader
ss << common
<< g_shaderSource
<< Osd::HLSLPatchShaderSource::GetDomainShaderSource(type);
config->CompileDomainShader("ds_5_0", "ds_main_patches", ss.str(),
g_pd3dDevice);
ss.str("");
}
2013-10-05 01:28:00 +00:00
// geometry shader
ss << common
<< g_shaderSource;
config->CompileGeometryShader("gs_5_0", "gs_main", ss.str(),
g_pd3dDevice);
ss.str("");
// pixel shader
ss << common
<< g_shaderSource;
config->CompilePixelShader("ps_5_0", "ps_main", ss.str(),
g_pd3dDevice);
ss.str("");
return config;
};
};
ShaderCache g_shaderCache;
2013-10-05 01:28:00 +00:00
//------------------------------------------------------------------------------
D3D11PtexMipmapTexture *
createPtex(const char *filename, int memLimit) {
2013-10-05 01:28:00 +00:00
Ptex::String ptexError;
printf("Loading ptex : %s\n", filename);
#define USE_PTEX_CACHE
#define PTEX_CACHE_SIZE (512*1024*1024)
#ifdef USE_PTEX_CACHE
PtexCache *cache = PtexCache::create(1, PTEX_CACHE_SIZE);
PtexTexture *ptex = cache->get(filename, ptexError);
#else
PtexTexture *ptex = PtexTexture::open(filename, ptexError, true);
#endif
if (ptex == NULL) {
printf("Error in reading %s\n", filename);
exit(1);
}
size_t targetMemory = memLimit * 1024 * 1024; // MB
D3D11PtexMipmapTexture *osdPtex = D3D11PtexMipmapTexture::Create(
g_pd3dDeviceContext, ptex, g_maxMipmapLevels, targetMemory);
2013-10-05 01:28:00 +00:00
ptex->release();
#ifdef USE_PTEX_CACHE
cache->release();
#endif
return osdPtex;
}
//------------------------------------------------------------------------------
2013-10-05 01:28:00 +00:00
void
createOsdMesh(int level, int kernel) {
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
using namespace OpenSubdiv;
2013-10-05 01:28:00 +00:00
Ptex::String ptexError;
PtexTexture *ptexColor = PtexTexture::open(g_ptexColorFilename, ptexError, true);
if (ptexColor == NULL) {
printf("Error in reading %s\n", g_ptexColorFilename);
exit(1);
}
// generate Hbr representation from ptex
Shape * shape = createPTexGeo(ptexColor);
if (!shape) {
return;
}
g_positions=shape->verts;
typedef OpenSubdiv::Far::ConstIndexArray IndexArray;
// create Far mesh (topology)
OpenSubdiv::Sdc::SchemeType sdctype = GetSdcType(*shape);
OpenSubdiv::Sdc::Options sdcoptions = GetSdcOptions(*shape);
OpenSubdiv::Far::TopologyRefiner * refiner =
OpenSubdiv::Far::TopologyRefinerFactory<Shape>::Create(*shape,
OpenSubdiv::Far::TopologyRefinerFactory<Shape>::Options(sdctype, sdcoptions));
// save coarse topology (used for coarse mesh drawing)
OpenSubdiv::Far::TopologyLevel const & refBaseLevel = refiner->GetLevel(0);
// create cage edge index
int nedges = refBaseLevel.GetNumEdges();
std::vector<int> edgeIndices(nedges*2);
for(int i=0; i<nedges; ++i) {
IndexArray verts = refBaseLevel.GetEdgeVertices(i);
edgeIndices[i*2 ]=verts[0];
edgeIndices[i*2+1]=verts[1];
}
delete shape;
2013-10-05 01:28:00 +00:00
g_normals.resize(g_positions.size(), 0.0f);
calcNormals(refiner, g_positions, g_normals);
2013-10-05 01:28:00 +00:00
delete g_mesh;
g_mesh = NULL;
// Adaptive refinement currently supported only for catmull-clark scheme
bool doAdaptive = (g_adaptive != 0 && g_scheme == 0);
2013-10-05 01:28:00 +00:00
OpenSubdiv::Osd::MeshBitset bits;
bits.set(OpenSubdiv::Osd::MeshAdaptive, doAdaptive);
bits.set(OpenSubdiv::Osd::MeshEndCapGregoryBasis, true);
2013-10-05 01:28:00 +00:00
int numVertexElements = 6; //g_adaptive ? 3 : 6;
int numVaryingElements = 0;
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
if (g_kernel == kCPU) {
g_mesh = new Osd::Mesh<Osd::CpuD3D11VertexBuffer,
Far::StencilTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
Osd::CpuEvaluator,
Osd::D3D11PatchTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
ID3D11DeviceContext>(
refiner,
numVertexElements,
numVaryingElements,
level, bits, NULL, g_pd3dDeviceContext);
2013-10-05 01:28:00 +00:00
#ifdef OPENSUBDIV_HAS_OPENMP
} else if (kernel == kOPENMP) {
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
g_mesh = new Osd::Mesh<Osd::CpuD3D11VertexBuffer,
Far::StencilTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
Osd::OmpEvaluator,
Osd::D3D11PatchTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
ID3D11DeviceContext>(
refiner,
numVertexElements,
numVaryingElements,
level, bits, NULL, g_pd3dDeviceContext);
#endif
#ifdef OPENSUBDIV_HAS_TBB
} else if (kernel == kTBB) {
g_mesh = new Osd::Mesh<Osd::CpuD3D11VertexBuffer,
Far::StencilTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
Osd::TbbEvaluator,
Osd::D3D11PatchTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
ID3D11DeviceContext>(
refiner,
numVertexElements,
numVaryingElements,
level, bits, NULL, g_pd3dDeviceContext);
2013-10-05 01:28:00 +00:00
#endif
#ifdef OPENSUBDIV_HAS_OPENCL
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
} else if(kernel == kCL) {
static Osd::EvaluatorCacheT<Osd::CLEvaluator> clEvaluatorCache;
g_mesh = new Osd::Mesh<Osd::CLD3D11VertexBuffer,
Osd::CLStencilTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
Osd::CLEvaluator,
Osd::D3D11PatchTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
CLD3D11DeviceContext>(
refiner,
numVertexElements,
numVaryingElements,
level, bits,
&clEvaluatorCache,
&g_clDeviceContext);
2013-10-05 01:28:00 +00:00
#endif
#ifdef OPENSUBDIV_HAS_CUDA
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
} else if (g_kernel == kCUDA) {
g_mesh = new Osd::Mesh<Osd::CudaD3D11VertexBuffer,
Osd::CudaStencilTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
Osd::CudaEvaluator,
Osd::D3D11PatchTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
ID3D11DeviceContext>(
refiner,
numVertexElements,
numVaryingElements,
level, bits, NULL, g_pd3dDeviceContext);
2013-10-05 01:28:00 +00:00
#endif
} else if (g_kernel == kDirectCompute) {
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
static Osd::EvaluatorCacheT<Osd::D3D11ComputeEvaluator> d3d11ComputeEvaluatorCache;
g_mesh = new Osd::Mesh<Osd::D3D11VertexBuffer,
Osd::D3D11StencilTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
Osd::D3D11ComputeEvaluator,
Osd::D3D11PatchTable,
Refurbish osd layer API. In OpenSubdiv 2.x, we encapsulated subdivision tables into compute context in osd layer since those tables are order-dependent and have to be applied in a certain manner. In 3.0, we adopted stencil table based refinement. It's more simple and such an encapsulation is no longer needed. Also 2.0 API has several ownership issues of GPU kernel caching, and forces unnecessary instantiation of controllers even though the cpu kernels typically don't need instances unlike GPU ones. This change completely revisit osd client facing APIs. All contexts and controllers were replaced with device-specific tables and evaluators. While we can still use consistent API across various device backends, unnecessary complexities have been removed. For example, cpu evaluator is just a set of static functions and also there's no need to replicate FarStencilTables to ComputeContext. Also the new API delegates the ownership of compiled GPU kernels to clients, for the better management of resources especially in multiple GPU environment. In addition to integrating ComputeController and EvalStencilController into a single function Evaluator::EvalStencils(), EvalLimit API is also added into Evaluator. This is working but still in progress, and we'll make a followup change for the complete implementation. -some naming convention changes: GLSLTransformFeedback to GLXFBEvaluator GLSLCompute to GLComputeEvaluator -move LimitLocation struct into examples/glEvalLimit. We're still discussing patch evaluation interface. Basically we'd like to tease all ptex-specific parametrization out of far/osd layer. TODO: -implments EvalPatches() in the right way -derivative evaluation API is still interim. -VertexBufferDescriptor needs a better API to advance its location -synchronization mechanism is not ideal (too global). -OsdMesh class is hacky. need to fix it.
