diff options
author | Andrew Knight <andrew.knight@digia.com> | 2013-04-03 19:57:09 +0300 |
---|---|---|
committer | The Qt Project <gerrit-noreply@qt-project.org> | 2013-04-08 11:35:44 +0200 |
commit | 710ad8ce1bd5d01ce048851d210ac3831ca17dde (patch) | |
tree | bb3f645e57a7887341113275c9caa2583f7e2dfa /src/3rdparty/angle/src/libGLESv2/renderer | |
parent | 8307ab784eb4d8d22441ee61f02cb381684ca3a2 (diff) |
Upgrade ANGLE to DX11 Proto
Upgrades ANGLE to dx11proto (dx11-MRT-support tag), which splits out support
for DirectX9 & DirectX11. The DX9 codepath is used by default;
CONFIG+=angle_d3d11 must be passed to the ANGLE project to build for DX11.
Existing patches to ANGLE have been updated (or removed if no longer
needed), and a patch has been added to make DX9/DX11 codepaths mutually
exclusive.
Change-Id: Ibe13befadb94f04883eca449d0ee1f0da955ff92
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Axel Waggershauser <awagger@gmail.com>
Diffstat (limited to 'src/3rdparty/angle/src/libGLESv2/renderer')
85 files changed, 21873 insertions, 0 deletions
diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Blit.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Blit.cpp new file mode 100644 index 0000000000..2a3ce39c63 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Blit.cpp @@ -0,0 +1,595 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2010 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Blit.cpp: Surface copy utility class. + +#include "libGLESv2/renderer/Blit.h" + +#include "libGLESv2/main.h" +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/renderer/TextureStorage9.h" +#include "libGLESv2/renderer/RenderTarget9.h" +#include "libGLESv2/renderer/Renderer9.h" +#include "libGLESv2/Framebuffer.h" +#include "libGLESv2/Renderbuffer.h" + +namespace +{ +#include "libGLESv2/renderer/shaders/compiled/standardvs.h" +#include "libGLESv2/renderer/shaders/compiled/flipyvs.h" +#include "libGLESv2/renderer/shaders/compiled/passthroughps.h" +#include "libGLESv2/renderer/shaders/compiled/luminanceps.h" +#include "libGLESv2/renderer/shaders/compiled/componentmaskps.h" + +const BYTE* const g_shaderCode[] = +{ + g_vs20_standardvs, + g_vs20_flipyvs, + g_ps20_passthroughps, + g_ps20_luminanceps, + g_ps20_componentmaskps +}; + +const size_t g_shaderSize[] = +{ + sizeof(g_vs20_standardvs), + sizeof(g_vs20_flipyvs), + sizeof(g_ps20_passthroughps), + sizeof(g_ps20_luminanceps), + sizeof(g_ps20_componentmaskps) +}; +} + +namespace rx +{ +Blit::Blit(rx::Renderer9 *renderer) + : mRenderer(renderer), mQuadVertexBuffer(NULL), mQuadVertexDeclaration(NULL), mSavedStateBlock(NULL), mSavedRenderTarget(NULL), mSavedDepthStencil(NULL) +{ + initGeometry(); + memset(mCompiledShaders, 0, sizeof(mCompiledShaders)); +} + +Blit::~Blit() +{ + if (mSavedStateBlock) mSavedStateBlock->Release(); + if (mQuadVertexBuffer) mQuadVertexBuffer->Release(); + if (mQuadVertexDeclaration) mQuadVertexDeclaration->Release(); + + for (int i = 0; i < SHADER_COUNT; i++) + { + if (mCompiledShaders[i]) + { + mCompiledShaders[i]->Release(); + } + } +} + +void Blit::initGeometry() +{ + static const float quad[] = + { + -1, -1, + -1, 1, + 1, -1, + 1, 1 + }; + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + HRESULT result = device->CreateVertexBuffer(sizeof(quad), D3DUSAGE_WRITEONLY, 0, D3DPOOL_DEFAULT, &mQuadVertexBuffer, NULL); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + return gl::error(GL_OUT_OF_MEMORY); + } + + void *lockPtr = NULL; + result = mQuadVertexBuffer->Lock(0, 0, &lockPtr, 0); + + if (FAILED(result) || lockPtr == NULL) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + return gl::error(GL_OUT_OF_MEMORY); + } + + memcpy(lockPtr, quad, sizeof(quad)); + mQuadVertexBuffer->Unlock(); + + static const D3DVERTEXELEMENT9 elements[] = + { + { 0, 0, D3DDECLTYPE_FLOAT2, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_POSITION, 0 }, + D3DDECL_END() + }; + + result = device->CreateVertexDeclaration(elements, &mQuadVertexDeclaration); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + return gl::error(GL_OUT_OF_MEMORY); + } +} + +template <class D3DShaderType> +bool Blit::setShader(ShaderId source, const char *profile, + D3DShaderType *(rx::Renderer9::*createShader)(const DWORD *, size_t length), + HRESULT (WINAPI IDirect3DDevice9::*setShader)(D3DShaderType*)) +{ + IDirect3DDevice9 *device = mRenderer->getDevice(); + + D3DShaderType *shader; + + if (mCompiledShaders[source] != NULL) + { + shader = static_cast<D3DShaderType*>(mCompiledShaders[source]); + } + else + { + const BYTE* shaderCode = g_shaderCode[source]; + size_t shaderSize = g_shaderSize[source]; + + shader = (mRenderer->*createShader)(reinterpret_cast<const DWORD*>(shaderCode), shaderSize); + if (!shader) + { + ERR("Failed to create shader for blit operation"); + return false; + } + + mCompiledShaders[source] = shader; + } + + HRESULT hr = (device->*setShader)(shader); + + if (FAILED(hr)) + { + ERR("Failed to set shader for blit operation"); + return false; + } + + return true; +} + +bool Blit::setVertexShader(ShaderId shader) +{ + return setShader<IDirect3DVertexShader9>(shader, "vs_2_0", &rx::Renderer9::createVertexShader, &IDirect3DDevice9::SetVertexShader); +} + +bool Blit::setPixelShader(ShaderId shader) +{ + return setShader<IDirect3DPixelShader9>(shader, "ps_2_0", &rx::Renderer9::createPixelShader, &IDirect3DDevice9::SetPixelShader); +} + +RECT Blit::getSurfaceRect(IDirect3DSurface9 *surface) const +{ + D3DSURFACE_DESC desc; + surface->GetDesc(&desc); + + RECT rect; + rect.left = 0; + rect.top = 0; + rect.right = desc.Width; + rect.bottom = desc.Height; + + return rect; +} + +bool Blit::boxFilter(IDirect3DSurface9 *source, IDirect3DSurface9 *dest) +{ + IDirect3DTexture9 *texture = copySurfaceToTexture(source, getSurfaceRect(source)); + if (!texture) + { + return false; + } + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + saveState(); + + device->SetTexture(0, texture); + device->SetRenderTarget(0, dest); + + setVertexShader(SHADER_VS_STANDARD); + setPixelShader(SHADER_PS_PASSTHROUGH); + + setCommonBlitState(); + device->SetSamplerState(0, D3DSAMP_MINFILTER, D3DTEXF_LINEAR); + device->SetSamplerState(0, D3DSAMP_MAGFILTER, D3DTEXF_LINEAR); + + setViewport(getSurfaceRect(dest), 0, 0); + + render(); + + texture->Release(); + + restoreState(); + + return true; +} + +bool Blit::copy(gl::Framebuffer *framebuffer, const RECT &sourceRect, GLenum destFormat, GLint xoffset, GLint yoffset, TextureStorageInterface2D *storage, GLint level) +{ + RenderTarget9 *renderTarget = NULL; + IDirect3DSurface9 *source = NULL; + gl::Renderbuffer *colorbuffer = framebuffer->getColorbuffer(0); + + if (colorbuffer) + { + renderTarget = RenderTarget9::makeRenderTarget9(colorbuffer->getRenderTarget()); + } + + if (renderTarget) + { + source = renderTarget->getSurface(); + } + + if (!source) + { + ERR("Failed to retrieve the render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + TextureStorage9_2D *storage9 = TextureStorage9_2D::makeTextureStorage9_2D(storage->getStorageInstance()); + IDirect3DSurface9 *destSurface = storage9->getSurfaceLevel(level, true); + bool result = false; + + if (destSurface) + { + result = copy(source, sourceRect, destFormat, xoffset, yoffset, destSurface); + destSurface->Release(); + } + + source->Release(); + return result; +} + +bool Blit::copy(gl::Framebuffer *framebuffer, const RECT &sourceRect, GLenum destFormat, GLint xoffset, GLint yoffset, TextureStorageInterfaceCube *storage, GLenum target, GLint level) +{ + RenderTarget9 *renderTarget = NULL; + IDirect3DSurface9 *source = NULL; + gl::Renderbuffer *colorbuffer = framebuffer->getColorbuffer(0); + + if (colorbuffer) + { + renderTarget = RenderTarget9::makeRenderTarget9(colorbuffer->getRenderTarget()); + } + + if (renderTarget) + { + source = renderTarget->getSurface(); + } + + if (!source) + { + ERR("Failed to retrieve the render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + TextureStorage9_Cube *storage9 = TextureStorage9_Cube::makeTextureStorage9_Cube(storage->getStorageInstance()); + IDirect3DSurface9 *destSurface = storage9->getCubeMapSurface(target, level, true); + bool result = false; + + if (destSurface) + { + result = copy(source, sourceRect, destFormat, xoffset, yoffset, destSurface); + destSurface->Release(); + } + + source->Release(); + return result; +} + +bool Blit::copy(IDirect3DSurface9 *source, const RECT &sourceRect, GLenum destFormat, GLint xoffset, GLint yoffset, IDirect3DSurface9 *dest) +{ + if (!dest) + { + return false; + } + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + D3DSURFACE_DESC sourceDesc; + D3DSURFACE_DESC destDesc; + source->GetDesc(&sourceDesc); + dest->GetDesc(&destDesc); + + if (sourceDesc.Format == destDesc.Format && destDesc.Usage & D3DUSAGE_RENDERTARGET && + d3d9_gl::IsFormatChannelEquivalent(destDesc.Format, destFormat)) // Can use StretchRect + { + RECT destRect = {xoffset, yoffset, xoffset + (sourceRect.right - sourceRect.left), yoffset + (sourceRect.bottom - sourceRect.top)}; + HRESULT result = device->StretchRect(source, &sourceRect, dest, &destRect, D3DTEXF_POINT); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + return gl::error(GL_OUT_OF_MEMORY, false); + } + } + else + { + return formatConvert(source, sourceRect, destFormat, xoffset, yoffset, dest); + } + return true; +} + +bool Blit::formatConvert(IDirect3DSurface9 *source, const RECT &sourceRect, GLenum destFormat, GLint xoffset, GLint yoffset, IDirect3DSurface9 *dest) +{ + IDirect3DTexture9 *texture = copySurfaceToTexture(source, sourceRect); + if (!texture) + { + return false; + } + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + saveState(); + + device->SetTexture(0, texture); + device->SetRenderTarget(0, dest); + + setViewport(sourceRect, xoffset, yoffset); + + setCommonBlitState(); + if (setFormatConvertShaders(destFormat)) + { + render(); + } + + texture->Release(); + + restoreState(); + + return true; +} + +bool Blit::setFormatConvertShaders(GLenum destFormat) +{ + bool okay = setVertexShader(SHADER_VS_STANDARD); + + switch (destFormat) + { + default: UNREACHABLE(); + case GL_RGBA: + case GL_BGRA_EXT: + case GL_RGB: + case GL_ALPHA: + okay = okay && setPixelShader(SHADER_PS_COMPONENTMASK); + break; + + case GL_LUMINANCE: + case GL_LUMINANCE_ALPHA: + okay = okay && setPixelShader(SHADER_PS_LUMINANCE); + break; + } + + if (!okay) + { + return false; + } + + enum { X = 0, Y = 1, Z = 2, W = 3 }; + + // The meaning of this constant depends on the shader that was selected. + // See the shader assembly code above for details. + float psConst0[4] = { 0, 0, 0, 0 }; + + switch (destFormat) + { + default: UNREACHABLE(); + case GL_RGBA: + case GL_BGRA_EXT: + psConst0[X] = 1; + psConst0[Z] = 1; + break; + + case GL_RGB: + psConst0[X] = 1; + psConst0[W] = 1; + break; + + case GL_ALPHA: + psConst0[Z] = 1; + break; + + case GL_LUMINANCE: + psConst0[Y] = 1; + break; + + case GL_LUMINANCE_ALPHA: + psConst0[X] = 1; + break; + } + + mRenderer->getDevice()->SetPixelShaderConstantF(0, psConst0, 1); + + return true; +} + +IDirect3DTexture9 *Blit::copySurfaceToTexture(IDirect3DSurface9 *surface, const RECT &sourceRect) +{ + if (!surface) + { + return NULL; + } + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + D3DSURFACE_DESC sourceDesc; + surface->GetDesc(&sourceDesc); + + // Copy the render target into a texture + IDirect3DTexture9 *texture; + HRESULT result = device->CreateTexture(sourceRect.right - sourceRect.left, sourceRect.bottom - sourceRect.top, 1, D3DUSAGE_RENDERTARGET, sourceDesc.Format, D3DPOOL_DEFAULT, &texture, NULL); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + return gl::error(GL_OUT_OF_MEMORY, (IDirect3DTexture9*)NULL); + } + + IDirect3DSurface9 *textureSurface; + result = texture->GetSurfaceLevel(0, &textureSurface); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + texture->Release(); + return gl::error(GL_OUT_OF_MEMORY, (IDirect3DTexture9*)NULL); + } + + mRenderer->endScene(); + result = device->StretchRect(surface, &sourceRect, textureSurface, NULL, D3DTEXF_NONE); + + textureSurface->Release(); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + texture->Release(); + return gl::error(GL_OUT_OF_MEMORY, (IDirect3DTexture9*)NULL); + } + + return texture; +} + +void Blit::setViewport(const RECT &sourceRect, GLint xoffset, GLint yoffset) +{ + IDirect3DDevice9 *device = mRenderer->getDevice(); + + D3DVIEWPORT9 vp; + vp.X = xoffset; + vp.Y = yoffset; + vp.Width = sourceRect.right - sourceRect.left; + vp.Height = sourceRect.bottom - sourceRect.top; + vp.MinZ = 0.0f; + vp.MaxZ = 1.0f; + device->SetViewport(&vp); + + float halfPixelAdjust[4] = { -1.0f/vp.Width, 1.0f/vp.Height, 0, 0 }; + device->SetVertexShaderConstantF(0, halfPixelAdjust, 1); +} + +void Blit::setCommonBlitState() +{ + IDirect3DDevice9 *device = mRenderer->getDevice(); + + device->SetDepthStencilSurface(NULL); + + device->SetRenderState(D3DRS_FILLMODE, D3DFILL_SOLID); + device->SetRenderState(D3DRS_ALPHATESTENABLE, FALSE); + device->SetRenderState(D3DRS_ALPHABLENDENABLE, FALSE); + device->SetRenderState(D3DRS_CULLMODE, D3DCULL_NONE); + device->SetRenderState(D3DRS_CLIPPLANEENABLE, 0); + device->SetRenderState(D3DRS_COLORWRITEENABLE, D3DCOLORWRITEENABLE_ALPHA | D3DCOLORWRITEENABLE_BLUE | D3DCOLORWRITEENABLE_GREEN | D3DCOLORWRITEENABLE_RED); + device->SetRenderState(D3DRS_SRGBWRITEENABLE, FALSE); + device->SetRenderState(D3DRS_SCISSORTESTENABLE, FALSE); + + device->SetSamplerState(0, D3DSAMP_MAGFILTER, D3DTEXF_POINT); + device->SetSamplerState(0, D3DSAMP_MINFILTER, D3DTEXF_POINT); + device->SetSamplerState(0, D3DSAMP_SRGBTEXTURE, FALSE); + device->SetSamplerState(0, D3DSAMP_ADDRESSU, D3DTADDRESS_CLAMP); + device->SetSamplerState(0, D3DSAMP_ADDRESSV, D3DTADDRESS_CLAMP); + + RECT scissorRect = {0}; // Scissoring is disabled for flipping, but we need this to capture and restore the old rectangle + device->SetScissorRect(&scissorRect); + + for(int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + device->SetStreamSourceFreq(i, 1); + } +} + +void Blit::render() +{ + IDirect3DDevice9 *device = mRenderer->getDevice(); + + HRESULT hr = device->SetStreamSource(0, mQuadVertexBuffer, 0, 2 * sizeof(float)); + hr = device->SetVertexDeclaration(mQuadVertexDeclaration); + + mRenderer->startScene(); + hr = device->DrawPrimitive(D3DPT_TRIANGLESTRIP, 0, 2); +} + +void Blit::saveState() +{ + IDirect3DDevice9 *device = mRenderer->getDevice(); + + HRESULT hr; + + device->GetDepthStencilSurface(&mSavedDepthStencil); + device->GetRenderTarget(0, &mSavedRenderTarget); + + if (mSavedStateBlock == NULL) + { + hr = device->BeginStateBlock(); + ASSERT(SUCCEEDED(hr) || hr == D3DERR_OUTOFVIDEOMEMORY || hr == E_OUTOFMEMORY); + + setCommonBlitState(); + + static const float dummyConst[4] = { 0, 0, 0, 0 }; + + device->SetVertexShader(NULL); + device->SetVertexShaderConstantF(0, dummyConst, 1); + device->SetPixelShader(NULL); + device->SetPixelShaderConstantF(0, dummyConst, 1); + + D3DVIEWPORT9 dummyVp; + dummyVp.X = 0; + dummyVp.Y = 0; + dummyVp.Width = 1; + dummyVp.Height = 1; + dummyVp.MinZ = 0; + dummyVp.MaxZ = 1; + + device->SetViewport(&dummyVp); + + device->SetTexture(0, NULL); + + device->SetStreamSource(0, mQuadVertexBuffer, 0, 0); + + device->SetVertexDeclaration(mQuadVertexDeclaration); + + hr = device->EndStateBlock(&mSavedStateBlock); + ASSERT(SUCCEEDED(hr) || hr == D3DERR_OUTOFVIDEOMEMORY || hr == E_OUTOFMEMORY); + } + + ASSERT(mSavedStateBlock != NULL); + + if (mSavedStateBlock != NULL) + { + hr = mSavedStateBlock->Capture(); + ASSERT(SUCCEEDED(hr)); + } +} + +void Blit::restoreState() +{ + IDirect3DDevice9 *device = mRenderer->getDevice(); + + device->SetDepthStencilSurface(mSavedDepthStencil); + if (mSavedDepthStencil != NULL) + { + mSavedDepthStencil->Release(); + mSavedDepthStencil = NULL; + } + + device->SetRenderTarget(0, mSavedRenderTarget); + if (mSavedRenderTarget != NULL) + { + mSavedRenderTarget->Release(); + mSavedRenderTarget = NULL; + } + + ASSERT(mSavedStateBlock != NULL); + + if (mSavedStateBlock != NULL) + { + mSavedStateBlock->Apply(); + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Blit.h b/src/3rdparty/angle/src/libGLESv2/renderer/Blit.h new file mode 100644 index 0000000000..3718028e66 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Blit.h @@ -0,0 +1,94 @@ +// +// Copyright (c) 2002-2010 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Blit.cpp: Surface copy utility class. + +#ifndef LIBGLESV2_BLIT_H_ +#define LIBGLESV2_BLIT_H_ + +#include "common/angleutils.h" + +namespace gl +{ +class Framebuffer; +} + +namespace rx +{ +class Renderer9; +class TextureStorageInterface2D; +class TextureStorageInterfaceCube; + +class Blit +{ + public: + explicit Blit(Renderer9 *renderer); + ~Blit(); + + // Copy from source surface to dest surface. + // sourceRect, xoffset, yoffset are in D3D coordinates (0,0 in upper-left) + bool copy(gl::Framebuffer *framebuffer, const RECT &sourceRect, GLenum destFormat, GLint xoffset, GLint yoffset, TextureStorageInterface2D *storage, GLint level); + bool copy(gl::Framebuffer *framebuffer, const RECT &sourceRect, GLenum destFormat, GLint xoffset, GLint yoffset, TextureStorageInterfaceCube *storage, GLenum target, GLint level); + + // Copy from source surface to dest surface. + // sourceRect, xoffset, yoffset are in D3D coordinates (0,0 in upper-left) + // source is interpreted as RGBA and destFormat specifies the desired result format. For example, if destFormat = GL_RGB, the alpha channel will be forced to 0. + bool formatConvert(IDirect3DSurface9 *source, const RECT &sourceRect, GLenum destFormat, GLint xoffset, GLint yoffset, IDirect3DSurface9 *dest); + + // 2x2 box filter sample from source to dest. + // Requires that source is RGB(A) and dest has the same format as source. + bool boxFilter(IDirect3DSurface9 *source, IDirect3DSurface9 *dest); + + private: + rx::Renderer9 *mRenderer; + + IDirect3DVertexBuffer9 *mQuadVertexBuffer; + IDirect3DVertexDeclaration9 *mQuadVertexDeclaration; + + void initGeometry(); + + bool setFormatConvertShaders(GLenum destFormat); + + bool copy(IDirect3DSurface9 *source, const RECT &sourceRect, GLenum destFormat, GLint xoffset, GLint yoffset, IDirect3DSurface9 *dest); + IDirect3DTexture9 *copySurfaceToTexture(IDirect3DSurface9 *surface, const RECT &sourceRect); + void setViewport(const RECT &sourceRect, GLint xoffset, GLint yoffset); + void setCommonBlitState(); + RECT getSurfaceRect(IDirect3DSurface9 *surface) const; + + // This enum is used to index mCompiledShaders and mShaderSource. + enum ShaderId + { + SHADER_VS_STANDARD, + SHADER_VS_FLIPY, + SHADER_PS_PASSTHROUGH, + SHADER_PS_LUMINANCE, + SHADER_PS_COMPONENTMASK, + SHADER_COUNT + }; + + // This actually contains IDirect3DVertexShader9 or IDirect3DPixelShader9 casted to IUnknown. + IUnknown *mCompiledShaders[SHADER_COUNT]; + + template <class D3DShaderType> + bool setShader(ShaderId source, const char *profile, + D3DShaderType *(Renderer9::*createShader)(const DWORD *, size_t length), + HRESULT (WINAPI IDirect3DDevice9::*setShader)(D3DShaderType*)); + + bool setVertexShader(ShaderId shader); + bool setPixelShader(ShaderId shader); + void render(); + + void saveState(); + void restoreState(); + IDirect3DStateBlock9 *mSavedStateBlock; + IDirect3DSurface9 *mSavedRenderTarget; + IDirect3DSurface9 *mSavedDepthStencil; + + DISALLOW_COPY_AND_ASSIGN(Blit); +}; +} + +#endif // LIBGLESV2_BLIT_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage.cpp new file mode 100644 index 0000000000..a49b7bab84 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage.cpp @@ -0,0 +1,40 @@ +#include "precompiled.h" +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// BufferStorage.cpp Defines the abstract BufferStorage class. + +#include "libGLESv2/renderer/BufferStorage.h" + +namespace rx +{ + +unsigned int BufferStorage::mNextSerial = 1; + +BufferStorage::BufferStorage() +{ + updateSerial(); +} + +BufferStorage::~BufferStorage() +{ +} + +unsigned int BufferStorage::getSerial() const +{ + return mSerial; +} + +void BufferStorage::updateSerial() +{ + mSerial = mNextSerial++; +} + +void BufferStorage::markBufferUsage() +{ +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage.h b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage.h new file mode 100644 index 0000000000..ace1a11bae --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage.h @@ -0,0 +1,44 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// BufferStorage.h Defines the abstract BufferStorage class. + +#ifndef LIBGLESV2_RENDERER_BUFFERSTORAGE_H_ +#define LIBGLESV2_RENDERER_BUFFERSTORAGE_H_ + +#include "common/angleutils.h" + +namespace rx +{ + +class BufferStorage +{ + public: + BufferStorage(); + virtual ~BufferStorage(); + + // The data returned is only guaranteed valid until next non-const method. + virtual void *getData() = 0; + virtual void setData(const void* data, unsigned int size, unsigned int offset) = 0; + virtual void clear() = 0; + virtual unsigned int getSize() const = 0; + virtual bool supportsDirectBinding() const = 0; + virtual void markBufferUsage(); + unsigned int getSerial() const; + + protected: + void updateSerial(); + + private: + DISALLOW_COPY_AND_ASSIGN(BufferStorage); + + unsigned int mSerial; + static unsigned int mNextSerial; +}; + +} + +#endif // LIBGLESV2_RENDERER_BUFFERSTORAGE_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage11.cpp new file mode 100644 index 0000000000..4c37bdbf60 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage11.cpp @@ -0,0 +1,342 @@ +#include "precompiled.h" +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// BufferStorage11.cpp Defines the BufferStorage11 class. + +#include "libGLESv2/renderer/BufferStorage11.h" +#include "libGLESv2/main.h" +#include "libGLESv2/renderer/Renderer11.h" + +namespace rx +{ + +BufferStorage11::BufferStorage11(Renderer11 *renderer) +{ + mRenderer = renderer; + + mStagingBuffer = NULL; + mStagingBufferSize = 0; + + mBuffer = NULL; + mBufferSize = 0; + + mSize = 0; + + mResolvedData = NULL; + mResolvedDataSize = 0; + mResolvedDataValid = false; + + mReadUsageCount = 0; + mWriteUsageCount = 0; +} + +BufferStorage11::~BufferStorage11() +{ + if (mStagingBuffer) + { + mStagingBuffer->Release(); + mStagingBuffer = NULL; + } + + if (mBuffer) + { + mBuffer->Release(); + mBuffer = NULL; + } + + if (mResolvedData) + { + free(mResolvedData); + mResolvedData = NULL; + } +} + +BufferStorage11 *BufferStorage11::makeBufferStorage11(BufferStorage *bufferStorage) +{ + ASSERT(HAS_DYNAMIC_TYPE(BufferStorage11*, bufferStorage)); + return static_cast<BufferStorage11*>(bufferStorage); +} + +void *BufferStorage11::getData() +{ + if (!mResolvedDataValid) + { + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + HRESULT result; + + if (!mStagingBuffer || mStagingBufferSize < mBufferSize) + { + if (mStagingBuffer) + { + mStagingBuffer->Release(); + mStagingBuffer = NULL; + mStagingBufferSize = 0; + } + + D3D11_BUFFER_DESC bufferDesc; + bufferDesc.ByteWidth = mSize; + bufferDesc.Usage = D3D11_USAGE_STAGING; + bufferDesc.BindFlags = 0; + bufferDesc.CPUAccessFlags = D3D11_CPU_ACCESS_READ | D3D11_CPU_ACCESS_WRITE; + bufferDesc.MiscFlags = 0; + bufferDesc.StructureByteStride = 0; + + result = device->CreateBuffer(&bufferDesc, NULL, &mStagingBuffer); + if (FAILED(result)) + { + return gl::error(GL_OUT_OF_MEMORY, (void*)NULL); + } + + mStagingBufferSize = bufferDesc.ByteWidth; + } + + if (!mResolvedData || mResolvedDataSize < mBufferSize) + { + free(mResolvedData); + mResolvedData = malloc(mSize); + mResolvedDataSize = mSize; + } + + D3D11_BOX srcBox; + srcBox.left = 0; + srcBox.right = mSize; + srcBox.top = 0; + srcBox.bottom = 1; + srcBox.front = 0; + srcBox.back = 1; + + context->CopySubresourceRegion(mStagingBuffer, 0, 0, 0, 0, mBuffer, 0, &srcBox); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + result = context->Map(mStagingBuffer, 0, D3D11_MAP_READ, 0, &mappedResource); + if (FAILED(result)) + { + return gl::error(GL_OUT_OF_MEMORY, (void*)NULL); + } + + memcpy(mResolvedData, mappedResource.pData, mSize); + + context->Unmap(mStagingBuffer, 0); + + mResolvedDataValid = true; + } + + mReadUsageCount = 0; + + return mResolvedData; +} + +void BufferStorage11::setData(const void* data, unsigned int size, unsigned int offset) +{ + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + HRESULT result; + + unsigned int requiredBufferSize = size + offset; + unsigned int requiredStagingSize = size; + bool directInitialization = offset == 0 && (!mBuffer || mBufferSize < size + offset); + + if (!directInitialization) + { + if (!mStagingBuffer || mStagingBufferSize < requiredStagingSize) + { + if (mStagingBuffer) + { + mStagingBuffer->Release(); + mStagingBuffer = NULL; + mStagingBufferSize = 0; + } + + D3D11_BUFFER_DESC bufferDesc; + bufferDesc.ByteWidth = size; + bufferDesc.Usage = D3D11_USAGE_STAGING; + bufferDesc.BindFlags = 0; + bufferDesc.CPUAccessFlags = D3D11_CPU_ACCESS_READ | D3D11_CPU_ACCESS_WRITE; + bufferDesc.MiscFlags = 0; + bufferDesc.StructureByteStride = 0; + + D3D11_SUBRESOURCE_DATA initialData; + initialData.pSysMem = data; + initialData.SysMemPitch = size; + initialData.SysMemSlicePitch = 0; + + result = device->CreateBuffer(&bufferDesc, &initialData, &mStagingBuffer); + if (FAILED(result)) + { + return gl::error(GL_OUT_OF_MEMORY); + } + + mStagingBufferSize = size; + } + else + { + D3D11_MAPPED_SUBRESOURCE mappedResource; + result = context->Map(mStagingBuffer, 0, D3D11_MAP_WRITE, 0, &mappedResource); + if (FAILED(result)) + { + return gl::error(GL_OUT_OF_MEMORY); + } + + memcpy(mappedResource.pData, data, size); + + context->Unmap(mStagingBuffer, 0); + } + } + + if (!mBuffer || mBufferSize < size + offset) + { + D3D11_BUFFER_DESC bufferDesc; + bufferDesc.ByteWidth = requiredBufferSize; + bufferDesc.Usage = D3D11_USAGE_DEFAULT; + bufferDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER | D3D11_BIND_INDEX_BUFFER; + bufferDesc.CPUAccessFlags = 0; + bufferDesc.MiscFlags = 0; + bufferDesc.StructureByteStride = 0; + + if (directInitialization) + { + // Since the data will fill the entire buffer (being larger than the initial size and having + // no offset), the buffer can be initialized with the data so no staging buffer is required + + // No longer need the old buffer + if (mBuffer) + { + mBuffer->Release(); + mBuffer = NULL; + mBufferSize = 0; + } + + D3D11_SUBRESOURCE_DATA initialData; + initialData.pSysMem = data; + initialData.SysMemPitch = size; + initialData.SysMemSlicePitch = 0; + + result = device->CreateBuffer(&bufferDesc, &initialData, &mBuffer); + if (FAILED(result)) + { + return gl::error(GL_OUT_OF_MEMORY); + } + } + else if (mBuffer && offset > 0) + { + // If offset is greater than zero and the buffer is non-null, need to preserve the data from + // the old buffer up to offset + ID3D11Buffer *newBuffer = NULL; + + result = device->CreateBuffer(&bufferDesc, NULL, &newBuffer); + if (FAILED(result)) + { + return gl::error(GL_OUT_OF_MEMORY); + } + + D3D11_BOX srcBox; + srcBox.left = 0; + srcBox.right = std::min(offset, mBufferSize); + srcBox.top = 0; + srcBox.bottom = 1; + srcBox.front = 0; + srcBox.back = 1; + + context->CopySubresourceRegion(newBuffer, 0, 0, 0, 0, mBuffer, 0, &srcBox); + + mBuffer->Release(); + mBuffer = newBuffer; + } + else + { + // Simple case, nothing needs to be copied from the old