2015-05-09 00:31:26 +00:00
ID3D11DeviceContext>(
refiner,
numVertexElements,
numVaryingElements,
level, bits,
&d3d11ComputeEvaluatorCache,
g_pd3dDeviceContext);
2013-10-05 01:28:00 +00:00
} else {
printf("Unsupported kernel %s\n", getKernelName(kernel));
}
updateGeom();
}
//------------------------------------------------------------------------------
static void
bindProgram(Effect effect, OpenSubdiv::Osd::PatchArray const & patch) {
2013-10-05 01:28:00 +00:00
EffectDesc effectDesc(patch.GetDescriptor(), effect);
D3D11DrawConfig *config = g_shaderCache.GetDrawConfig(effectDesc);
2013-10-05 01:28:00 +00:00
assert(g_pInputLayout);
// Update transform state
{
__declspec(align(16))
struct CB_PER_FRAME_CONSTANTS
{
float ModelViewMatrix[16];
float ProjectionMatrix[16];
float ModelViewProjectionMatrix[16];
float ModelViewInverseMatrix[16];
2013-10-05 01:28:00 +00:00
};
if (! g_pcbPerFrame) {
D3D11_BUFFER_DESC cbDesc;
ZeroMemory(&cbDesc, sizeof(cbDesc));
cbDesc.Usage = D3D11_USAGE_DYNAMIC;
cbDesc.BindFlags = D3D11_BIND_CONSTANT_BUFFER;
cbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
cbDesc.MiscFlags = 0;
cbDesc.ByteWidth = sizeof(CB_PER_FRAME_CONSTANTS);
g_pd3dDevice->CreateBuffer(&cbDesc, NULL, &g_pcbPerFrame);
}
assert(g_pcbPerFrame);
D3D11_MAPPED_SUBRESOURCE MappedResource;
g_pd3dDeviceContext->Map(g_pcbPerFrame, 0, D3D11_MAP_WRITE_DISCARD, 0, &MappedResource);
CB_PER_FRAME_CONSTANTS* pData = ( CB_PER_FRAME_CONSTANTS* )MappedResource.pData;
float aspect = (g_height > 0) ? (float)g_width / g_height : 1.0f;
identity(pData->ModelViewMatrix);
translate(pData->ModelViewMatrix, -g_pan[0], -g_pan[1], -g_dolly);
rotate(pData->ModelViewMatrix, g_rotate[1], 1, 0, 0);
rotate(pData->ModelViewMatrix, g_rotate[0], 0, 1, 0);
translate(pData->ModelViewMatrix, -g_center[0], -g_center[1], -g_center[2]);
2013-10-05 01:28:00 +00:00
identity(pData->ProjectionMatrix);
perspective(pData->ProjectionMatrix, 45.0, aspect, g_size*0.001f, g_size+g_dolly);
multMatrix(pData->ModelViewProjectionMatrix,
pData->ModelViewMatrix,
pData->ProjectionMatrix);
inverseMatrix(pData->ModelViewInverseMatrix,
pData->ModelViewMatrix);
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->Unmap( g_pcbPerFrame, 0 );
}
// Update tessellation state
{
__declspec(align(16))
struct Tessellation {
float TessLevel;
int PrimitiveIdBase;
};
if (! g_pcbTessellation) {
D3D11_BUFFER_DESC cbDesc;
ZeroMemory(&cbDesc, sizeof(cbDesc));
cbDesc.Usage = D3D11_USAGE_DYNAMIC;
cbDesc.BindFlags = D3D11_BIND_CONSTANT_BUFFER;
cbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
cbDesc.MiscFlags = 0;
cbDesc.ByteWidth = sizeof(Tessellation);
g_pd3dDevice->CreateBuffer(&cbDesc, NULL, &g_pcbTessellation);
}
assert(g_pcbTessellation);
D3D11_MAPPED_SUBRESOURCE MappedResource;
g_pd3dDeviceContext->Map(g_pcbTessellation, 0, D3D11_MAP_WRITE_DISCARD, 0, &MappedResource);
Tessellation * pData = ( Tessellation* )MappedResource.pData;
pData->TessLevel = static_cast<float>(1 << g_tessLevel);
pData->PrimitiveIdBase = patch.GetPrimitiveIdBase();
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->Unmap( g_pcbTessellation, 0 );
}
// Update config state
{
__declspec(align(16))
struct Config {
float displacementScale;
float mipmapBias;
};
if (! g_pcbConfig) {
D3D11_BUFFER_DESC cbDesc;
ZeroMemory(&cbDesc, sizeof(cbDesc));
cbDesc.Usage = D3D11_USAGE_DYNAMIC;
cbDesc.BindFlags = D3D11_BIND_CONSTANT_BUFFER;
cbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
cbDesc.MiscFlags = 0;
cbDesc.ByteWidth = sizeof(Config);
g_pd3dDevice->CreateBuffer(&cbDesc, NULL, &g_pcbConfig);
}
assert(g_pcbConfig);
D3D11_MAPPED_SUBRESOURCE MappedResource;
g_pd3dDeviceContext->Map(g_pcbConfig, 0, D3D11_MAP_WRITE_DISCARD, 0, &MappedResource);
Config * pData = ( Config* )MappedResource.pData;
pData->displacementScale = g_displacementScale;
pData->mipmapBias = g_mipmapBias;
g_pd3dDeviceContext->Unmap( g_pcbConfig, 0 );
}
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->IASetInputLayout(g_pInputLayout);
g_pd3dDeviceContext->VSSetShader(config->GetVertexShader(), NULL, 0);
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->VSSetConstantBuffers(0, 1, &g_pcbPerFrame);
g_pd3dDeviceContext->HSSetShader(config->GetHullShader(), NULL, 0);
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->HSSetConstantBuffers(0, 1, &g_pcbPerFrame);
g_pd3dDeviceContext->HSSetConstantBuffers(1, 1, &g_pcbTessellation);
g_pd3dDeviceContext->DSSetShader(config->GetDomainShader(), NULL, 0);
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->DSSetConstantBuffers(0, 1, &g_pcbPerFrame);
g_pd3dDeviceContext->DSSetConstantBuffers(3, 1, &g_pcbConfig);
g_pd3dDeviceContext->GSSetShader(config->GetGeometryShader(), NULL, 0);
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->GSSetConstantBuffers(0, 1, &g_pcbPerFrame);
g_pd3dDeviceContext->PSSetShader(config->GetPixelShader(), NULL, 0);
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->PSSetConstantBuffers(0, 1, &g_pcbPerFrame);
g_pd3dDeviceContext->PSSetConstantBuffers(2, 1, &g_pcbLighting);
g_pd3dDeviceContext->PSSetConstantBuffers(3, 1, &g_pcbConfig);
2013-10-05 01:28:00 +00:00
ID3D11ShaderResourceView *srv = g_mesh->GetPatchTable()->GetPatchParamSRV();
if (srv) {
g_pd3dDeviceContext->HSSetShaderResources(0, 1, &srv);
g_pd3dDeviceContext->DSSetShaderResources(0, 1, &srv);
g_pd3dDeviceContext->GSSetShaderResources(0, 1, &srv);
g_pd3dDeviceContext->PSSetShaderResources(0, 1, &srv);
2013-10-05 01:28:00 +00:00
}
g_pd3dDeviceContext->PSSetShaderResources(4, 1, g_osdPTexImage->GetTexelsSRV());
g_pd3dDeviceContext->PSSetShaderResources(5, 1, g_osdPTexImage->GetLayoutSRV());
if (g_osdPTexDisplacement) {
g_pd3dDeviceContext->DSSetShaderResources(6, 1, g_osdPTexDisplacement->GetTexelsSRV());