buffer to the new one, just create + // a new buffer + + // No longer need the old buffer + if (mBuffer) + { + mBuffer->Release(); + mBuffer = NULL; + mBufferSize = 0; + } + + // Create a new buffer for data storage + result = device->CreateBuffer(&bufferDesc, NULL, &mBuffer); + if (FAILED(result)) + { + return gl::error(GL_OUT_OF_MEMORY); + } + } + + updateSerial(); + mBufferSize = bufferDesc.ByteWidth; + } + + if (!directInitialization) + { + ASSERT(mStagingBuffer && mStagingBufferSize >= requiredStagingSize); + + // Data is already put into the staging buffer, copy it over to the data buffer + D3D11_BOX srcBox; + srcBox.left = 0; + srcBox.right = size; + srcBox.top = 0; + srcBox.bottom = 1; + srcBox.front = 0; + srcBox.back = 1; + + context->CopySubresourceRegion(mBuffer, 0, offset, 0, 0, mStagingBuffer, 0, &srcBox); + } + + mSize = std::max(mSize, offset + size); + + mWriteUsageCount = 0; + + mResolvedDataValid = false; +} + +void BufferStorage11::clear() +{ + mResolvedDataValid = false; + mSize = 0; +} + +unsigned int BufferStorage11::getSize() const +{ + return mSize; +} + +bool BufferStorage11::supportsDirectBinding() const +{ + return true; +} + +void BufferStorage11::markBufferUsage() +{ + mReadUsageCount++; + mWriteUsageCount++; + + static const unsigned int usageLimit = 5; + + if (mReadUsageCount > usageLimit && mResolvedData) + { + free(mResolvedData); + mResolvedData = NULL; + mResolvedDataSize = 0; + mResolvedDataValid = false; + } + + if (mReadUsageCount > usageLimit && mWriteUsageCount > usageLimit && mStagingBuffer) + { + mStagingBuffer->Release(); + mStagingBuffer = NULL; + mStagingBufferSize = 0; + } +} + +ID3D11Buffer *BufferStorage11::getBuffer() const +{ + return mBuffer; +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage11.h b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage11.h new file mode 100644 index 0000000000..b62348b0c9 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage11.h @@ -0,0 +1,56 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// BufferStorage11.h Defines the BufferStorage11 class. + +#ifndef LIBGLESV2_RENDERER_BUFFERSTORAGE11_H_ +#define LIBGLESV2_RENDERER_BUFFERSTORAGE11_H_ + +#include "libGLESv2/renderer/BufferStorage.h" + +namespace rx +{ +class Renderer11; + +class BufferStorage11 : public BufferStorage +{ + public: + explicit BufferStorage11(Renderer11 *renderer); + virtual ~BufferStorage11(); + + static BufferStorage11 *makeBufferStorage11(BufferStorage *bufferStorage); + + virtual void *getData(); + virtual void setData(const void* data, unsigned int size, unsigned int offset); + virtual void clear(); + virtual unsigned int getSize() const; + virtual bool supportsDirectBinding() const; + virtual void markBufferUsage(); + + ID3D11Buffer *getBuffer() const; + + private: + Renderer11 *mRenderer; + + ID3D11Buffer *mStagingBuffer; + unsigned int mStagingBufferSize; + + ID3D11Buffer *mBuffer; + unsigned int mBufferSize; + + unsigned int mSize; + + void *mResolvedData; + unsigned int mResolvedDataSize; + bool mResolvedDataValid; + + unsigned int mReadUsageCount; + unsigned int mWriteUsageCount; +}; + +} + +#endif // LIBGLESV2_RENDERER_BUFFERSTORAGE11_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage9.cpp new file mode 100644 index 0000000000..7fc14fc073 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage9.cpp @@ -0,0 +1,75 @@ +#include "precompiled.h" +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// BufferStorage9.cpp Defines the BufferStorage9 class. + +#include "libGLESv2/renderer/BufferStorage9.h" +#include "common/debug.h" + +namespace rx +{ + +BufferStorage9::BufferStorage9() +{ + mMemory = NULL; + mAllocatedSize = 0; + mSize = 0; +} + +BufferStorage9::~BufferStorage9() +{ + delete[] mMemory; +} + +BufferStorage9 *BufferStorage9::makeBufferStorage9(BufferStorage *bufferStorage) +{ + ASSERT(HAS_DYNAMIC_TYPE(BufferStorage9*, bufferStorage)); + return static_cast<BufferStorage9*>(bufferStorage); +} + +void *BufferStorage9::getData() +{ + return mMemory; +} + +void BufferStorage9::setData(const void* data, unsigned int size, unsigned int offset) +{ + if (!mMemory || offset + size > mAllocatedSize) + { + unsigned int newAllocatedSize = offset + size; + void *newMemory = new char[newAllocatedSize]; + + if (offset > 0 && mMemory && mAllocatedSize > 0) + { + memcpy(newMemory, mMemory, std::min(offset, mAllocatedSize)); + } + + delete[] mMemory; + mMemory = newMemory; + mAllocatedSize = newAllocatedSize; + } + + mSize = std::max(mSize, offset + size); + memcpy(reinterpret_cast<char*>(mMemory) + offset, data, size); +} + +void BufferStorage9::clear() +{ + mSize = 0; +} + +unsigned int BufferStorage9::getSize() const +{ + return mSize; +} + +bool BufferStorage9::supportsDirectBinding() const +{ + return false; +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage9.h b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage9.h new file mode 100644 index 0000000000..3e803969bc --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/BufferStorage9.h @@ -0,0 +1,42 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// BufferStorage9.h Defines the BufferStorage9 class. + +#ifndef LIBGLESV2_RENDERER_BUFFERSTORAGE9_H_ +#define LIBGLESV2_RENDERER_BUFFERSTORAGE9_H_ + +#include "libGLESv2/renderer/BufferStorage.h" + +namespace rx +{ + +class BufferStorage9 : public BufferStorage +{ + public: + BufferStorage9(); + virtual ~BufferStorage9(); + + static BufferStorage9 *makeBufferStorage9(BufferStorage *bufferStorage); + + virtual void *getData(); + virtual void setData(const void* data, unsigned int size, unsigned int offset); + virtual void clear(); + virtual unsigned int getSize() const; + virtual bool supportsDirectBinding() const; + + private: + DISALLOW_COPY_AND_ASSIGN(BufferStorage9); + + void *mMemory; + unsigned int mAllocatedSize; + + unsigned int mSize; +}; + +} + +#endif // LIBGLESV2_RENDERER_BUFFERSTORAGE9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Fence11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Fence11.cpp new file mode 100644 index 0000000000..9d11c9a0fc --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Fence11.cpp @@ -0,0 +1,134 @@ +#include "precompiled.h" +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Fence11.cpp: Defines the rx::Fence11 class which implements rx::FenceImpl. + +#include "libGLESv2/renderer/Fence11.h" +#include "libGLESv2/main.h" +#include "libGLESv2/renderer/Renderer11.h" + +namespace rx +{ + +Fence11::Fence11(rx::Renderer11 *renderer) +{ + mRenderer = renderer; + mQuery = NULL; +} + +Fence11::~Fence11() +{ + if (mQuery) + { + mQuery->Release(); + mQuery = NULL; + } +} + +GLboolean Fence11::isFence() +{ + // GL_NV_fence spec: + // A name returned by GenFencesNV, but not yet set via SetFenceNV, is not the name of an existing fence. + return mQuery != NULL; +} + +void Fence11::setFence(GLenum condition) +{ + if (!mQuery) + { + D3D11_QUERY_DESC queryDesc; + queryDesc.Query = D3D11_QUERY_EVENT; + queryDesc.MiscFlags = 0; + + if (FAILED(mRenderer->getDevice()->CreateQuery(&queryDesc, &mQuery))) + { + return gl::error(GL_OUT_OF_MEMORY); + } + } + + mRenderer->getDeviceContext()->End(mQuery); + + setCondition(condition); + setStatus(GL_FALSE); +} + +GLboolean Fence11::testFence() +{ + if (mQuery == NULL) + { + return gl::error(GL_INVALID_OPERATION, GL_TRUE); + } + + HRESULT result = mRenderer->getDeviceContext()->GetData(mQuery, NULL, 0, 0); + + if (mRenderer->isDeviceLost()) + { + return gl::error(GL_OUT_OF_MEMORY, GL_TRUE); + } + + ASSERT(result == S_OK || result == S_FALSE); + setStatus(result == S_OK); + return getStatus(); +} + +void Fence11::finishFence() +{ + if (mQuery == NULL) + { + return gl::error(GL_INVALID_OPERATION); + } + + while (!testFence()) + { + Sleep(0); + } +} + +void Fence11::getFenceiv(GLenum pname, GLint *params) +{ + if (mQuery == NULL) + { + return gl::error(GL_INVALID_OPERATION); + } + + switch (pname) + { + case GL_FENCE_STATUS_NV: + { + // GL_NV_fence spec: + // Once the status of a fence has been finished (via FinishFenceNV) or tested and the returned status is TRUE (via either TestFenceNV + // or GetFenceivNV querying the FENCE_STATUS_NV), the status remains TRUE until the next SetFenceNV of the fence. + if (getStatus()) + { + params[0] = GL_TRUE; + return; + } + + HRESULT result = mRenderer->getDeviceContext()->GetData(mQuery, NULL, 0, D3D11_ASYNC_GETDATA_DONOTFLUSH); + + if (mRenderer->isDeviceLost()) + { + params[0] = GL_TRUE; + return gl::error(GL_OUT_OF_MEMORY); + } + + ASSERT(result == S_OK || result == S_FALSE); + setStatus(result == S_OK); + params[0] = getStatus(); + + break; + } + case GL_FENCE_CONDITION_NV: + params[0] = getCondition(); + break; + default: + return gl::error(GL_INVALID_ENUM); + break; + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Fence11.h b/src/3rdparty/angle/src/libGLESv2/renderer/Fence11.h new file mode 100644 index 0000000000..a5398bca14 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Fence11.h @@ -0,0 +1,39 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Fence11.h: Defines the rx::Fence11 class which implements rx::FenceImpl. + +#ifndef LIBGLESV2_RENDERER_Fence11_H_ +#define LIBGLESV2_RENDERER_Fence11_H_ + +#include "libGLESv2/renderer/FenceImpl.h" + +namespace rx +{ +class Renderer11; + +class Fence11 : public FenceImpl +{ + public: + explicit Fence11(rx::Renderer11 *renderer); + virtual ~Fence11(); + + GLboolean isFence(); + void setFence(GLenum condition); + GLboolean testFence(); + void finishFence(); + void getFenceiv(GLenum pname, GLint *params); + + private: + DISALLOW_COPY_AND_ASSIGN(Fence11); + + rx::Renderer11 *mRenderer; + ID3D11Query *mQuery; +}; + +} + +#endif // LIBGLESV2_RENDERER_FENCE11_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Fence9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Fence9.cpp new file mode 100644 index 0000000000..86064d7e52 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Fence9.cpp @@ -0,0 +1,135 @@ +#include "precompiled.h" +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Fence9.cpp: Defines the rx::Fence9 class. + +#include "libGLESv2/renderer/Fence9.h" +#include "libGLESv2/main.h" +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/renderer/Renderer9.h" + +namespace rx +{ + +Fence9::Fence9(rx::Renderer9 *renderer) +{ + mRenderer = renderer; + mQuery = NULL; +} + +Fence9::~Fence9() +{ + if (mQuery) + { + mRenderer->freeEventQuery(mQuery); + mQuery = NULL; + } +} + +GLboolean Fence9::isFence() +{ + // GL_NV_fence spec: + // A name returned by GenFencesNV, but not yet set via SetFenceNV, is not the name of an existing fence. + return mQuery != NULL; +} + +void Fence9::setFence(GLenum condition) +{ + if (!mQuery) + { + mQuery = mRenderer->allocateEventQuery(); + if (!mQuery) + { + return gl::error(GL_OUT_OF_MEMORY); + } + } + + HRESULT result = mQuery->Issue(D3DISSUE_END); + ASSERT(SUCCEEDED(result)); + + setCondition(condition); + setStatus(GL_FALSE); +} + +GLboolean Fence9::testFence() +{ + if (mQuery == NULL) + { + return gl::error(GL_INVALID_OPERATION, GL_TRUE); + } + + HRESULT result = mQuery->GetData(NULL, 0, D3DGETDATA_FLUSH); + + if (d3d9::isDeviceLostError(result)) + { + mRenderer->notifyDeviceLost(); + return gl::error(GL_OUT_OF_MEMORY, GL_TRUE); + } + + ASSERT(result == S_OK || result == S_FALSE); + setStatus(result == S_OK); + return getStatus(); +} + +void Fence9::finishFence() +{ + if (mQuery == NULL) + { + return gl::error(GL_INVALID_OPERATION); + } + + while (!testFence()) + { + Sleep(0); + } +} + +void Fence9::getFenceiv(GLenum pname, GLint *params) +{ + if (mQuery == NULL) + { + return gl::error(GL_INVALID_OPERATION); + } + + switch (pname) + { + case GL_FENCE_STATUS_NV: + { + // GL_NV_fence spec: + // Once the status of a fence has been finished (via FinishFenceNV) or tested and the returned status is TRUE (via either TestFenceNV + // or GetFenceivNV querying the FENCE_STATUS_NV), the status remains TRUE until the next SetFenceNV of the fence. + if (getStatus()) + { + params[0] = GL_TRUE; + return; + } + + HRESULT result = mQuery->GetData(NULL, 0, 0); + + if (d3d9::isDeviceLostError(result)) + { + params[0] = GL_TRUE; + mRenderer->notifyDeviceLost(); + return gl::error(GL_OUT_OF_MEMORY); + } + + ASSERT(result == S_OK || result == S_FALSE); + setStatus(result == S_OK); + params[0] = getStatus(); + + break; + } + case GL_FENCE_CONDITION_NV: + params[0] = getCondition(); + break; + default: + return gl::error(GL_INVALID_ENUM); + break; + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Fence9.h b/src/3rdparty/angle/src/libGLESv2/renderer/Fence9.h new file mode 100644 index 0000000000..9f17641e51 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Fence9.h @@ -0,0 +1,39 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Fence9.h: Defines the rx::Fence9 class which implements rx::FenceImpl. + +#ifndef LIBGLESV2_RENDERER_FENCE9_H_ +#define LIBGLESV2_RENDERER_FENCE9_H_ + +#include "libGLESv2/renderer/FenceImpl.h" + +namespace rx +{ +class Renderer9; + +class Fence9 : public FenceImpl +{ + public: + explicit Fence9(rx::Renderer9 *renderer); + virtual ~Fence9(); + + GLboolean isFence(); + void setFence(GLenum condition); + GLboolean testFence(); + void finishFence(); + void getFenceiv(GLenum pname, GLint *params); + + private: + DISALLOW_COPY_AND_ASSIGN(Fence9); + + rx::Renderer9 *mRenderer; + IDirect3DQuery9 *mQuery; +}; + +} + +#endif // LIBGLESV2_RENDERER_FENCE9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/FenceImpl.h b/src/3rdparty/angle/src/libGLESv2/renderer/FenceImpl.h new file mode 100644 index 0000000000..d7f2102a2e --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/FenceImpl.h @@ -0,0 +1,45 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// FenceImpl.h: Defines the rx::FenceImpl class. + +#ifndef LIBGLESV2_RENDERER_FENCEIMPL_H_ +#define LIBGLESV2_RENDERER_FENCEIMPL_H_ + +#include "common/angleutils.h" + +namespace rx +{ + +class FenceImpl +{ + public: + FenceImpl() : mStatus(GL_FALSE), mCondition(GL_NONE) { }; + virtual ~FenceImpl() { }; + + virtual GLboolean isFence() = 0; + virtual void setFence(GLenum condition) = 0; + virtual GLboolean testFence() = 0; + virtual void finishFence() = 0; + virtual void getFenceiv(GLenum pname, GLint *params) = 0; + + protected: + void setStatus(GLboolean status) { mStatus = status; } + GLboolean getStatus() const { return mStatus; } + + void setCondition(GLuint condition) { mCondition = condition; } + GLuint getCondition() const { return mCondition; } + + private: + DISALLOW_COPY_AND_ASSIGN(FenceImpl); + + GLboolean mStatus; + GLenum mCondition; +}; + +} + +#endif // LIBGLESV2_RENDERER_FENCEIMPL_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Image.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Image.cpp new file mode 100644 index 0000000000..57239ef74f --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Image.cpp @@ -0,0 +1,548 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Image.h: Implements the rx::Image class, an abstract base class for the +// renderer-specific classes which will define the interface to the underlying +// surfaces or resources. + +#include "libGLESv2/renderer/Image.h" + +namespace rx +{ + +Image::Image() +{ + mWidth = 0; + mHeight = 0; + mInternalFormat = GL_NONE; + mActualFormat = GL_NONE; +} + +void Image::loadAlphaDataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned char *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = 0; + dest[4 * x + 1] = 0; + dest[4 * x + 2] = 0; + dest[4 * x + 3] = source[x]; + } + } +} + +void Image::loadAlphaDataToNative(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned char *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = static_cast<unsigned char*>(output) + y * outputPitch; + memcpy(dest, source, width); + } +} + +void Image::loadAlphaFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const float *source = NULL; + float *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const float*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<float*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = 0; + dest[4 * x + 1] = 0; + dest[4 * x + 2] = 0; + dest[4 * x + 3] = source[x]; + } + } +} + +void Image::loadAlphaHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned short *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<unsigned short*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = 0; + dest[4 * x + 1] = 0; + dest[4 * x + 2] = 0; + dest[4 * x + 3] = source[x]; + } + } +} + +void Image::loadLuminanceDataToNativeOrBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output, bool native) +{ + const unsigned char *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = static_cast<unsigned char*>(output) + y * outputPitch; + + if (!native) // BGRA8 destination format + { + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[x]; + dest[4 * x + 1] = source[x]; + dest[4 * x + 2] = source[x]; + dest[4 * x + 3] = 0xFF; + } + } + else // L8 destination format + { + memcpy(dest, source, width); + } + } +} + +void Image::loadLuminanceFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const float *source = NULL; + float *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const float*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<float*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[x]; + dest[4 * x + 1] = source[x]; + dest[4 * x + 2] = source[x]; + dest[4 * x + 3] = 1.0f; + } + } +} + +void Image::loadLuminanceFloatDataToRGB(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const float *source = NULL; + float *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const float*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<float*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[3 * x + 0] = source[x]; + dest[3 * x + 1] = source[x]; + dest[3 * x + 2] = source[x]; + } + } +} + +void Image::loadLuminanceHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned short *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<unsigned short*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[x]; + dest[4 * x + 1] = source[x]; + dest[4 * x + 2] = source[x]; + dest[4 * x + 3] = 0x3C00; // SEEEEEMMMMMMMMMM, S = 0, E = 15, M = 0: 16bit flpt representation of 1 + } + } +} + +void Image::loadLuminanceAlphaDataToNativeOrBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output, bool native) +{ + const unsigned char *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = static_cast<unsigned char*>(output) + y * outputPitch; + + if (!native) // BGRA8 destination format + { + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[2*x+0]; + dest[4 * x + 1] = source[2*x+0]; + dest[4 * x + 2] = source[2*x+0]; + dest[4 * x + 3] = source[2*x+1]; + } + } + else + { + memcpy(dest, source, width * 2); + } + } +} + +void Image::loadLuminanceAlphaFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const float *source = NULL; + float *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const float*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<float*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[2*x+0]; + dest[4 * x + 1] = source[2*x+0]; + dest[4 * x + 2] = source[2*x+0]; + dest[4 * x + 3] = source[2*x+1]; + } + } +} + +void Image::loadLuminanceAlphaHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned short *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<unsigned short*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[2*x+0]; + dest[4 * x + 1] = source[2*x+0]; + dest[4 * x + 2] = source[2*x+0]; + dest[4 * x + 3] = source[2*x+1]; + } + } +} + +void Image::loadRGBUByteDataToBGRX(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned char *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[x * 3 + 2]; + dest[4 * x + 1] = source[x * 3 + 1]; + dest[4 * x + 2] = source[x * 3 + 0]; + dest[4 * x + 3] = 0xFF; + } + } +} + +void Image::loadRGBUByteDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned char *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[x * 3 + 0]; + dest[4 * x + 1] = source[x * 3 + 1]; + dest[4 * x + 2] = source[x * 3 + 2]; + dest[4 * x + 3] = 0xFF; + } + } +} + +void Image::loadRGB565DataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + unsigned short rgba = source[x]; + dest[4 * x + 0] = ((rgba & 0x001F) << 3) | ((rgba & 0x001F) >> 2); + dest[4 * x + 1] = ((rgba & 0x07E0) >> 3) | ((rgba & 0x07E0) >> 9); + dest[4 * x + 2] = ((rgba & 0xF800) >> 8) | ((rgba & 0xF800) >> 13); + dest[4 * x + 3] = 0xFF; + } + } +} + +void Image::loadRGB565DataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + unsigned short rgba = source[x]; + dest[4 * x + 0] = ((rgba & 0xF800) >> 8) | ((rgba & 0xF800) >> 13); + dest[4 * x + 1] = ((rgba & 0x07E0) >> 3) | ((rgba & 0x07E0) >> 9); + dest[4 * x + 2] = ((rgba & 0x001F) << 3) | ((rgba & 0x001F) >> 2); + dest[4 * x + 3] = 0xFF; + } + } +} + +void Image::loadRGBFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const float *source = NULL; + float *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const float*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<float*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[x * 3 + 0]; + dest[4 * x + 1] = source[x * 3 + 1]; + dest[4 * x + 2] = source[x * 3 + 2]; + dest[4 * x + 3] = 1.0f; + } + } +} + +void Image::loadRGBFloatDataToNative(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const float *source = NULL; + float *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const float*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<float*>(static_cast<unsigned char*>(output) + y * outputPitch); + memcpy(dest, source, width * 12); + } +} + +void Image::loadRGBHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned short *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<unsigned short*>(static_cast<unsigned char*>(output) + y * outputPitch); + for (int x = 0; x < width; x++) + { + dest[4 * x + 0] = source[x * 3 + 0]; + dest[4 * x + 1] = source[x * 3 + 1]; + dest[4 * x + 2] = source[x * 3 + 2]; + dest[4 * x + 3] = 0x3C00; // SEEEEEMMMMMMMMMM, S = 0, E = 15, M = 0: 16bit flpt representation of 1 + } + } +} + +void Image::loadRGBAUByteDataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned int *source = NULL; + unsigned int *dest = NULL; + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned int*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<unsigned int*>(static_cast<unsigned char*>(output) + y * outputPitch); + + for (int x = 0; x < width; x++) + { + unsigned int rgba = source[x]; + dest[x] = (_rotl(rgba, 16) & 0x00ff00ff) | (rgba & 0xff00ff00); + } + } +} + +void Image::loadRGBAUByteDataToNative(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned int *source = NULL; + unsigned int *dest = NULL; + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned int*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<unsigned int*>(static_cast<unsigned char*>(output) + y * outputPitch); + + memcpy(dest, source, width * 4); + } +} + +void Image::loadRGBA4444DataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + unsigned short rgba = source[x]; + dest[4 * x + 0] = ((rgba & 0x00F0) << 0) | ((rgba & 0x00F0) >> 4); + dest[4 * x + 1] = ((rgba & 0x0F00) >> 4) | ((rgba & 0x0F00) >> 8); + dest[4 * x + 2] = ((rgba & 0xF000) >> 8) | ((rgba & 0xF000) >> 12); + dest[4 * x + 3] = ((rgba & 0x000F) << 4) | ((rgba & 0x000F) >> 0); + } + } +} + +void Image::loadRGBA4444DataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + unsigned short rgba = source[x]; + dest[4 * x + 0] = ((rgba & 0xF000) >> 8) | ((rgba & 0xF000) >> 12); + dest[4 * x + 1] = ((rgba & 0x0F00) >> 4) | ((rgba & 0x0F00) >> 8); + dest[4 * x + 2] = ((rgba & 0x00F0) << 0) | ((rgba & 0x00F0) >> 4); + dest[4 * x + 3] = ((rgba & 0x000F) << 4) | ((rgba & 0x000F) >> 0); + } + } +} + +void Image::loadRGBA5551DataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + unsigned short rgba = source[x]; + dest[4 * x + 0] = ((rgba & 0x003E) << 2) | ((rgba & 0x003E) >> 3); + dest[4 * x + 1] = ((rgba & 0x07C0) >> 3) | ((rgba & 0x07C0) >> 8); + dest[4 * x + 2] = ((rgba & 0xF800) >> 8) | ((rgba & 0xF800) >> 13); + dest[4 * x + 3] = (rgba & 0x0001) ? 0xFF : 0; + } + } +} + +void Image::loadRGBA5551DataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned short *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned short*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = static_cast<unsigned char*>(output) + y * outputPitch; + for (int x = 0; x < width; x++) + { + unsigned short rgba = source[x]; + dest[4 * x + 0] = ((rgba & 0xF800) >> 8) | ((rgba & 0xF800) >> 13); + dest[4 * x + 1] = ((rgba & 0x07C0) >> 3) | ((rgba & 0x07C0) >> 8); + dest[4 * x + 2] = ((rgba & 0x003E) << 2) | ((rgba & 0x003E) >> 3); + dest[4 * x + 3] = (rgba & 0x0001) ? 0xFF : 0; + } + } +} + +void Image::loadRGBAFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const float *source = NULL; + float *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const float*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<float*>(static_cast<unsigned char*>(output) + y * outputPitch); + memcpy(dest, source, width * 16); + } +} + +void Image::loadRGBAHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned char *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = static_cast<unsigned char*>(output) + y * outputPitch; + memcpy(dest, source, width * 8); + } +} + +void Image::loadBGRADataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned char *source = NULL; + unsigned char *dest = NULL; + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = static_cast<unsigned char*>(output) + y * outputPitch; + memcpy(dest, source, width*4); + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Image.h b/src/3rdparty/angle/src/libGLESv2/renderer/Image.h new file mode 100644 index 0000000000..454e83e21e --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Image.h @@ -0,0 +1,131 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Image.h: Defines the rx::Image class, an abstract base class for the +// renderer-specific classes which will define the interface to the underlying +// surfaces or resources. + +#ifndef LIBGLESV2_RENDERER_IMAGE_H_ +#define LIBGLESV2_RENDERER_IMAGE_H_ + +#include "common/debug.h" + +namespace gl +{ +class Framebuffer; +} + +namespace rx +{ +class Renderer; +class TextureStorageInterface2D; +class TextureStorageInterfaceCube; + +class Image +{ + public: + Image(); + virtual ~Image() {}; + + GLsizei getWidth() const { return mWidth; } + GLsizei getHeight() const { return mHeight; } + GLenum getInternalFormat() const { return mInternalFormat; } + GLenum getActualFormat() const { return mActualFormat; } + + void markDirty() {mDirty = true;} + void markClean() {mDirty = false;} + virtual bool isDirty() const = 0; + + virtual void setManagedSurface(TextureStorageInterface2D *storage, int level) {}; + virtual void setManagedSurface(TextureStorageInterfaceCube *storage, int face, int level) {}; + virtual bool updateSurface(TextureStorageInterface2D *storage, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height) = 0; + virtual bool updateSurface(TextureStorageInterfaceCube *storage, int face, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height) = 0; + + virtual bool redefine(Renderer *renderer, GLint internalformat, GLsizei width, GLsizei height, bool forceRelease) = 0; + + virtual bool isRenderableFormat() const = 0; + + virtual void