g_pd3dDeviceContext->DSSetShaderResources(7, 1, g_osdPTexDisplacement->GetLayoutSRV());
g_pd3dDeviceContext->PSSetShaderResources(6, 1, g_osdPTexDisplacement->GetTexelsSRV());
g_pd3dDeviceContext->PSSetShaderResources(7, 1, g_osdPTexDisplacement->GetLayoutSRV());
}
if (g_osdPTexOcclusion) {
g_pd3dDeviceContext->PSSetShaderResources(8, 1, g_osdPTexOcclusion->GetTexelsSRV());
g_pd3dDeviceContext->PSSetShaderResources(9, 1, g_osdPTexOcclusion->GetLayoutSRV());
}
if (g_osdPTexSpecular) {
g_pd3dDeviceContext->PSSetShaderResources(10, 1, g_osdPTexSpecular->GetTexelsSRV());
g_pd3dDeviceContext->PSSetShaderResources(11, 1, g_osdPTexSpecular->GetLayoutSRV());
}
if (g_ibl) {
if (g_diffuseEnvironmentMap) {
g_pd3dDeviceContext->PSSetShaderResources(12, 1, &g_diffuseEnvironmentSRV);
}
if (g_specularEnvironmentMap) {
g_pd3dDeviceContext->PSSetShaderResources(13, 1, &g_specularEnvironmentSRV);
}
assert(g_iblSampler);
g_pd3dDeviceContext->PSSetSamplers(0, 1, &g_iblSampler);
}
2013-10-05 01:28:00 +00:00
}
//------------------------------------------------------------------------------
2013-10-05 01:28:00 +00:00
static void
drawModel() {
2013-10-05 01:28:00 +00:00
ID3D11Buffer *buffer = g_mesh->BindVertexBuffer();
assert(buffer);
UINT hStrides = 6*sizeof(float);
UINT hOffsets = 0;
g_pd3dDeviceContext->IASetVertexBuffers(0, 1, &buffer, &hStrides, &hOffsets);
OpenSubdiv::Osd::PatchArrayVector const & patches =
g_mesh->GetPatchTable()->GetPatchArrays();
2013-10-05 01:28:00 +00:00
g_pd3dDeviceContext->IASetIndexBuffer(
g_mesh->GetPatchTable()->GetPatchIndexBuffer(),
DXGI_FORMAT_R32_UINT, 0);
2013-10-05 01:28:00 +00:00
// patch drawing
for (int i = 0; i < (int)patches.size(); ++i) {
OpenSubdiv::Osd::PatchArray const & patch = patches[i];
OpenSubdiv::Far::PatchDescriptor desc = patch.GetDescriptor();
OpenSubdiv::Far::PatchDescriptor::Type patchType = desc.GetType();
2013-10-05 01:28:00 +00:00
D3D11_PRIMITIVE_TOPOLOGY topology;
switch (patchType) {
case OpenSubdiv::Far::PatchDescriptor::TRIANGLES:
topology = D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST;
break;
case OpenSubdiv::Far::PatchDescriptor::QUADS:
topology = D3D11_PRIMITIVE_TOPOLOGY_LINELIST_ADJ;
break;
default:
switch (desc.GetNumControlVertices()) {
2013-10-05 01:28:00 +00:00
case 4:
topology = D3D11_PRIMITIVE_TOPOLOGY_4_CONTROL_POINT_PATCHLIST;
break;
case 9:
topology = D3D11_PRIMITIVE_TOPOLOGY_9_CONTROL_POINT_PATCHLIST;
break;
case 12:
topology = D3D11_PRIMITIVE_TOPOLOGY_12_CONTROL_POINT_PATCHLIST;
break;
case 16:
topology = D3D11_PRIMITIVE_TOPOLOGY_16_CONTROL_POINT_PATCHLIST;
break;
case 20:
topology = D3D11_PRIMITIVE_TOPOLOGY_20_CONTROL_POINT_PATCHLIST;
break;
2013-10-05 01:28:00 +00:00
default:
assert(false);
break;
}
break;
}
2013-10-05 01:28:00 +00:00
Effect effect;
effect.value = 0;
effect.color = g_color;
effect.displacement = g_displacement;
effect.occlusion = g_occlusion;
effect.normal = g_normal;
effect.specular = g_specular;
effect.patchCull = g_patchCull;
effect.screenSpaceTess = g_screenSpaceTess;
effect.fractionalSpacing = g_fractionalSpacing;
effect.ibl = g_ibl;
effect.wire = g_wire;
2013-10-16 00:47:58 +00:00
effect.seamless = g_seamless;
2013-10-05 01:28:00 +00:00
bindProgram(effect, patch);
g_pd3dDeviceContext->IASetPrimitiveTopology(topology);
g_pd3dDeviceContext->DrawIndexed(
patch.GetNumPatches() * desc.GetNumControlVertices(),
patch.GetIndexBase(), 0);
2013-10-05 01:28:00 +00:00
}
}
//------------------------------------------------------------------------------
static void
display() {
Stopwatch s;
s.Start();
2013-10-05 01:28:00 +00:00
float color[4] = {0.006f, 0.006f, 0.006f, 1.0f};
g_pd3dDeviceContext->ClearRenderTargetView(g_pSwapChainRTV, color);
// Clear the depth buffer.
g_pd3dDeviceContext->ClearDepthStencilView(g_pDepthStencilView, D3D11_CLEAR_DEPTH, 1.0f, 0);
if (g_ibl && g_sky) {
float modelView[16], projection[16], mvp[16];
double aspect = g_width/(double)g_height;
identity(modelView);
rotate(modelView, g_rotate[1], 1, 0, 0);
rotate(modelView, g_rotate[0], 0, 1, 0);
perspective(projection, 45.0f, (float)aspect, g_size*0.001f, g_size+g_dolly);
multMatrix(mvp, modelView, projection);
g_sky->Draw(g_pd3dDeviceContext, mvp);
}
2013-10-05 01:28:00 +00:00
// this query can be slow : comment out #define to turn off
#define PIPELINE_STATISTICS
#ifdef PIPELINE_STATISTICS
g_pd3dDeviceContext->Begin(g_pipelineStatsQuery);
#endif // PIPELINE_STATISTICS
g_pd3dDeviceContext->OMSetDepthStencilState(g_pDepthStencilState, 1);
g_pd3dDeviceContext->RSSetState(g_pRasterizerState);
2013-10-05 01:28:00 +00:00
drawModel();
#ifdef PIPELINE_STATISTICS
g_pd3dDeviceContext->End(g_pipelineStatsQuery);
#endif // PIPELINE_STATISTICS
s.Stop();
float drawCpuTime = float(s.GetElapsed() * 1000.0f);
int timeElapsed = 0;
// XXXX TODO GPU cycles elapsed query
float drawGpuTime = timeElapsed / 1000.0f / 1000.0f;
UINT64 numPrimsGenerated = 0;
#ifdef PIPELINE_STATISTICS
D3D11_QUERY_DATA_PIPELINE_STATISTICS pipelineStats;
ZeroMemory(&pipelineStats, sizeof(pipelineStats));
while (S_OK != g_pd3dDeviceContext->GetData(g_pipelineStatsQuery, &pipelineStats, g_pipelineStatsQuery->GetDataSize(), 0));
numPrimsGenerated = pipelineStats.GSPrimitives;
#endif // PIPELINE_STATISTICS
2013-10-05 01:28:00 +00:00
g_fpsTimer.Stop();
float elapsed = (float)g_fpsTimer.GetElapsed();
if (!g_freeze)
g_animTime += elapsed;
2013-10-05 01:28:00 +00:00
g_fpsTimer.Start();
if (g_hud->IsVisible()) {
double fps = 1.0/elapsed;
// Avereage fps over a defined number of time samples for
// easier reading in the HUD
g_fpsTimeSamples[g_currentFpsTimeSample++] = float(fps);
if (g_currentFpsTimeSample >= NUM_FPS_TIME_SAMPLES)
g_currentFpsTimeSample = 0;