loadData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + GLint unpackAlignment, const void *input) = 0; + virtual void loadCompressedData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + const void *input) = 0; + + virtual void copy(GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height, gl::Framebuffer *source) = 0; + + static void loadAlphaDataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadAlphaDataToNative(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadAlphaDataToBGRASSE2(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadAlphaFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadAlphaHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadLuminanceDataToNativeOrBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output, bool native); + static void loadLuminanceFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadLuminanceFloatDataToRGB(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadLuminanceHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadLuminanceAlphaDataToNativeOrBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output, bool native); + static void loadLuminanceAlphaFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadLuminanceAlphaHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBUByteDataToBGRX(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBUByteDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGB565DataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGB565DataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBFloatDataToNative(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBAUByteDataToBGRASSE2(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBAUByteDataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBAUByteDataToNative(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBA4444DataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBA4444DataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBA5551DataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBA5551DataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBAFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadRGBAHalfFloatDataToRGBA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + static void loadBGRADataToBGRA(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output); + + protected: + GLsizei mWidth; + GLsizei mHeight; + GLint mInternalFormat; + GLenum mActualFormat; + + bool mDirty; + + private: + DISALLOW_COPY_AND_ASSIGN(Image); +}; + +} + +#endif // LIBGLESV2_RENDERER_IMAGE_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Image11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Image11.cpp new file mode 100644 index 0000000000..8c78c7d750 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Image11.cpp @@ -0,0 +1,457 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Image11.h: Implements the rx::Image11 class, which acts as the interface to +// the actual underlying resources of a Texture + +#include "libGLESv2/renderer/Renderer11.h" +#include "libGLESv2/renderer/Image11.h" +#include "libGLESv2/renderer/TextureStorage11.h" +#include "libGLESv2/Framebuffer.h" +#include "libGLESv2/Renderbuffer.h" + +#include "libGLESv2/main.h" +#include "libGLESv2/utilities.h" +#include "libGLESv2/renderer/renderer11_utils.h" +#include "libGLESv2/renderer/generatemip.h" + +namespace rx +{ + +Image11::Image11() +{ + mStagingTexture = NULL; + mRenderer = NULL; + mDXGIFormat = DXGI_FORMAT_UNKNOWN; +} + +Image11::~Image11() +{ + if (mStagingTexture) + { + mStagingTexture->Release(); + } +} + +Image11 *Image11::makeImage11(Image *img) +{ + ASSERT(HAS_DYNAMIC_TYPE(rx::Image11*, img)); + return static_cast<rx::Image11*>(img); +} + +void Image11::generateMipmap(Image11 *dest, Image11 *src) +{ + ASSERT(src->getDXGIFormat() == dest->getDXGIFormat()); + ASSERT(src->getWidth() == 1 || src->getWidth() / 2 == dest->getWidth()); + ASSERT(src->getHeight() == 1 || src->getHeight() / 2 == dest->getHeight()); + + D3D11_MAPPED_SUBRESOURCE destMapped, srcMapped; + dest->map(&destMapped); + src->map(&srcMapped); + + const unsigned char *sourceData = reinterpret_cast<const unsigned char*>(srcMapped.pData); + unsigned char *destData = reinterpret_cast<unsigned char*>(destMapped.pData); + + if (sourceData && destData) + { + switch (src->getDXGIFormat()) + { + case DXGI_FORMAT_R8G8B8A8_UNORM: + case DXGI_FORMAT_B8G8R8A8_UNORM: + GenerateMip<R8G8B8A8>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_A8_UNORM: + GenerateMip<A8>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R8_UNORM: + GenerateMip<R8>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R32G32B32A32_FLOAT: + GenerateMip<A32B32G32R32F>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R32G32B32_FLOAT: + GenerateMip<R32G32B32F>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R16G16B16A16_FLOAT: + GenerateMip<A16B16G16R16F>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R8G8_UNORM: + GenerateMip<R8G8>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R16_FLOAT: + GenerateMip<R16F>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R16G16_FLOAT: + GenerateMip<R16G16F>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R32_FLOAT: + GenerateMip<R32F>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + case DXGI_FORMAT_R32G32_FLOAT: + GenerateMip<R32G32F>(src->getWidth(), src->getHeight(), sourceData, srcMapped.RowPitch, destData, destMapped.RowPitch); + break; + default: + UNREACHABLE(); + break; + } + + dest->unmap(); + src->unmap(); + } + + dest->markDirty(); +} + +bool Image11::isDirty() const +{ + return (mStagingTexture && mDirty); +} + +bool Image11::updateSurface(TextureStorageInterface2D *storage, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height) +{ + TextureStorage11_2D *storage11 = TextureStorage11_2D::makeTextureStorage11_2D(storage->getStorageInstance()); + return storage11->updateSubresourceLevel(getStagingTexture(), getStagingSubresource(), level, 0, xoffset, yoffset, width, height); +} + +bool Image11::updateSurface(TextureStorageInterfaceCube *storage, int face, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height) +{ + TextureStorage11_Cube *storage11 = TextureStorage11_Cube::makeTextureStorage11_Cube(storage->getStorageInstance()); + return storage11->updateSubresourceLevel(getStagingTexture(), getStagingSubresource(), level, face, xoffset, yoffset, width, height); +} + +bool Image11::redefine(Renderer *renderer, GLint internalformat, GLsizei width, GLsizei height, bool forceRelease) +{ + if (mWidth != width || + mHeight != height || + mInternalFormat != internalformat || + forceRelease) + { + mRenderer = Renderer11::makeRenderer11(renderer); + + mWidth = width; + mHeight = height; + mInternalFormat = internalformat; + // compute the d3d format that will be used + mDXGIFormat = gl_d3d11::ConvertTextureFormat(internalformat); + mActualFormat = d3d11_gl::ConvertTextureInternalFormat(mDXGIFormat); + + if (mStagingTexture) + { + mStagingTexture->Release(); + mStagingTexture = NULL; + } + + return true; + } + + return false; +} + +bool Image11::isRenderableFormat() const +{ + return TextureStorage11::IsTextureFormatRenderable(mDXGIFormat); +} + +DXGI_FORMAT Image11::getDXGIFormat() const +{ + // this should only happen if the image hasn't been redefined first + // which would be a bug by the caller + ASSERT(mDXGIFormat != DXGI_FORMAT_UNKNOWN); + + return mDXGIFormat; +} + +// Store the pixel rectangle designated by xoffset,yoffset,width,height with pixels stored as format/type at input +// into the target pixel rectangle. +void Image11::loadData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + GLint unpackAlignment, const void *input) +{ + D3D11_MAPPED_SUBRESOURCE mappedImage; + HRESULT result = map(&mappedImage); + if (FAILED(result)) + { + ERR("Could not map image for loading."); + return; + } + + GLsizei inputPitch = gl::ComputePitch(width, mInternalFormat, unpackAlignment); + size_t pixelSize = d3d11::ComputePixelSizeBits(mDXGIFormat) / 8; + void* offsetMappedData = (void*)((BYTE *)mappedImage.pData + (yoffset * mappedImage.RowPitch + xoffset * pixelSize)); + + switch (mInternalFormat) + { + case GL_ALPHA8_EXT: + loadAlphaDataToNative(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_LUMINANCE8_EXT: + loadLuminanceDataToNativeOrBGRA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData, false); + break; + case GL_ALPHA32F_EXT: + loadAlphaFloatDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_LUMINANCE32F_EXT: + loadLuminanceFloatDataToRGB(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_ALPHA16F_EXT: + loadAlphaHalfFloatDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_LUMINANCE16F_EXT: + loadLuminanceHalfFloatDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_LUMINANCE8_ALPHA8_EXT: + loadLuminanceAlphaDataToNativeOrBGRA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData, false); + break; + case GL_LUMINANCE_ALPHA32F_EXT: + loadLuminanceAlphaFloatDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_LUMINANCE_ALPHA16F_EXT: + loadLuminanceAlphaHalfFloatDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGB8_OES: + loadRGBUByteDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGB565: + loadRGB565DataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGBA8_OES: + loadRGBAUByteDataToNative(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGBA4: + loadRGBA4444DataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGB5_A1: + loadRGBA5551DataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_BGRA8_EXT: + loadBGRADataToBGRA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGB32F_EXT: + loadRGBFloatDataToNative(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGB16F_EXT: + loadRGBHalfFloatDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGBA32F_EXT: + loadRGBAFloatDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + case GL_RGBA16F_EXT: + loadRGBAHalfFloatDataToRGBA(width, height, inputPitch, input, mappedImage.RowPitch, offsetMappedData); + break; + default: UNREACHABLE(); + } + + unmap(); +} + +void Image11::loadCompressedData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + const void *input) +{ + ASSERT(xoffset % 4 == 0); + ASSERT(yoffset % 4 == 0); + + D3D11_MAPPED_SUBRESOURCE mappedImage; + HRESULT result = map(&mappedImage); + if (FAILED(result)) + { + ERR("Could not map image for loading."); + return; + } + + // Size computation assumes a 4x4 block compressed texture format + size_t blockSize = d3d11::ComputeBlockSizeBits(mDXGIFormat) / 8; + void* offsetMappedData = (void*)((BYTE *)mappedImage.pData + ((yoffset / 4) * mappedImage.RowPitch + (xoffset / 4) * blockSize)); + + GLsizei inputSize = gl::ComputeCompressedSize(width, height, mInternalFormat); + GLsizei inputPitch = gl::ComputeCompressedPitch(width, mInternalFormat); + int rows = inputSize / inputPitch; + for (int i = 0; i < rows; ++i) + { + memcpy((void*)((BYTE*)offsetMappedData + i * mappedImage.RowPitch), (void*)((BYTE*)input + i * inputPitch), inputPitch); + } + + unmap(); +} + +void Image11::copy(GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height, gl::Framebuffer *source) +{ + gl::Renderbuffer *colorbuffer = source->getReadColorbuffer(); + + if (colorbuffer && colorbuffer->getActualFormat() == (GLuint)mActualFormat) + { + // No conversion needed-- use copyback fastpath + ID3D11Texture2D *colorBufferTexture = NULL; + unsigned int subresourceIndex = 0; + + if (mRenderer->getRenderTargetResource(colorbuffer, &subresourceIndex, &colorBufferTexture)) + { + D3D11_TEXTURE2D_DESC textureDesc; + colorBufferTexture->GetDesc(&textureDesc); + + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + ID3D11Texture2D* srcTex = NULL; + if (textureDesc.SampleDesc.Count > 1) + { + D3D11_TEXTURE2D_DESC resolveDesc; + resolveDesc.Width = textureDesc.Width; + resolveDesc.Height = textureDesc.Height; + resolveDesc.MipLevels = 1; + resolveDesc.ArraySize = 1; + resolveDesc.Format = textureDesc.Format; + resolveDesc.SampleDesc.Count = 1; + resolveDesc.SampleDesc.Quality = 0; + resolveDesc.Usage = D3D11_USAGE_DEFAULT; + resolveDesc.BindFlags = 0; + resolveDesc.CPUAccessFlags = 0; + resolveDesc.MiscFlags = 0; + + HRESULT result = device->CreateTexture2D(&resolveDesc, NULL, &srcTex); + if (FAILED(result)) + { + ERR("Failed to create resolve texture for Image11::copy, HRESULT: 0x%X.", result); + return; + } + + deviceContext->ResolveSubresource(srcTex, 0, colorBufferTexture, subresourceIndex, textureDesc.Format); + subresourceIndex = 0; + } + else + { + srcTex = colorBufferTexture; + srcTex->AddRef(); + } + + D3D11_BOX srcBox; + srcBox.left = x; + srcBox.right = x + width; + srcBox.top = y; + srcBox.bottom = y + height; + srcBox.front = 0; + srcBox.back = 1; + + deviceContext->CopySubresourceRegion(mStagingTexture, 0, xoffset, yoffset, 0, srcTex, subresourceIndex, &srcBox); + + srcTex->Release(); + colorBufferTexture->Release(); + } + } + else + { + // This format requires conversion, so we must copy the texture to staging and manually convert via readPixels + D3D11_MAPPED_SUBRESOURCE mappedImage; + HRESULT result = map(&mappedImage); + + // determine the offset coordinate into the destination buffer + GLsizei rowOffset = gl::ComputePixelSize(mActualFormat) * xoffset; + void *dataOffset = static_cast<unsigned char*>(mappedImage.pData) + mappedImage.RowPitch * yoffset + rowOffset; + + mRenderer->readPixels(source, x, y, width, height, gl::ExtractFormat(mInternalFormat), + gl::ExtractType(mInternalFormat), mappedImage.RowPitch, false, 4, dataOffset); + + unmap(); + } +} + +ID3D11Texture2D *Image11::getStagingTexture() +{ + createStagingTexture(); + + return mStagingTexture; +} + +unsigned int Image11::getStagingSubresource() +{ + createStagingTexture(); + + return mStagingSubresource; +} + +void Image11::createStagingTexture() +{ + if (mStagingTexture) + { + return; + } + + ID3D11Texture2D *newTexture = NULL; + int lodOffset = 1; + const DXGI_FORMAT dxgiFormat = getDXGIFormat(); + ASSERT(!d3d11::IsDepthStencilFormat(dxgiFormat)); // We should never get here for depth textures + + if (mWidth != 0 && mHeight != 0) + { + GLsizei width = mWidth; + GLsizei height = mHeight; + + // adjust size if needed for compressed textures + gl::MakeValidSize(false, d3d11::IsCompressed(dxgiFormat), &width, &height, &lodOffset); + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = width; + desc.Height = height; + desc.MipLevels = lodOffset + 1; + desc.ArraySize = 1; + desc.Format = dxgiFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_STAGING; + desc.BindFlags = 0; + desc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + desc.MiscFlags = 0; + + HRESULT result = device->CreateTexture2D(&desc, NULL, &newTexture); + + if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + ERR("Creating image failed."); + return gl::error(GL_OUT_OF_MEMORY); + } + } + + mStagingTexture = newTexture; + mStagingSubresource = D3D11CalcSubresource(lodOffset, 0, lodOffset + 1); + mDirty = false; +} + +HRESULT Image11::map(D3D11_MAPPED_SUBRESOURCE *map) +{ + createStagingTexture(); + + HRESULT result = E_FAIL; + + if (mStagingTexture) + { + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + result = deviceContext->Map(mStagingTexture, mStagingSubresource, D3D11_MAP_WRITE, 0, map); + + // this can fail if the device is removed (from TDR) + if (d3d11::isDeviceLostError(result)) + { + mRenderer->notifyDeviceLost(); + } + else if (SUCCEEDED(result)) + { + mDirty = true; + } + } + + return result; +} + +void Image11::unmap() +{ + if (mStagingTexture) + { + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + deviceContext->Unmap(mStagingTexture, mStagingSubresource); + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Image11.h b/src/3rdparty/angle/src/libGLESv2/renderer/Image11.h new file mode 100644 index 0000000000..4d5f1c1780 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Image11.h @@ -0,0 +1,76 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Image11.h: Defines the rx::Image11 class, which acts as the interface to +// the actual underlying resources of a Texture + +#ifndef LIBGLESV2_RENDERER_IMAGE11_H_ +#define LIBGLESV2_RENDERER_IMAGE11_H_ + +#include "libGLESv2/renderer/Image.h" + +#include "common/debug.h" + +namespace gl +{ +class Framebuffer; +} + +namespace rx +{ +class Renderer; +class Renderer11; +class TextureStorageInterface2D; +class TextureStorageInterfaceCube; + +class Image11 : public Image +{ + public: + Image11(); + virtual ~Image11(); + + static Image11 *makeImage11(Image *img); + + static void generateMipmap(Image11 *dest, Image11 *src); + + virtual bool isDirty() const; + + virtual bool updateSurface(TextureStorageInterface2D *storage, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height); + virtual bool updateSurface(TextureStorageInterfaceCube *storage, int face, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height); + + virtual bool redefine(Renderer *renderer, GLint internalformat, GLsizei width, GLsizei height, bool forceRelease); + + virtual bool isRenderableFormat() const; + DXGI_FORMAT getDXGIFormat() const; + + virtual void loadData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + GLint unpackAlignment, const void *input); + virtual void loadCompressedData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + const void *input); + + virtual void copy(GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height, gl::Framebuffer *source); + + protected: + HRESULT map(D3D11_MAPPED_SUBRESOURCE *map); + void unmap(); + + private: + DISALLOW_COPY_AND_ASSIGN(Image11); + + ID3D11Texture2D *getStagingTexture(); + unsigned int getStagingSubresource(); + void createStagingTexture(); + + Renderer11 *mRenderer; + + DXGI_FORMAT mDXGIFormat; + ID3D11Texture2D *mStagingTexture; + unsigned int mStagingSubresource; +}; + +} + +#endif // LIBGLESV2_RENDERER_IMAGE11_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Image9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Image9.cpp new file mode 100644 index 0000000000..53030b7f1e --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Image9.cpp @@ -0,0 +1,736 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Image9.cpp: Implements the rx::Image9 class, which acts as the interface to +// the actual underlying surfaces of a Texture. + +#include "libGLESv2/renderer/Image9.h" + +#include "libGLESv2/main.h" +#include "libGLESv2/Framebuffer.h" +#include "libGLESv2/Renderbuffer.h" +#include "libGLESv2/renderer/Renderer9.h" +#include "libGLESv2/renderer/RenderTarget9.h" +#include "libGLESv2/renderer/TextureStorage9.h" + +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/renderer/generatemip.h" + +namespace rx +{ + +Image9::Image9() +{ + mSurface = NULL; + mRenderer = NULL; + + mD3DPool = D3DPOOL_SYSTEMMEM; + mD3DFormat = D3DFMT_UNKNOWN; +} + +Image9::~Image9() +{ + if (mSurface) + { + mSurface->Release(); + } +} + +void Image9::generateMip(IDirect3DSurface9 *destSurface, IDirect3DSurface9 *sourceSurface) +{ + D3DSURFACE_DESC destDesc; + HRESULT result = destSurface->GetDesc(&destDesc); + ASSERT(SUCCEEDED(result)); + + D3DSURFACE_DESC sourceDesc; + result = sourceSurface->GetDesc(&sourceDesc); + ASSERT(SUCCEEDED(result)); + + ASSERT(sourceDesc.Format == destDesc.Format); + ASSERT(sourceDesc.Width == 1 || sourceDesc.Width / 2 == destDesc.Width); + ASSERT(sourceDesc.Height == 1 || sourceDesc.Height / 2 == destDesc.Height); + + D3DLOCKED_RECT sourceLocked = {0}; + result = sourceSurface->LockRect(&sourceLocked, NULL, D3DLOCK_READONLY); + ASSERT(SUCCEEDED(result)); + + D3DLOCKED_RECT destLocked = {0}; + result = destSurface->LockRect(&destLocked, NULL, 0); + ASSERT(SUCCEEDED(result)); + + const unsigned char *sourceData = reinterpret_cast<const unsigned char*>(sourceLocked.pBits); + unsigned char *destData = reinterpret_cast<unsigned char*>(destLocked.pBits); + + if (sourceData && destData) + { + switch (sourceDesc.Format) + { + case D3DFMT_L8: + GenerateMip<L8>(sourceDesc.Width, sourceDesc.Height, sourceData, sourceLocked.Pitch, destData, destLocked.Pitch); + break; + case D3DFMT_A8L8: + GenerateMip<A8L8>(sourceDesc.Width, sourceDesc.Height, sourceData, sourceLocked.Pitch, destData, destLocked.Pitch); + break; + case D3DFMT_A8R8G8B8: + case D3DFMT_X8R8G8B8: + GenerateMip<A8R8G8B8>(sourceDesc.Width, sourceDesc.Height, sourceData, sourceLocked.Pitch, destData, destLocked.Pitch); + break; + case D3DFMT_A16B16G16R16F: + GenerateMip<A16B16G16R16F>(sourceDesc.Width, sourceDesc.Height, sourceData, sourceLocked.Pitch, destData, destLocked.Pitch); + break; + case D3DFMT_A32B32G32R32F: + GenerateMip<A32B32G32R32F>(sourceDesc.Width, sourceDesc.Height, sourceData, sourceLocked.Pitch, destData, destLocked.Pitch); + break; + default: + UNREACHABLE(); + break; + } + + destSurface->UnlockRect(); + sourceSurface->UnlockRect(); + } +} + +Image9 *Image9::makeImage9(Image *img) +{ + ASSERT(HAS_DYNAMIC_TYPE(rx::Image9*, img)); + return static_cast<rx::Image9*>(img); +} + +void Image9::generateMipmap(Image9 *dest, Image9 *source) +{ + IDirect3DSurface9 *sourceSurface = source->getSurface(); + if (sourceSurface == NULL) + return gl::error(GL_OUT_OF_MEMORY); + + IDirect3DSurface9 *destSurface = dest->getSurface(); + generateMip(destSurface, sourceSurface); + + dest->markDirty(); +} + +void Image9::copyLockableSurfaces(IDirect3DSurface9 *dest, IDirect3DSurface9 *source) +{ + D3DLOCKED_RECT sourceLock = {0}; + D3DLOCKED_RECT destLock = {0}; + + source->LockRect(&sourceLock, NULL, 0); + dest->LockRect(&destLock, NULL, 0); + + if (sourceLock.pBits && destLock.pBits) + { + D3DSURFACE_DESC desc; + source->GetDesc(&desc); + + int rows = d3d9::IsCompressedFormat(desc.Format) ? desc.Height / 4 : desc.Height; + int bytes = d3d9::ComputeRowSize(desc.Format, desc.Width); + ASSERT(bytes <= sourceLock.Pitch && bytes <= destLock.Pitch); + + for(int i = 0; i < rows; i++) + { + memcpy((char*)destLock.pBits + destLock.Pitch * i, (char*)sourceLock.pBits + sourceLock.Pitch * i, bytes); + } + + source->UnlockRect(); + dest->UnlockRect(); + } + else UNREACHABLE(); +} + +bool Image9::redefine(rx::Renderer *renderer, GLint internalformat, GLsizei width, GLsizei height, bool forceRelease) +{ + if (mWidth != width || + mHeight != height || + mInternalFormat != internalformat || + forceRelease) + { + mRenderer = Renderer9::makeRenderer9(renderer); + + mWidth = width; + mHeight = height; + mInternalFormat = internalformat; + // compute the d3d format that will be used + mD3DFormat = mRenderer->ConvertTextureInternalFormat(internalformat); + mActualFormat = d3d9_gl::GetEquivalentFormat(mD3DFormat); + + if (mSurface) + { + mSurface->Release(); + mSurface = NULL; + } + + return true; + } + + return false; +} + +void Image9::createSurface() +{ + if(mSurface) + { + return; + } + + IDirect3DTexture9 *newTexture = NULL; + IDirect3DSurface9 *newSurface = NULL; + const D3DPOOL poolToUse = D3DPOOL_SYSTEMMEM; + const D3DFORMAT d3dFormat = getD3DFormat(); + ASSERT(d3dFormat != D3DFMT_INTZ); // We should never get here for depth textures + + if (mWidth != 0 && mHeight != 0) + { + int levelToFetch = 0; + GLsizei requestWidth = mWidth; + GLsizei requestHeight = mHeight; + gl::MakeValidSize(true, gl::IsCompressed(mInternalFormat), &requestWidth, &requestHeight, &levelToFetch); + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + HRESULT result = device->CreateTexture(requestWidth, requestHeight, levelToFetch + 1, 0, d3dFormat, + poolToUse, &newTexture, NULL); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + ERR("Creating image surface failed."); + return gl::error(GL_OUT_OF_MEMORY); + } + + newTexture->GetSurfaceLevel(levelToFetch, &newSurface); + newTexture->Release(); + } + + mSurface = newSurface; + mDirty = false; + mD3DPool = poolToUse; +} + +HRESULT Image9::lock(D3DLOCKED_RECT *lockedRect, const RECT *rect) +{ + createSurface(); + + HRESULT result = D3DERR_INVALIDCALL; + + if (mSurface) + { + result = mSurface->LockRect(lockedRect, rect, 0); + ASSERT(SUCCEEDED(result)); + + mDirty = true; + } + + return result; +} + +void Image9::unlock() +{ + if (mSurface) + { + HRESULT result = mSurface->UnlockRect(); + ASSERT(SUCCEEDED(result)); + } +} + +bool Image9::isRenderableFormat() const +{ + return TextureStorage9::IsTextureFormatRenderable(getD3DFormat()); +} + +D3DFORMAT Image9::getD3DFormat() const +{ + // this should only happen if the image hasn't been redefined first + // which would be a bug by the caller + ASSERT(mD3DFormat != D3DFMT_UNKNOWN); + + return mD3DFormat; +} + +IDirect3DSurface9 *Image9::getSurface() +{ + createSurface(); + + return mSurface; +} + +void Image9::setManagedSurface(TextureStorageInterface2D *storage, int level) +{ + TextureStorage9_2D *storage9 = TextureStorage9_2D::makeTextureStorage9_2D(storage->getStorageInstance()); + setManagedSurface(storage9->getSurfaceLevel(level, false)); +} + +void Image9::setManagedSurface(TextureStorageInterfaceCube *storage, int face, int level) +{ + TextureStorage9_Cube *storage9 = TextureStorage9_Cube::makeTextureStorage9_Cube(storage->getStorageInstance()); + setManagedSurface(storage9->getCubeMapSurface(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face, level, false)); +} + +void Image9::setManagedSurface(IDirect3DSurface9 *surface) +{ + D3DSURFACE_DESC desc; + surface->GetDesc(&desc); + ASSERT(desc.Pool == D3DPOOL_MANAGED); + + if ((GLsizei)desc.Width == mWidth && (GLsizei)desc.Height == mHeight) + { + if (mSurface) + { + copyLockableSurfaces(surface, mSurface); + mSurface->Release(); + } + + mSurface = surface; + mD3DPool = desc.Pool; + } +} + +bool Image9::updateSurface(TextureStorageInterface2D *storage, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height) +{ + ASSERT(getSurface() != NULL); + TextureStorage9_2D *storage9 = TextureStorage9_2D::makeTextureStorage9_2D(storage->getStorageInstance()); + return updateSurface(storage9->getSurfaceLevel(level, true), xoffset, yoffset, width, height); +} + +bool Image9::updateSurface(TextureStorageInterfaceCube *storage, int face, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height) +{ + ASSERT(getSurface() != NULL); + TextureStorage9_Cube *storage9 = TextureStorage9_Cube::makeTextureStorage9_Cube(storage->getStorageInstance()); + return updateSurface(storage9->getCubeMapSurface(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face, level, true), xoffset, yoffset, width, height); +} + +bool Image9::updateSurface(IDirect3DSurface9 *destSurface, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height) +{ + if (!destSurface) + return false; + + IDirect3DSurface9 *sourceSurface = getSurface(); + + if (sourceSurface && sourceSurface != destSurface) + { + RECT rect; + rect.left = xoffset; + rect.top = yoffset; + rect.right = xoffset + width; + rect.bottom = yoffset + height; + + POINT point = {rect.left, rect.top}; + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + if (mD3DPool == D3DPOOL_MANAGED) + { + D3DSURFACE_DESC desc; + sourceSurface->GetDesc(&desc); + + IDirect3DSurface9 *surf = 0; + HRESULT result = device->CreateOffscreenPlainSurface(desc.Width, desc.Height, desc.Format, D3DPOOL_SYSTEMMEM, &surf, NULL); + + if (SUCCEEDED(result)) + { + copyLockableSurfaces(surf, sourceSurface); + result = device->UpdateSurface(surf, &rect, destSurface, &point); + ASSERT(SUCCEEDED(result)); + surf->Release(); + } + } + else + { + // UpdateSurface: source must be SYSTEMMEM, dest must be DEFAULT pools + HRESULT result = device->UpdateSurface(sourceSurface, &rect, destSurface, &point); + ASSERT(SUCCEEDED(result)); + } + } + + destSurface->Release(); + return true; +} + +// Store the pixel rectangle designated by xoffset,yoffset,width,height with pixels stored as format/type at input +// into the target pixel rectangle. +void Image9::loadData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + GLint unpackAlignment, const void *input) +{ + RECT lockRect = + { + xoffset, yoffset, + xoffset + width, yoffset + height + }; + + D3DLOCKED_RECT locked; + HRESULT result = lock(&locked, &lockRect); + if (FAILED(result)) + { + return; + } + + + GLsizei inputPitch = gl::ComputePitch(width, mInternalFormat, unpackAlignment); + + switch (mInternalFormat) + { + case GL_ALPHA8_EXT: +#if defined(__SSE2__) + if (gl::supportsSSE2()) + { + loadAlphaDataToBGRASSE2(width, height, inputPitch, input, locked.Pitch, locked.pBits); + } + else +#endif + { + loadAlphaDataToBGRA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + } + break; + case GL_LUMINANCE8_EXT: + loadLuminanceDataToNativeOrBGRA(width, height, inputPitch, input, locked.Pitch, locked.pBits, getD3DFormat() == D3DFMT_L8); + break; + case GL_ALPHA32F_EXT: + loadAlphaFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_LUMINANCE32F_EXT: + loadLuminanceFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_ALPHA16F_EXT: + loadAlphaHalfFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_LUMINANCE16F_EXT: + loadLuminanceHalfFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_LUMINANCE8_ALPHA8_EXT: + loadLuminanceAlphaDataToNativeOrBGRA(width, height, inputPitch, input, locked.Pitch, locked.pBits, getD3DFormat() == D3DFMT_A8L8); + break; + case GL_LUMINANCE_ALPHA32F_EXT: + loadLuminanceAlphaFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_LUMINANCE_ALPHA16F_EXT: + loadLuminanceAlphaHalfFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_RGB8_OES: + loadRGBUByteDataToBGRX(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_RGB565: + loadRGB565DataToBGRA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_RGBA8_OES: +#if defined(__SSE2__) + if (gl::supportsSSE2()) + { + loadRGBAUByteDataToBGRASSE2(width, height, inputPitch, input, locked.Pitch, locked.pBits); + } + else +#endif + { + loadRGBAUByteDataToBGRA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + } + break; + case GL_RGBA4: + loadRGBA4444DataToBGRA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_RGB5_A1: + loadRGBA5551DataToBGRA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_BGRA8_EXT: + loadBGRADataToBGRA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + // float textures are converted to RGBA, not BGRA, as they're stored that way in D3D + case GL_RGB32F_EXT: + loadRGBFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_RGB16F_EXT: + loadRGBHalfFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_RGBA32F_EXT: + loadRGBAFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + case GL_RGBA16F_EXT: + loadRGBAHalfFloatDataToRGBA(width, height, inputPitch, input, locked.Pitch, locked.pBits); + break; + default: UNREACHABLE(); + } + + unlock(); +} + +void Image9::loadCompressedData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + const void *input) +{ + ASSERT(xoffset % 4 == 0); + ASSERT(yoffset % 4 == 0); + + RECT lockRect = { + xoffset, yoffset, + xoffset + width, yoffset + height + }; + + D3DLOCKED_RECT locked; + HRESULT result = lock(&locked, &lockRect); + if (FAILED(result)) + { + return; + } + + GLsizei inputSize = gl::ComputeCompressedSize(width, height, mInternalFormat); + GLsizei inputPitch = gl::ComputeCompressedPitch(width, mInternalFormat); + int rows = inputSize / inputPitch; + for (int i = 0; i < rows; ++i) + { + memcpy((void*)((BYTE*)locked.pBits + i * locked.Pitch), (void*)((BYTE*)input + i * inputPitch), inputPitch); + } + + unlock(); +} + +// This implements glCopyTex[Sub]Image2D for non-renderable internal texture formats and incomplete textures +void Image9::copy(GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height, gl::Framebuffer *source) +{ + RenderTarget9 *renderTarget = NULL; + IDirect3DSurface9 *surface = NULL; + gl::Renderbuffer *colorbuffer = source->getColorbuffer(0); + + if (colorbuffer) + { + renderTarget = RenderTarget9::makeRenderTarget9(colorbuffer->getRenderTarget()); + } + + if (renderTarget) + { + surface = renderTarget->getSurface(); + } + + if (!surface) + { + ERR("Failed to retrieve the render target."); + return gl::error(GL_OUT_OF_MEMORY); + } + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + IDirect3DSurface9 *renderTargetData = NULL; + D3DSURFACE_DESC description; + surface->GetDesc(&description); + + HRESULT result = device->CreateOffscreenPlainSurface(description.Width, description.Height, description.Format, D3DPOOL_SYSTEMMEM, &renderTargetData, NULL); + + if (FAILED(result)) + { + ERR("Could not create matching destination surface."); + surface->Release(); + return gl::error(GL_OUT_OF_MEMORY); + } + + result = device->GetRenderTargetData(surface, renderTargetData); + + if (FAILED(result)) + { + ERR("GetRenderTargetData unexpectedly failed."); + renderTargetData->Release(); + surface->Release(); + return gl::error(GL_OUT_OF_MEMORY); + } + + RECT sourceRect = {x, y, x + width, y + height}; + RECT destRect = {xoffset, yoffset, xoffset + width, yoffset + height}; + + D3DLOCKED_RECT sourceLock = {0}; + result = renderTargetData->LockRect(&sourceLock, &sourceRect, 0); + + if (FAILED(result)) + { + ERR("Failed to lock the source surface (rectangle might be invalid)."); + renderTargetData->Release(); + surface->Release(); + return gl::error(GL_OUT_OF_MEMORY); + } + + D3DLOCKED_RECT destLock = {0}; + result = lock(&destLock, &destRect); + + if (FAILED(result)) + { + ERR("Failed to lock the destination surface (rectangle might be invalid)."); + renderTargetData->UnlockRect(); + renderTargetData->Release(); + surface->Release(); + return gl::error(GL_OUT_OF_MEMORY); + } + + if (destLock.pBits && sourceLock.pBits) + { + unsigned char *source = (unsigned char*)sourceLock.pBits; + unsigned char *dest = (unsigned char*)destLock.pBits; + + switch (description.Format) + { + case D3DFMT_X8R8G8B8: + case D3DFMT_A8R8G8B8: + switch(getD3DFormat()) + { + case D3DFMT_X8R8G8B8: + case D3DFMT_A8R8G8B8: + for(int y = 0; y < height; y++) + { + memcpy(dest, source, 4 * width); + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + case D3DFMT_L8: + for(int y = 0; y < height; y++) + { + for(int x = 0; x < width; x++) + { + dest[x] = source[x * 4 + 2]; + } + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + case D3DFMT_A8L8: + for(int y = 0; y < height; y++) + { + for(int x = 0; x < width; x++) + { + dest[x * 2 + 0] = source[x * 4 + 2]; + dest[x * 2 + 1] = source[x * 4 + 3]; + } + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + default: + UNREACHABLE(); + } + break; + case D3DFMT_R5G6B5: + switch(getD3DFormat()) + { + case D3DFMT_X8R8G8B8: + for(int y = 0; y < height; y++) + { + for(int x = 0; x < width; x++) + { + unsigned short rgb = ((unsigned short*)source)[x]; + unsigned char red = (rgb & 0xF800) >> 8; + unsigned char green = (rgb & 0x07E0) >> 3; + unsigned char blue = (rgb & 0x001F) << 3; + dest[x + 0] = blue | (blue >> 5); + dest[x + 1] = green | (green >> 6); + dest[x + 2] = red | (red >> 5); + dest[x + 3] = 0xFF; + } + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + case D3DFMT_L8: + for(int y = 0; y < height; y++) + { + for(int x = 0; x < width; x++) + { + unsigned char red = source[x * 2 + 1] & 0xF8; + dest[x] = red | (red >> 5); + } + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + default: + UNREACHABLE(); + } + break; + case D3DFMT_A1R5G5B5: + switch(getD3DFormat()) + { + case D3DFMT_X8R8G8B8: + for(int y = 0; y < height; y++) + { + for(int x = 0; x < width; x++) + { + unsigned short argb = ((unsigned short*)source)[x]; + unsigned char red = (argb & 0x7C00) >> 7; + unsigned char green = (argb & 0x03E0) >> 2; + unsigned char blue = (argb & 0x001F) << 3; + dest[x + 0] = blue | (blue >> 5); + dest[x + 1] = green | (green >> 5); + dest[x + 2] = red | (red >> 5); + dest[x + 3] = 0xFF; + } + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + case D3DFMT_A8R8G8B8: + for(int y = 0; y < height; y++) + { + for(int x = 0; x < width; x++) + { + unsigned short argb = ((unsigned short*)source)[x]; + unsigned char red = (argb & 0x7C00) >> 7; + unsigned char green = (argb & 0x03E0) >> 2; + unsigned char blue = (argb & 0x001F) << 3; + unsigned char alpha = (signed short)argb >> 15; + dest[x + 0] = blue | (blue >> 5); + dest[x + 1] = green | (green >> 5); + dest[x + 2] = red | (red >> 5); + dest[x + 3] = alpha; + } + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + case D3DFMT_L8: + for(int y = 0; y < height; y++) + { + for(int x = 0; x < width; x++) + { + unsigned char red = source[x * 2 + 1] & 0x7C; + dest[x] = (red << 1) | (red >> 4); + } + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + case D3DFMT_A8L8: + for(int y = 0; y < height; y++) + { + for(int x = 0; x < width; x++) + { + unsigned char red = source[x * 2 + 1] & 0x7C; + dest[x * 2 + 0] = (red << 1) | (red >> 4); + dest[x * 2 + 1] = (signed char)source[x * 2 + 1] >> 7; + } + + source += sourceLock.Pitch; + dest += destLock.Pitch; + } + break; + default: + UNREACHABLE(); + } + break; + default: + UNREACHABLE(); + } + } + + unlock(); + renderTargetData->UnlockRect(); + + renderTargetData->Release(); + surface->Release(); + + mDirty = true; +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Image9.h b/src/3rdparty/angle/src/libGLESv2/renderer/Image9.h new file mode 100644 index 0000000000..2fbbca3124 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Image9.h @@ -0,0 +1,79 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Image9.h: Defines the rx::Image9 class, which acts as the interface to +// the actual underlying surfaces of a Texture. + +#ifndef LIBGLESV2_RENDERER_IMAGE9_H_ +#define LIBGLESV2_RENDERER_IMAGE9_H_ + +#include "libGLESv2/renderer/Image.h" +#include "common/debug.h" + +namespace gl +{ +class Framebuffer; +} + +namespace rx +{ +class Renderer; +class Renderer9; +class TextureStorageInterface2D; +class TextureStorageInterfaceCube; + +class Image9 : public Image +{ + public: + Image9(); + ~Image9(); + + static Image9 *makeImage9(Image *img); + + static void generateMipmap(Image9 *dest, Image9 *source); + static void generateMip(IDirect3DSurface9 *destSurface, IDirect3DSurface9 *sourceSurface); + static void copyLockableSurfaces(IDirect3DSurface9 *dest, IDirect3DSurface9 *source); + + virtual bool redefine(Renderer *renderer, GLint internalformat, GLsizei width, GLsizei height, bool forceRelease); + + virtual bool isRenderableFormat() const; + D3DFORMAT getD3DFormat() const; + + virtual bool isDirty() const {return mSurface && mDirty;} + IDirect3DSurface9 *getSurface(); + + virtual void setManagedSurface(TextureStorageInterface2D *storage, int level); + virtual void setManagedSurface(TextureStorageInterfaceCube *storage, int face, int level); + virtual bool updateSurface(TextureStorageInterface2D *storage, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height); + virtual bool updateSurface(TextureStorageInterfaceCube *storage, int face, int level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height); + + virtual void loadData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + GLint unpackAlignment, const void *input); + virtual void loadCompressedData(GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, + const void *input); + + virtual void copy(GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height, gl::Framebuffer *source); + + private: + DISALLOW_COPY_AND_ASSIGN(Image9); + + void createSurface(); + void setManagedSurface(IDirect3DSurface9 *surface); + bool updateSurface(IDirect3DSurface9 *dest, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height); + + HRESULT lock(D3DLOCKED_RECT *lockedRect, const RECT *rect); + void unlock(); + + Renderer9 *mRenderer; + + D3DPOOL mD3DPool; // can only be D3DPOOL_SYSTEMMEM or D3DPOOL_MANAGED since it needs to be lockable. + D3DFORMAT mD3DFormat; + + IDirect3DSurface9 *mSurface; +}; +} + +#endif // LIBGLESV2_RENDERER_IMAGE9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/ImageSSE2.