double averageFps = 0;
for (int i = 0; i < NUM_FPS_TIME_SAMPLES; ++i) {
averageFps += g_fpsTimeSamples[i]/(float)NUM_FPS_TIME_SAMPLES;
}
g_hud->DrawString(10, -220, "Ptex memory use : %.1f mb", g_ptexMemoryUsage/1024.0/1024.0);
g_hud->DrawString(10, -180, "Tess level (+/-): %d", g_tessLevel);
#ifdef PIPELINE_STATISTICS
if (numPrimsGenerated > 1000000) {
g_hud->DrawString(10, -160, "Primitives : %3.1f million",
(float)numPrimsGenerated/1000000.0);
} else if (numPrimsGenerated > 1000) {
g_hud->DrawString(10, -160, "Primitives : %3.1f thousand",
(float)numPrimsGenerated/1000.0);
} else {
g_hud->DrawString(10, -160, "Primitives : %d", numPrimsGenerated);
}
#endif // PIPELINE_STATISTICS
g_hud->DrawString(10, -140, "Vertices : %d", g_mesh->GetNumVertices());
g_hud->DrawString(10, -120, "Scheme : %s", g_scheme == 0 ? "CATMARK" : "LOOP");
g_hud->DrawString(10, -100, "GPU Kernel : %.3f ms", g_gpuTime);
g_hud->DrawString(10, -80, "CPU Kernel : %.3f ms", g_cpuTime);
g_hud->DrawString(10, -60, "GPU Draw : %.3f ms", drawGpuTime);
g_hud->DrawString(10, -40, "CPU Draw : %.3f ms", drawCpuTime);
g_hud->DrawString(10, -20, "FPS : %3.1f", fps);
2013-10-05 01:28:00 +00:00
}
g_hud->Flush();
g_pSwapChain->Present(0, 0);
}
//------------------------------------------------------------------------------
static void
mouse(int button, int state, int x, int y) {
if (state == 0)
g_hud->MouseRelease();
if (button == 0 && state == 1 && g_hud->MouseClick(x, y)) return;
if (button < 3) {
g_prev_x = float(x);
g_prev_y = float(y);
g_mbutton[button] = state;
}
}
//------------------------------------------------------------------------------
static void
motion(int x, int y) {
if (g_hud->MouseCapture()) {
// check gui
g_hud->MouseMotion(x, y);
} else if (g_mbutton[0] && !g_mbutton[1] && !g_mbutton[2]) {
// orbit
g_rotate[0] += x - g_prev_x;
g_rotate[1] += y - g_prev_y;
} else if (!g_mbutton[0] && g_mbutton[1] && !g_mbutton[2]) {
// pan
g_pan[0] -= g_dolly*(x - g_prev_x)/g_width;
g_pan[1] += g_dolly*(y - g_prev_y)/g_height;
} else if ((g_mbutton[0] && g_mbutton[1] && !g_mbutton[2]) ||
2013-10-05 01:28:00 +00:00
(!g_mbutton[0] && !g_mbutton[1] && g_mbutton[2])) {
// dolly
g_dolly -= g_dolly*0.01f*(x - g_prev_x);
if(g_dolly <= 0.01) g_dolly = 0.01f;
}
g_prev_x = float(x);
g_prev_y = float(y);
}
//-----------------------------------------------------------------------------
static void
quit() {
g_bDone = true;
if (g_osdPTexImage) delete g_osdPTexImage;
if (g_osdPTexDisplacement) delete g_osdPTexDisplacement;
if (g_osdPTexOcclusion) delete g_osdPTexOcclusion;
if (g_osdPTexSpecular) delete g_osdPTexSpecular;
if (g_mesh) delete g_mesh;
if (g_hud) delete g_hud;
SAFE_RELEASE(g_pRasterizerState);
SAFE_RELEASE(g_pInputLayout);
SAFE_RELEASE(g_pDepthStencilState);
SAFE_RELEASE(g_pcbPerFrame);
SAFE_RELEASE(g_pcbTessellation);
SAFE_RELEASE(g_pcbLighting);
SAFE_RELEASE(g_pcbConfig);
2013-10-05 01:28:00 +00:00
SAFE_RELEASE(g_pDepthStencilView);
SAFE_RELEASE(g_pSwapChainRTV);
SAFE_RELEASE(g_pSwapChain);
SAFE_RELEASE(g_pd3dDeviceContext);
SAFE_RELEASE(g_pd3dDevice);
2013-10-05 01:28:00 +00:00
PostQuitMessage(0);
exit(0);
}
//------------------------------------------------------------------------------
static void
keyboard(char key) {
if (g_hud->KeyDown((int)tolower(key))) return;
switch (key) {
case 'Q': quit();
case 'F': fitFrame(); break;
case '+':
case '=': g_tessLevel++; break;
case '-': g_tessLevel = std::max(1, g_tessLevel-1); break;
case 0x1b: g_hud->SetVisible(!g_hud->IsVisible()); break;
}
}
//------------------------------------------------------------------------------
static void
callbackWireframe(int b) {
2013-10-05 01:28:00 +00:00
g_wire = b;
}
static void
callbackKernel(int k) {
2013-10-05 01:28:00 +00:00
g_kernel = k;
#ifdef OPENSUBDIV_HAS_OPENCL
if (g_kernel == kCL && (!g_clDeviceContext.IsInitialized())) {
if (g_clDeviceContext.Initialize(g_pd3dDeviceContext) == false) {
2013-10-05 01:28:00 +00:00
printf("Error in initializing OpenCL\n");
exit(1);
}
}
#endif
#ifdef OPENSUBDIV_HAS_CUDA
if (g_kernel == kCUDA && (!g_cudaDeviceContext.IsInitialized())) {
if (g_cudaDeviceContext.Initialize(g_pd3dDevice) == false) {
printf("Error in initializing Cuda\n");
exit(1);
}
2013-10-05 01:28:00 +00:00
}
#endif
createOsdMesh(g_level, g_kernel);
}
static void
callbackScheme(int s) {
2013-10-05 01:28:00 +00:00
g_scheme = s;
createOsdMesh(g_level, g_kernel);
}
static void
callbackLevel(int l) {
2013-10-05 01:28:00 +00:00
g_level = l;
createOsdMesh(g_level, g_kernel);
}
static void
callbackColor(int c) {
2013-10-05 01:28:00 +00:00
g_color = c;
}
static void
callbackDisplacement(int d) {
2013-10-05 01:28:00 +00:00
g_displacement = d;
}
static void
callbackNormal(int n) {
2013-10-05 01:28:00 +00:00
g_normal = n;
}
static void
callbackCheckBox(bool checked, int button) {
2013-10-05 01:28:00 +00:00
bool rebuild = false;
switch (button) {
case HUD_CB_ADAPTIVE:
g_adaptive = checked;
rebuild = true;
break;
case HUD_CB_DISPLAY_OCCLUSION:
g_occlusion = checked;
break;
case HUD_CB_DISPLAY_SPECULAR:
g_specular = checked;
break;
case HUD_CB_ANIMATE_VERTICES:
g_moveScale = checked ? 1.0f : 0.0f;
g_animTime = 0;
break;
case HUD_CB_VIEW_LOD:
g_screenSpaceTess = checked;
break;
case HUD_CB_FRACTIONAL_SPACING:
g_fractionalSpacing = checked;
break;
case HUD_CB_PATCH_CULL:
g_patchCull = checked;
break;
case HUD_CB_IBL:
g_ibl = checked;
break;
2013-10-16 00:47:58 +00:00
case HUD_CB_SEAMLESS_MIPMAP:
g_seamless = checked;
break;
2013-10-05 01:28:00 +00:00
case HUD_CB_FREEZE:
g_freeze = checked;
break;
}
if (rebuild)
createOsdMesh(g_level, g_kernel);
}
static void
callbackSlider(float value, int data) {
2013-10-05 01:28:00 +00:00
switch (data) {
case 0:
g_mipmapBias = value;
break;
case 1:
g_displacementScale = value;
break;
}
}
static void
initHUD() {
2013-10-05 01:28:00 +00:00
g_hud = new D3D11hud(g_pd3dDeviceContext);
g_hud->Init(g_width, g_height);
if (g_osdPTexOcclusion != NULL) {
g_hud->AddCheckBox("Ambient Occlusion (A)", g_occlusion,
-200, 570, callbackCheckBox, HUD_CB_DISPLAY_OCCLUSION, 'a');
}
if (g_osdPTexSpecular != NULL)
g_hud->AddCheckBox("Specular (S)", g_specular,
-200, 590, callbackCheckBox, HUD_CB_DISPLAY_SPECULAR, 's');
2013-10-05 01:28:00 +00:00
if (g_diffuseEnvironmentMap || g_diffuseEnvironmentMap) {
g_hud->AddCheckBox("IBL (I)", g_ibl,
-200, 610, callbackCheckBox, HUD_CB_IBL, 'i');
}
g_hud->AddCheckBox("Animate vertices (M)", g_moveScale != 0.0,
10, 30, callbackCheckBox, HUD_CB_ANIMATE_VERTICES, 'm');
g_hud->AddCheckBox("Screen space LOD (V)", g_screenSpaceTess,
10, 50, callbackCheckBox, HUD_CB_VIEW_LOD, 'v');
g_hud->AddCheckBox("Fractional spacing (T)", g_fractionalSpacing,
10, 70, callbackCheckBox, HUD_CB_FRACTIONAL_SPACING, 't');
g_hud->AddCheckBox("Frustum Patch Culling (B)", g_patchCull,
10, 90, callbackCheckBox, HUD_CB_PATCH_CULL, 'b');
g_hud->AddCheckBox("Bloom (Y)", g_bloom,
10, 110, callbackCheckBox, HUD_CB_BLOOM, 'y');
g_hud->AddCheckBox("Freeze (spc)", g_freeze,
10, 130, callbackCheckBox, HUD_CB_FREEZE, ' ');
2013-10-05 01:28:00 +00:00
g_hud->AddRadioButton(HUD_RB_SCHEME, "CATMARK", true, 10, 190, callbackScheme, 0);
g_hud->AddRadioButton(HUD_RB_SCHEME, "BILINEAR", false, 10, 210, callbackScheme, 1);
g_hud->AddCheckBox("Adaptive (`)", g_adaptive,
10, 300, callbackCheckBox, HUD_CB_ADAPTIVE, '`');
2013-10-05 01:28:00 +00:00
for (int i = 1; i < 8; ++i) {
char level[16];
sprintf(level, "Lv. %d", i);
g_hud->AddRadioButton(HUD_RB_LEVEL, level, i == g_level,
10, 320+i*20, callbackLevel, i, '0'+i);
2013-10-05 01:28:00 +00:00
}
int compute_pulldown = g_hud->AddPullDown("Compute (K)", 475, 10, 300, callbackKernel, 'k');
g_hud->AddPullDownButton(compute_pulldown, "CPU (K)", kCPU);
#ifdef OPENSUBDIV_HAS_OPENMP
g_hud->AddPullDownButton(compute_pulldown, "OPENMP", kOPENMP);
#endif
#ifdef OPENSUBDIV_HAS_TBB
g_hud->AddPullDownButton(compute_pulldown, "TBB", kTBB);
#endif
#ifdef OPENSUBDIV_HAS_CUDA
g_hud->AddPullDownButton(compute_pulldown, "CUDA", kCUDA);
#endif
#ifdef OPENSUBDIV_HAS_OPENCL
if (CLDeviceContext::HAS_CL_VERSION_1_1()) {
g_hud->AddPullDownButton(compute_pulldown, "OPENCL", kCL);
}
#endif
g_hud->AddPullDownButton(compute_pulldown, "DirectCompute", kDirectCompute);
int shading_pulldown = g_hud->AddPullDown("Shading (W)", 250, 10, 250, callbackWireframe, 'w');
g_hud->AddPullDownButton(shading_pulldown, "Wire", DISPLAY_WIRE, g_wire==DISPLAY_WIRE);
g_hud->AddPullDownButton(shading_pulldown, "Shaded", DISPLAY_SHADED, g_wire==DISPLAY_SHADED);
g_hud->AddPullDownButton(shading_pulldown, "Wire+Shaded", DISPLAY_WIRE_ON_SHADED, g_wire==DISPLAY_WIRE_ON_SHADED);
2013-10-05 01:28:00 +00:00
g_hud->AddLabel("Color (C)", -200, 10);
g_hud->AddRadioButton(HUD_RB_COLOR, "None", (g_color == COLOR_NONE),
-200, 30, callbackColor, COLOR_NONE, 'c');
g_hud->AddRadioButton(HUD_RB_COLOR, "Ptex Nearest", (g_color == COLOR_PTEX_NEAREST),
-200, 50, callbackColor, COLOR_PTEX_NEAREST, 'c');
// Commented out : we need to add a texture sampler state to make this work
//g_hud->AddRadioButton(HUD_RB_COLOR, "Ptex HW bilinear", (g_color == COLOR_PTEX_HW_BILINEAR),
// -200, 70, callbackColor, COLOR_PTEX_HW_BILINEAR, 'c');
2013-10-05 01:28:00 +00:00
g_hud->AddRadioButton(HUD_RB_COLOR, "Ptex bilinear", (g_color == COLOR_PTEX_BILINEAR),
-200, 90, callbackColor, COLOR_PTEX_BILINEAR, 'c');
g_hud->AddRadioButton(HUD_RB_COLOR, "Ptex biquadratic", (g_color == COLOR_PTEX_BIQUADRATIC),
-200, 110, callbackColor, COLOR_PTEX_BIQUADRATIC, 'c');
g_hud->AddRadioButton(HUD_RB_COLOR, "Patch type", (g_color == COLOR_PATCHTYPE),
-200, 130, callbackColor, COLOR_PATCHTYPE, 'c');
g_hud->AddRadioButton(HUD_RB_COLOR, "Patch coord", (g_color == COLOR_PATCHCOORD),
-200, 150, callbackColor, COLOR_PATCHCOORD, 'c');
g_hud->AddRadioButton(HUD_RB_COLOR, "Normal", (g_color == COLOR_NORMAL),
-200, 170, callbackColor, COLOR_NORMAL, 'c');
if (g_osdPTexDisplacement != NULL) {
g_hud->AddLabel("Displacement (D)", -200, 200);
g_hud->AddRadioButton(HUD_RB_DISPLACEMENT, "None",
(g_displacement == DISPLACEMENT_NONE),
-200, 220, callbackDisplacement, DISPLACEMENT_NONE, 'd');
// Commented out : we need to add a texture sampler state to make this work
//g_hud->AddRadioButton(HUD_RB_DISPLACEMENT, "HW bilinear",
// (g_displacement == DISPLACEMENT_HW_BILINEAR),
// -200, 240, callbackDisplacement, DISPLACEMENT_HW_BILINEAR, 'd');
2013-10-05 01:28:00 +00:00
g_hud->AddRadioButton(HUD_RB_DISPLACEMENT, "Bilinear",
(g_displacement == DISPLACEMENT_BILINEAR),
-200, 260, callbackDisplacement, DISPLACEMENT_BILINEAR, 'd');
g_hud->AddRadioButton(HUD_RB_DISPLACEMENT, "Biquadratic",
(g_displacement == DISPLACEMENT_BIQUADRATIC),
-200, 280, callbackDisplacement, DISPLACEMENT_BIQUADRATIC, 'd');
g_hud->AddLabel("Normal (N)", -200, 310);
g_hud->AddRadioButton(HUD_RB_NORMAL, "Surface",
(g_normal == NORMAL_SURFACE),
-200, 330, callbackNormal, NORMAL_SURFACE, 'n');
g_hud->AddRadioButton(HUD_RB_NORMAL, "Facet",
(g_normal == NORMAL_FACET),
-200, 350, callbackNormal, NORMAL_FACET, 'n');
g_hud->AddRadioButton(HUD_RB_NORMAL, "HW Screen space",
(g_normal == NORMAL_HW_SCREENSPACE),
-200, 370, callbackNormal, NORMAL_HW_SCREENSPACE, 'n');
g_hud->AddRadioButton(HUD_RB_NORMAL, "Screen space",
(g_normal == NORMAL_SCREENSPACE),
-200, 390, callbackNormal, NORMAL_SCREENSPACE, 'n');
g_hud->AddRadioButton(HUD_RB_NORMAL, "Biquadratic",
(g_normal == NORMAL_BIQUADRATIC),
-200, 410, callbackNormal, NORMAL_BIQUADRATIC, 'n');
g_hud->AddRadioButton(HUD_RB_NORMAL, "Biquadratic WG",
(g_normal == NORMAL_BIQUADRATIC_WG),
-200, 430, callbackNormal, NORMAL_BIQUADRATIC_WG, 'n');
}
g_hud->AddSlider("Mipmap Bias", 0, 5, 0,
-200, 450, 20, false, callbackSlider, 0);
g_hud->AddSlider("Displacement", 0, 5, 1,
2013-10-05 01:28:00 +00:00
-200, 490, 20, false, callbackSlider, 1);
2013-10-16 00:47:58 +00:00
g_hud->AddCheckBox("Seamless Mipmap", g_seamless,
-200, 530, callbackCheckBox, HUD_CB_SEAMLESS_MIPMAP, 'j');
2013-10-05 01:28:00 +00:00
}
//------------------------------------------------------------------------------
static bool
initD3D11(HWND hWnd) {
2013-10-05 01:28:00 +00:00
D3D_DRIVER_TYPE driverTypes[] = {
D3D_DRIVER_TYPE_HARDWARE,
D3D_DRIVER_TYPE_WARP,
D3D_DRIVER_TYPE_REFERENCE,
};
UINT numDriverTypes = ARRAYSIZE(driverTypes);
int width = g_width,
height = g_height;
if (g_fullscreen) {
HWND const desktop = GetDesktopWindow();
RECT rect;
GetWindowRect(desktop, &rect);
width = (int)rect.right;
height = (int)rect.bottom;
}
2013-10-05 01:28:00 +00:00
DXGI_SWAP_CHAIN_DESC hDXGISwapChainDesc;
hDXGISwapChainDesc.BufferDesc.Width = width;
hDXGISwapChainDesc.BufferDesc.Height = height;
2013-10-05 01:28:00 +00:00
hDXGISwapChainDesc.BufferDesc.RefreshRate.Numerator = 0;
hDXGISwapChainDesc.BufferDesc.RefreshRate.Denominator = 1;
hDXGISwapChainDesc.BufferDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM_SRGB;
hDXGISwapChainDesc.BufferDesc.ScanlineOrdering = DXGI_MODE_SCANLINE_ORDER_UNSPECIFIED;
hDXGISwapChainDesc.BufferDesc.Scaling = DXGI_MODE_SCALING_UNSPECIFIED;
hDXGISwapChainDesc.SampleDesc.Count = 1;
hDXGISwapChainDesc.SampleDesc.Quality = 0;
hDXGISwapChainDesc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT;
hDXGISwapChainDesc.BufferCount = 1;
hDXGISwapChainDesc.OutputWindow = hWnd;
hDXGISwapChainDesc.Windowed = !g_fullscreen;
2013-10-05 01:28:00 +00:00
hDXGISwapChainDesc.SwapEffect = DXGI_SWAP_EFFECT_DISCARD;
hDXGISwapChainDesc.Flags = DXGI_SWAP_CHAIN_FLAG_ALLOW_MODE_SWITCH;
2013-10-05 01:28:00 +00:00
// create device and swap chain
HRESULT hr;
D3D_DRIVER_TYPE hDriverType = D3D_DRIVER_TYPE_NULL;
D3D_FEATURE_LEVEL hFeatureLevel = D3D_FEATURE_LEVEL_11_0;
for(UINT driverTypeIndex=0; driverTypeIndex < numDriverTypes; driverTypeIndex++){
hDriverType = driverTypes[driverTypeIndex];
unsigned int deviceFlags = 0;
#ifndef NDEBUG
// XXX: this is problematic in some environments.
// deviceFlags |= D3D11_CREATE_DEVICE_DEBUG;
#endif
2013-10-05 01:28:00 +00:00
hr = D3D11CreateDeviceAndSwapChain(NULL,
hDriverType, NULL, deviceFlags, NULL, 0,
2013-10-05 01:28:00 +00:00
D3D11_SDK_VERSION, &hDXGISwapChainDesc,
&g_pSwapChain, &g_pd3dDevice,
&hFeatureLevel, &g_pd3dDeviceContext);
if(SUCCEEDED(hr)){
break;
}
}
if(FAILED(hr)){
MessageBoxW(hWnd, L"D3D11CreateDeviceAndSwapChain", L"Err", MB_ICONSTOP);
return false;
}
#ifndef NDEBUG
// set break points on directx errors
ID3D11Debug *d3dDebug = nullptr;
hr = g_pd3dDevice->QueryInterface(__uuidof(ID3D11Debug), (void**)&d3dDebug);
if (SUCCEEDED(hr)) {
ID3D11InfoQueue *d3dInfoQueue = nullptr;
hr = d3dDebug->QueryInterface(__uuidof(ID3D11InfoQueue), (void**)&d3dInfoQueue);
if (SUCCEEDED(hr)) {
d3dInfoQueue->SetBreakOnSeverity(D3D11_MESSAGE_SEVERITY_CORRUPTION, true);
d3dInfoQueue->SetBreakOnSeverity(D3D11_MESSAGE_SEVERITY_ERROR, true);
d3dInfoQueue->SetBreakOnSeverity(D3D11_MESSAGE_SEVERITY_WARNING, true);
D3D11_MESSAGE_ID denied[] = { D3D11_MESSAGE_ID_SETPRIVATEDATA_CHANGINGPARAMS };
D3D11_INFO_QUEUE_FILTER filter;
memset(&filter, 0, sizeof(filter));
filter.DenyList.NumIDs = _countof(denied);
filter.DenyList.pIDList = denied;
d3dInfoQueue->AddStorageFilterEntries(&filter);
d3dInfoQueue->Release();
}
d3dDebug->Release();
}
#endif
2013-10-05 01:28:00 +00:00
// create rasterizer
D3D11_RASTERIZER_DESC rasterDesc;
ZeroMemory(&rasterDesc, sizeof(rasterDesc));
rasterDesc.AntialiasedLineEnable = false;
rasterDesc.CullMode = D3D11_CULL_BACK;
2013-10-05 01:28:00 +00:00
rasterDesc.DepthBias = 0;
rasterDesc.DepthBiasClamp = 0.0f;
rasterDesc.DepthClipEnable = true;
rasterDesc.FillMode = D3D11_FILL_SOLID;
rasterDesc.FrontCounterClockwise = true;
rasterDesc.MultisampleEnable = false;
rasterDesc.ScissorEnable = false;
rasterDesc.SlopeScaledDepthBias = 0.0f;
g_pd3dDevice->CreateRasterizerState(&rasterDesc, &g_pRasterizerState);
assert(g_pRasterizerState);
__declspec(align(16))
struct Lighting {
struct Light {
float position[4];
float ambient[4];
float diffuse[4];
float specular[4];
} lightSource[2];
} lightingData = {
0.5, 0.2f, 1.0f, 0.0f,
0.1f, 0.1f, 0.1f, 1.0f,
0.7f, 0.7f, 0.7f, 1.0f,
0.8f, 0.8f, 0.8f, 1.0f,
-0.8f, 0.4f, -1.0f, 0.0f,
0.0f, 0.0f, 0.0f, 1.0f,
0.5f, 0.5f, 0.5f, 1.0f,
0.8f, 0.8f, 0.8f, 1.0f,
};
D3D11_BUFFER_DESC cbDesc;
ZeroMemory(&cbDesc, sizeof(cbDesc));
cbDesc.Usage = D3D11_USAGE_DYNAMIC;
cbDesc.BindFlags = D3D11_BIND_CONSTANT_BUFFER;
cbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
cbDesc.MiscFlags = 0;
cbDesc.ByteWidth = sizeof(lightingData);
D3D11_SUBRESOURCE_DATA initData;
initData.pSysMem = &lightingData;
g_pd3dDevice->CreateBuffer(&cbDesc, &initData, &g_pcbLighting);
assert(g_pcbLighting);
// create depth stencil state
D3D11_DEPTH_STENCIL_DESC depthStencilDesc;
ZeroMemory(&depthStencilDesc, sizeof(depthStencilDesc));
depthStencilDesc.DepthEnable = true;
depthStencilDesc.DepthWriteMask = D3D11_DEPTH_WRITE_MASK_ALL;
depthStencilDesc.DepthFunc = D3D11_COMPARISON_LESS_EQUAL;
depthStencilDesc.StencilEnable = false;
g_pd3dDevice->CreateDepthStencilState(&depthStencilDesc, &g_pDepthStencilState);
assert(g_pDepthStencilState);
#ifdef PIPELINE_STATISTICS
// Create pipeline statistics query
D3D11_QUERY_DESC queryDesc;
ZeroMemory(&queryDesc, sizeof(D3D11_QUERY_DESC));
queryDesc.Query = D3D11_QUERY_PIPELINE_STATISTICS;
g_pd3dDevice->CreateQuery(&queryDesc, &g_pipelineStatsQuery);
#endif // PIPELINE_STATISTICS
2013-10-05 01:28:00 +00:00
return true;
}
static bool
updateRenderTarget(HWND hWnd) {
2013-10-05 01:28:00 +00:00
RECT rc;
GetClientRect(hWnd, &rc);
UINT width = rc.right - rc.left;
UINT height = rc.bottom - rc.top;
if (g_pSwapChainRTV && (g_width == width) && (g_height == height)) {
return true;
}
g_width = width;
g_height = height;
g_hud->Rebuild(g_width, g_height);
SAFE_RELEASE(g_pSwapChainRTV);
g_pSwapChain->ResizeBuffers(0, g_width, g_height, DXGI_FORMAT_UNKNOWN, 0);
// get backbuffer of swap chain
ID3D11Texture2D* hpBackBuffer = NULL;
if(FAILED(g_pSwapChain->GetBuffer(0, __uuidof(ID3D11Texture2D), (void**)&hpBackBuffer))){
MessageBoxW(hWnd, L"SwpChain GetBuffer", L"Err", MB_ICONSTOP);
return false;
}
2013-10-05 01:28:00 +00:00
// create render target from the back buffer
if(FAILED(g_pd3dDevice->CreateRenderTargetView(hpBackBuffer, NULL, &g_pSwapChainRTV))){
MessageBoxW(hWnd, L"CreateRenderTargetView", L"Err", MB_ICONSTOP);
return false;
}
SAFE_RELEASE(hpBackBuffer);
// create depth buffer
D3D11_TEXTURE2D_DESC depthBufferDesc;
ZeroMemory(&depthBufferDesc, sizeof(depthBufferDesc));
depthBufferDesc.Width = g_width;
depthBufferDesc.Height = g_height;
depthBufferDesc.MipLevels = 1;
depthBufferDesc.ArraySize = 1;
depthBufferDesc.Format = DXGI_FORMAT_D24_UNORM_S8_UINT;
depthBufferDesc.SampleDesc.Count = 1;
depthBufferDesc.SampleDesc.Quality = 0;
depthBufferDesc.Usage = D3D11_USAGE_DEFAULT;
depthBufferDesc.BindFlags = D3D11_BIND_DEPTH_STENCIL;
depthBufferDesc.CPUAccessFlags = 0;
depthBufferDesc.MiscFlags = 0;
g_pd3dDevice->CreateTexture2D(&depthBufferDesc, NULL, &g_pDepthStencilBuffer);
assert(g_pDepthStencilBuffer);
D3D11_DEPTH_STENCIL_VIEW_DESC depthStencilViewDesc;
ZeroMemory(&depthStencilViewDesc, sizeof(depthStencilViewDesc));
depthStencilViewDesc.Format = DXGI_FORMAT_D24_UNORM_S8_UINT;
depthStencilViewDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2D;
depthStencilViewDesc.Texture2D.MipSlice = 0;
g_pd3dDevice->CreateDepthStencilView(g_pDepthStencilBuffer, &depthStencilViewDesc, &g_pDepthStencilView);
assert(g_pDepthStencilView);
// set device context to the render target
g_pd3dDeviceContext->OMSetRenderTargets(1, &g_pSwapChainRTV, g_pDepthStencilView);
// init viewport
D3D11_VIEWPORT vp;
vp.TopLeftX = 0;
vp.TopLeftY = 0;
vp.Width = (float)g_width;
vp.Height = (float)g_height;
vp.MinDepth = 0.0f;
vp.MaxDepth = 1.0f;
g_pd3dDeviceContext->RSSetViewports(1, &vp);
2013-10-05 01:28:00 +00:00
return true;
}
//------------------------------------------------------------------------------
static void
callbackError(OpenSubdiv::Far::ErrorType err, const char *message) {
2013-10-05 01:28:00 +00:00
std::ostringstream s;
s << "Error: " << err << "\n";
2013-10-05 01:28:00 +00:00
s << message;
OutputDebugString(s.str().c_str());
}
//------------------------------------------------------------------------------
static LRESULT WINAPI
msgProc(HWND hWnd, UINT msg, WPARAM wParam, LPARAM lParam) {
2013-10-05 01:28:00 +00:00
switch(msg)
{
case WM_KEYDOWN:
keyboard(MapVirtualKey(UINT(wParam), MAPVK_VK_TO_CHAR));
break;
case WM_DESTROY:
quit();
return 0;
case WM_MOUSEMOVE:
motion(LOWORD(lParam), HIWORD(lParam));
return 0;
case WM_LBUTTONDOWN:
mouse(0, 1, LOWORD(lParam), HIWORD(lParam));
return 0;
case WM_LBUTTONUP:
mouse(0, 0, LOWORD(lParam), HIWORD(lParam));
return 0;
case WM_MBUTTONDOWN:
mouse(1, 1, LOWORD(lParam), HIWORD(lParam));
return 0;
case WM_MBUTTONUP:
mouse(1, 0, LOWORD(lParam), HIWORD(lParam));
return 0;
case WM_RBUTTONDOWN:
mouse(2, 1, LOWORD(lParam), HIWORD(lParam));
return 0;
case WM_RBUTTONUP:
mouse(2, 0, LOWORD(lParam), HIWORD(lParam));
return 0;
case WM_PAINT:
ValidateRect(hWnd, NULL);
return 0;
}
return DefWindowProc(hWnd, msg, wParam, lParam);
}
static std::vector<std::string>
tokenize(std::string const & src) {
2013-10-05 01:28:00 +00:00
std::vector<std::string> result;
std::stringstream input(src);
std::copy(std::istream_iterator<std::string>(input),
std::istream_iterator<std::string>(),
std::back_inserter< std::vector<std::string> >(result));
return result;
}
//------------------------------------------------------------------------------
void usage(const char *program) {
printf("Usage: %s [options] <color.ptx> [<displacement.ptx>] [occlusion.ptx>] "
"[specular.ptx] [pose.obj]...\n", program);
printf("Options: -l level : subdivision level\n");
printf(" -c count : frame count until exit (for profiler)\n");
printf(" -d <diffseEnvMap.hdr> : diffuse environment map for IBL\n");
printf(" -e <specularEnvMap.hdr> : specular environment map for IBL\n");
printf(" -s <shaderfile.glsl> : custom shader file\n");
printf(" -y : Y-up model\n");
printf(" -m level : max mimmap level (default=10)\n");
printf(" -x <ptex limit MB> : ptex target memory size\n");
printf(" --disp <scale> : Displacment scale\n");
}
//------------------------------------------------------------------------------
2013-10-05 01:28:00 +00:00
int WINAPI
WinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPTSTR lpCmdLine, int nCmdShow) {
2013-10-05 01:28:00 +00:00
// register window class
TCHAR szWindowClass[] = "OPENSUBDIV_EXAMPLE";
WNDCLASS wcex;
wcex.style = CS_HREDRAW | CS_VREDRAW;
wcex.lpfnWndProc = msgProc;
wcex.cbClsExtra = 0;
wcex.cbWndExtra = 0;
wcex.hInstance = hInstance;
wcex.hIcon = NULL;
wcex.hCursor = LoadCursor(NULL, IDC_ARROW);
2013-10-05 01:28:00 +00:00
wcex.hbrBackground = (HBRUSH)(COLOR_WINDOW+1);
wcex.lpszMenuName = NULL;
wcex.lpszClassName = szWindowClass;
RegisterClass(&wcex);
// crete window
RECT rect = { 0, 0, g_width, g_height };
AdjustWindowRect(&rect, WS_OVERLAPPEDWINDOW, FALSE);
static const char windowTitle[] = "OpenSubdiv dxPtexViewer " OPENSUBDIV_VERSION_STRING;
2013-10-05 01:28:00 +00:00
HWND hWnd = CreateWindow(szWindowClass,
windowTitle,
2013-10-05 01:28:00 +00:00
WS_OVERLAPPEDWINDOW | WS_VISIBLE,
CW_USEDEFAULT,
CW_USEDEFAULT,
rect.right - rect.left,
rect.bottom - rect.top,
NULL,
NULL,
hInstance,
NULL);
std::vector<std::string> argv = tokenize(lpCmdLine);
std::vector<std::string> animobjs;
const char *diffuseEnvironmentMap = NULL, *specularEnvironmentMap = NULL;
const char *colorFilename = NULL, *displacementFilename = NULL,
*occlusionFilename = NULL, *specularFilename = NULL;
int memLimit;
2013-10-05 01:28:00 +00:00
for (int i = 0; i < (int)argv.size(); ++i) {
if (strstr(argv[i].c_str(), ".obj"))
animobjs.push_back(argv[i]);
else if (argv[i] == "-l")
g_level = atoi(argv[++i].c_str());
else if (argv[i] == "-c")
g_repeatCount = atoi(argv[++i].c_str());
else if (argv[i] == "-d")
diffuseEnvironmentMap = argv[++i].c_str();
else if (argv[i] == "-e")
specularEnvironmentMap = argv[++i].c_str();
else if (argv[i] == "-f")
g_fullscreen = true;
2013-10-05 01:28:00 +00:00
else if (argv[i] == "-y")
g_yup = true;
else if (argv[i] == "-m")
g_maxMipmapLevels = atoi(argv[++i].c_str());
else if (argv[i] == "-x")
memLimit = atoi(argv[++i].c_str());
2013-10-05 01:28:00 +00:00
else if (argv[i] == "--disp")
g_displacementScale = (float)atof(argv[++i].c_str());
else if (colorFilename == NULL)
colorFilename = argv[i].c_str();
else if (displacementFilename == NULL) {
displacementFilename = argv[i].c_str();
g_displacement = DISPLACEMENT_BILINEAR;
g_normal = NORMAL_BIQUADRATIC;
} else if (occlusionFilename == NULL) {
occlusionFilename = argv[i].c_str();
g_occlusion = 1;
} else if (specularFilename == NULL) {
specularFilename = argv[i].c_str();
g_specular = 1;
}
}
OpenSubdiv::Far::SetErrorCallback(callbackError);
2013-10-05 01:28:00 +00:00
g_ptexColorFilename = colorFilename;
if (g_ptexColorFilename == NULL) {
usage(argv[0].c_str());
2013-10-05 01:28:00 +00:00
return 1;
}
initD3D11(hWnd);
createOsdMesh(g_level, g_kernel);
// load ptex files
g_osdPTexImage = createPtex(colorFilename, memLimit);
if (displacementFilename)
g_osdPTexDisplacement = createPtex(displacementFilename, memLimit);
if (occlusionFilename)
g_osdPTexOcclusion = createPtex(occlusionFilename, memLimit);
if (specularFilename)
g_osdPTexSpecular = createPtex(specularFilename, memLimit);
g_ptexMemoryUsage =
(g_osdPTexImage ? g_osdPTexImage->GetMemoryUsage() : 0)
+ (g_osdPTexDisplacement ? g_osdPTexDisplacement->GetMemoryUsage() : 0)
+ (g_osdPTexOcclusion ? g_osdPTexOcclusion->GetMemoryUsage() : 0)
+ (g_osdPTexSpecular ? g_osdPTexSpecular->GetMemoryUsage() : 0);
// load animation obj sequences (optional)
if (!animobjs.empty()) {
for (int i = 0; i < (int)animobjs.size(); ++i) {
std::ifstream ifs(animobjs[i].c_str());
if (ifs) {
std::stringstream ss;
ss << ifs.rdbuf();
ifs.close();
printf("Reading %s\r", animobjs[i].c_str());
std::string str = ss.str();
Shape *shape = Shape::parseObj(str.c_str(), kCatmark);
if (shape->verts.size() != g_positions.size()) {
printf("Error: vertex count doesn't match.\n");
goto end;
}
g_animPositions.push_back(shape->verts);
delete shape;
} else {
printf("Error in reading %s\n", animobjs[i].c_str());
goto end;
}
}
printf("\n");
}
// IBL textures
if (diffuseEnvironmentMap) {
g_diffuseEnvironmentMap =
createHDRTexture(g_pd3dDevice, diffuseEnvironmentMap);
assert(g_diffuseEnvironmentMap);
D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc;
ZeroMemory(&srvDesc, sizeof(srvDesc));
srvDesc.Format = DXGI_FORMAT_R32G32B32A32_FLOAT;
srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D;
srvDesc.Texture2D.MostDetailedMip = 0;
srvDesc.Texture2D.MipLevels = 1;
g_pd3dDevice->CreateShaderResourceView(g_diffuseEnvironmentMap, &srvDesc, &g_diffuseEnvironmentSRV);
assert(g_diffuseEnvironmentSRV);
}
if (specularEnvironmentMap) {
g_specularEnvironmentMap =
createHDRTexture(g_pd3dDevice, specularEnvironmentMap);
assert(g_specularEnvironmentMap);
D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc;
ZeroMemory(&srvDesc, sizeof(srvDesc));
srvDesc.Format = DXGI_FORMAT_R32G32B32A32_FLOAT;
srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D;
srvDesc.Texture2D.MostDetailedMip = 0;
srvDesc.Texture2D.MipLevels = 1;
g_pd3dDevice->CreateShaderResourceView(g_specularEnvironmentMap, &srvDesc, &g_specularEnvironmentSRV);
assert(g_specularEnvironmentSRV);
}
if (g_diffuseEnvironmentMap || g_specularEnvironmentMap) {
// texture sampler
D3D11_SAMPLER_DESC samplerDesc;
ZeroMemory(&samplerDesc, sizeof(samplerDesc));
samplerDesc.Filter = D3D11_FILTER_MIN_MAG_LINEAR_MIP_POINT;
samplerDesc.AddressU = samplerDesc.AddressV = samplerDesc.AddressW = D3D11_TEXTURE_ADDRESS_WRAP;
samplerDesc.MaxAnisotropy = 0;
samplerDesc.ComparisonFunc = D3D11_COMPARISON_NEVER;
samplerDesc.MinLOD = 0;
samplerDesc.MaxLOD = D3D11_FLOAT32_MAX;
samplerDesc.BorderColor[0] = samplerDesc.BorderColor[1] = samplerDesc.BorderColor[2] = samplerDesc.BorderColor[3] = 0.0f;
g_pd3dDevice->CreateSamplerState(&samplerDesc, &g_iblSampler);
g_sky = new Sky(g_pd3dDevice, g_specularEnvironmentMap!=0 ?
g_specularEnvironmentMap : g_diffuseEnvironmentMap);
}
initHUD();
2013-10-05 01:28:00 +00:00
fitFrame();
// main loop
while (g_bDone == false) {
MSG msg;
ZeroMemory(&msg, sizeof(msg));
while (msg.message != WM_QUIT) {
while (PeekMessage(&msg, NULL, 0U, 0U, PM_REMOVE)) {
if (msg.message == WM_QUIT) goto end;
TranslateMessage(&msg);
DispatchMessage(&msg);
}
if (!g_freeze)
2013-10-05 01:28:00 +00:00
g_frame++;
updateGeom();
updateRenderTarget(hWnd);
display();
}
}
end:
quit();
return 0;
}
//------------------------------------------------------------------------------