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/ImageSSE2.cpp new file mode 100644 index 0000000000..b2a90ca961 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/ImageSSE2.cpp @@ -0,0 +1,100 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ImageSSE2.cpp: Implements SSE2-based functions of rx::Image class. It's +// in a separated file for GCC, which can enable SSE usage only per-file, +// not for code blocks that use SSE2 explicitly. + +#include "libGLESv2/Texture.h" +#include "libGLESv2/renderer/Image.h" + +namespace rx +{ + +void Image::loadRGBAUByteDataToBGRASSE2(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned int *source = NULL; + unsigned int *dest = NULL; + __m128i brMask = _mm_set1_epi32(0x00ff00ff); + + for (int y = 0; y < height; y++) + { + source = reinterpret_cast<const unsigned int*>(static_cast<const unsigned char*>(input) + y * inputPitch); + dest = reinterpret_cast<unsigned int*>(static_cast<unsigned char*>(output) + y * outputPitch); + int x = 0; + + // Make output writes aligned + for (x = 0; ((reinterpret_cast<intptr_t>(&dest[x]) & 15) != 0) && x < width; x++) + { + unsigned int rgba = source[x]; + dest[x] = (_rotl(rgba, 16) & 0x00ff00ff) | (rgba & 0xff00ff00); + } + + for (; x + 3 < width; x += 4) + { + __m128i sourceData = _mm_loadu_si128(reinterpret_cast<const __m128i*>(&source[x])); + // Mask out g and a, which don't change + __m128i gaComponents = _mm_andnot_si128(brMask, sourceData); + // Mask out b and r + __m128i brComponents = _mm_and_si128(sourceData, brMask); + // Swap b and r + __m128i brSwapped = _mm_shufflehi_epi16(_mm_shufflelo_epi16(brComponents, _MM_SHUFFLE(2, 3, 0, 1)), _MM_SHUFFLE(2, 3, 0, 1)); + __m128i result = _mm_or_si128(gaComponents, brSwapped); + _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x]), result); + } + + // Perform leftover writes + for (; x < width; x++) + { + unsigned int rgba = source[x]; + dest[x] = (_rotl(rgba, 16) & 0x00ff00ff) | (rgba & 0xff00ff00); + } + } +} + +void Image::loadAlphaDataToBGRASSE2(GLsizei width, GLsizei height, + int inputPitch, const void *input, size_t outputPitch, void *output) +{ + const unsigned char *source = NULL; + unsigned int *dest = NULL; + __m128i zeroWide = _mm_setzero_si128(); + + for (int y = 0; y < height; y++) + { + source = static_cast<const unsigned char*>(input) + y * inputPitch; + dest = reinterpret_cast<unsigned int*>(static_cast<unsigned char*>(output) + y * outputPitch); + + int x; + // Make output writes aligned + for (x = 0; ((reinterpret_cast<intptr_t>(&dest[x]) & 0xF) != 0 && x < width); x++) + { + dest[x] = static_cast<unsigned int>(source[x]) << 24; + } + + for (; x + 7 < width; x += 8) + { + __m128i sourceData = _mm_loadl_epi64(reinterpret_cast<const __m128i*>(&source[x])); + // Interleave each byte to 16bit, make the lower byte to zero + sourceData = _mm_unpacklo_epi8(zeroWide, sourceData); + // Interleave each 16bit to 32bit, make the lower 16bit to zero + __m128i lo = _mm_unpacklo_epi16(zeroWide, sourceData); + __m128i hi = _mm_unpackhi_epi16(zeroWide, sourceData); + + _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x]), lo); + _mm_store_si128(reinterpret_cast<__m128i*>(&dest[x + 4]), hi); + } + + // Handle the remainder + for (; x < width; x++) + { + dest[x] = static_cast<unsigned int>(source[x]) << 24; + } + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer.cpp new file mode 100644 index 0000000000..16fd782315 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer.cpp @@ -0,0 +1,202 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// IndexBuffer.cpp: Defines the abstract IndexBuffer class and IndexBufferInterface +// class with derivations, classes that perform graphics API agnostic index buffer operations. + +#include "libGLESv2/renderer/IndexBuffer.h" +#include "libGLESv2/renderer/Renderer.h" + +namespace rx +{ + +unsigned int IndexBuffer::mNextSerial = 1; + +IndexBuffer::IndexBuffer() +{ + updateSerial(); +} + +IndexBuffer::~IndexBuffer() +{ +} + +unsigned int IndexBuffer::getSerial() const +{ + return mSerial; +} + +void IndexBuffer::updateSerial() +{ + mSerial = mNextSerial++; +} + + +IndexBufferInterface::IndexBufferInterface(Renderer *renderer, bool dynamic) : mRenderer(renderer) +{ + mIndexBuffer = renderer->createIndexBuffer(); + + mDynamic = dynamic; + mWritePosition = 0; +} + +IndexBufferInterface::~IndexBufferInterface() +{ + if (mIndexBuffer) + { + delete mIndexBuffer; + } +} + +GLenum IndexBufferInterface::getIndexType() const +{ + return mIndexBuffer->getIndexType(); +} + +unsigned int IndexBufferInterface::getBufferSize() const +{ + return mIndexBuffer->getBufferSize(); +} + +unsigned int IndexBufferInterface::getSerial() const +{ + return mIndexBuffer->getSerial(); +} + +int IndexBufferInterface::mapBuffer(unsigned int size, void** outMappedMemory) +{ + if (!mIndexBuffer->mapBuffer(mWritePosition, size, outMappedMemory)) + { + *outMappedMemory = NULL; + return -1; + } + + int oldWritePos = static_cast<int>(mWritePosition); + mWritePosition += size; + + return oldWritePos; +} + +bool IndexBufferInterface::unmapBuffer() +{ + return mIndexBuffer->unmapBuffer(); +} + +IndexBuffer * IndexBufferInterface::getIndexBuffer() const +{ + return mIndexBuffer; +} + +unsigned int IndexBufferInterface::getWritePosition() const +{ + return mWritePosition; +} + +void IndexBufferInterface::setWritePosition(unsigned int writePosition) +{ + mWritePosition = writePosition; +} + +bool IndexBufferInterface::discard() +{ + return mIndexBuffer->discard(); +} + +bool IndexBufferInterface::setBufferSize(unsigned int bufferSize, GLenum indexType) +{ + if (mIndexBuffer->getBufferSize() == 0) + { + return mIndexBuffer->initialize(bufferSize, indexType, mDynamic); + } + else + { + return mIndexBuffer->setSize(bufferSize, indexType); + } +} + +StreamingIndexBufferInterface::StreamingIndexBufferInterface(Renderer *renderer) : IndexBufferInterface(renderer, true) +{ +} + +StreamingIndexBufferInterface::~StreamingIndexBufferInterface() +{ +} + +bool StreamingIndexBufferInterface::reserveBufferSpace(unsigned int size, GLenum indexType) +{ + bool result = true; + unsigned int curBufferSize = getBufferSize(); + if (size > curBufferSize) + { + result = setBufferSize(std::max(size, 2 * curBufferSize), indexType); + setWritePosition(0); + } + else if (getWritePosition() + size > curBufferSize) + { + if (!discard()) + { + return false; + } + setWritePosition(0); + } + + return result; +} + + +StaticIndexBufferInterface::StaticIndexBufferInterface(Renderer *renderer) : IndexBufferInterface(renderer, false) +{ +} + +StaticIndexBufferInterface::~StaticIndexBufferInterface() +{ +} + +bool StaticIndexBufferInterface::reserveBufferSpace(unsigned int size, GLenum indexType) +{ + unsigned int curSize = getBufferSize(); + if (curSize == 0) + { + return setBufferSize(size, indexType); + } + else if (curSize >= size && indexType == getIndexType()) + { + return true; + } + else + { + ERR("Static index buffers can't be resized"); + UNREACHABLE(); + return false; + } +} + +unsigned int StaticIndexBufferInterface::lookupRange(intptr_t offset, GLsizei count, unsigned int *minIndex, unsigned int *maxIndex) +{ + IndexRange range = {offset, count}; + + std::map<IndexRange, IndexResult>::iterator res = mCache.find(range); + + if (res == mCache.end()) + { + return -1; + } + + *minIndex = res->second.minIndex; + *maxIndex = res->second.maxIndex; + return res->second.streamOffset; +} + +void StaticIndexBufferInterface::addRange(intptr_t offset, GLsizei count, unsigned int minIndex, unsigned int maxIndex, unsigned int streamOffset) +{ + IndexRange indexRange = {offset, count}; + IndexResult indexResult = {minIndex, maxIndex, streamOffset}; + mCache[indexRange] = indexResult; +} + +} + diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer.h b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer.h new file mode 100644 index 0000000000..1afbd626fa --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer.h @@ -0,0 +1,137 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// IndexBuffer.h: Defines the abstract IndexBuffer class and IndexBufferInterface +// class with derivations, classes that perform graphics API agnostic index buffer operations. + +#ifndef LIBGLESV2_RENDERER_INDEXBUFFER_H_ +#define LIBGLESV2_RENDERER_INDEXBUFFER_H_ + +#include "common/angleutils.h" + +namespace rx +{ +class Renderer; + +class IndexBuffer +{ + public: + IndexBuffer(); + virtual ~IndexBuffer(); + + virtual bool initialize(unsigned int bufferSize, GLenum indexType, bool dynamic) = 0; + + virtual bool mapBuffer(unsigned int offset, unsigned int size, void** outMappedMemory) = 0; + virtual bool unmapBuffer() = 0; + + virtual bool discard() = 0; + + virtual GLenum getIndexType() const = 0; + virtual unsigned int getBufferSize() const = 0; + virtual bool setSize(unsigned int bufferSize, GLenum indexType) = 0; + + unsigned int getSerial() const; + + protected: + void updateSerial(); + + private: + DISALLOW_COPY_AND_ASSIGN(IndexBuffer); + + unsigned int mSerial; + static unsigned int mNextSerial; +}; + +class IndexBufferInterface +{ + public: + IndexBufferInterface(Renderer *renderer, bool dynamic); + virtual ~IndexBufferInterface(); + + virtual bool reserveBufferSpace(unsigned int size, GLenum indexType) = 0; + + GLenum getIndexType() const; + unsigned int getBufferSize() const; + + unsigned int getSerial() const; + + int mapBuffer(unsigned int size, void** outMappedMemory); + bool unmapBuffer(); + + IndexBuffer *getIndexBuffer() const; + + protected: + unsigned int getWritePosition() const; + void setWritePosition(unsigned int writePosition); + + bool discard(); + + bool setBufferSize(unsigned int bufferSize, GLenum indexType); + + private: + DISALLOW_COPY_AND_ASSIGN(IndexBufferInterface); + + rx::Renderer *const mRenderer; + + IndexBuffer* mIndexBuffer; + + unsigned int mWritePosition; + bool mDynamic; +}; + +class StreamingIndexBufferInterface : public IndexBufferInterface +{ + public: + StreamingIndexBufferInterface(Renderer *renderer); + ~StreamingIndexBufferInterface(); + + virtual bool reserveBufferSpace(unsigned int size, GLenum indexType); +}; + +class StaticIndexBufferInterface : public IndexBufferInterface +{ + public: + explicit StaticIndexBufferInterface(Renderer *renderer); + ~StaticIndexBufferInterface(); + + virtual bool reserveBufferSpace(unsigned int size, GLenum indexType); + + unsigned int lookupRange(intptr_t offset, GLsizei count, unsigned int *minIndex, unsigned int *maxIndex); // Returns the offset into the index buffer, or -1 if not found + void addRange(intptr_t offset, GLsizei count, unsigned int minIndex, unsigned int maxIndex, unsigned int streamOffset); + + private: + struct IndexRange + { + intptr_t offset; + GLsizei count; + + bool operator<(const IndexRange& rhs) const + { + if (offset != rhs.offset) + { + return offset < rhs.offset; + } + if (count != rhs.count) + { + return count < rhs.count; + } + return false; + } + }; + + struct IndexResult + { + unsigned int minIndex; + unsigned int maxIndex; + unsigned int streamOffset; + }; + + std::map<IndexRange, IndexResult> mCache; +}; + +} + +#endif // LIBGLESV2_RENDERER_INDEXBUFFER_H_
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer11.cpp new file mode 100644 index 0000000000..2a442ecd1a --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer11.cpp @@ -0,0 +1,182 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// IndexBuffer11.cpp: Defines the D3D11 IndexBuffer implementation. + +#include "libGLESv2/renderer/IndexBuffer11.h" +#include "libGLESv2/renderer/Renderer11.h" + +namespace rx +{ + +IndexBuffer11::IndexBuffer11(Renderer11 *const renderer) : mRenderer(renderer) +{ + mBuffer = NULL; + mBufferSize = 0; + mDynamicUsage = false; +} + +IndexBuffer11::~IndexBuffer11() +{ + if (mBuffer) + { + mBuffer->Release(); + mBuffer = NULL; + } +} + +bool IndexBuffer11::initialize(unsigned int bufferSize, GLenum indexType, bool dynamic) +{ + if (mBuffer) + { + mBuffer->Release(); + mBuffer = NULL; + } + + updateSerial(); + + if (bufferSize > 0) + { + ID3D11Device* dxDevice = mRenderer->getDevice(); + + D3D11_BUFFER_DESC bufferDesc; + bufferDesc.ByteWidth = bufferSize; + bufferDesc.Usage = D3D11_USAGE_DYNAMIC; + bufferDesc.BindFlags = D3D11_BIND_INDEX_BUFFER; + bufferDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + bufferDesc.MiscFlags = 0; + bufferDesc.StructureByteStride = 0; + + HRESULT result = dxDevice->CreateBuffer(&bufferDesc, NULL, &mBuffer); + if (FAILED(result)) + { + return false; + } + } + + mBufferSize = bufferSize; + mIndexType = indexType; + mDynamicUsage = dynamic; + + return true; +} + +IndexBuffer11 *IndexBuffer11::makeIndexBuffer11(IndexBuffer *indexBuffer) +{ + ASSERT(HAS_DYNAMIC_TYPE(IndexBuffer11*, indexBuffer)); + return static_cast<IndexBuffer11*>(indexBuffer); +} + +bool IndexBuffer11::mapBuffer(unsigned int offset, unsigned int size, void** outMappedMemory) +{ + if (mBuffer) + { + if (offset + size > mBufferSize) + { + ERR("Index buffer map range is not inside the buffer."); + return false; + } + + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_NO_OVERWRITE, 0, &mappedResource); + if (FAILED(result)) + { + ERR("Index buffer map failed with error 0x%08x", result); + return false; + } + + *outMappedMemory = reinterpret_cast<char*>(mappedResource.pData) + offset; + return true; + } + else + { + ERR("Index buffer not initialized."); + return false; + } +} + +bool IndexBuffer11::unmapBuffer() +{ + if (mBuffer) + { + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + dxContext->Unmap(mBuffer, 0); + return true; + } + else + { + ERR("Index buffer not initialized."); + return false; + } +} + +GLenum IndexBuffer11::getIndexType() const +{ + return mIndexType; +} + +unsigned int IndexBuffer11::getBufferSize() const +{ + return mBufferSize; +} + +bool IndexBuffer11::setSize(unsigned int bufferSize, GLenum indexType) +{ + if (bufferSize > mBufferSize || indexType != mIndexType) + { + return initialize(bufferSize, indexType, mDynamicUsage); + } + else + { + return true; + } +} + +bool IndexBuffer11::discard() +{ + if (mBuffer) + { + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + ERR("Index buffer map failed with error 0x%08x", result); + return false; + } + + dxContext->Unmap(mBuffer, 0); + + return true; + } + else + { + ERR("Index buffer not initialized."); + return false; + } +} + +DXGI_FORMAT IndexBuffer11::getIndexFormat() const +{ + switch (mIndexType) + { + case GL_UNSIGNED_BYTE: return DXGI_FORMAT_R16_UINT; + case GL_UNSIGNED_SHORT: return DXGI_FORMAT_R16_UINT; + case GL_UNSIGNED_INT: return DXGI_FORMAT_R32_UINT; + default: UNREACHABLE(); return DXGI_FORMAT_UNKNOWN; + } +} + +ID3D11Buffer *IndexBuffer11::getBuffer() const +{ + return mBuffer; +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer11.h b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer11.h new file mode 100644 index 0000000000..39a61946ad --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer11.h @@ -0,0 +1,53 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// IndexBuffer11.h: Defines the D3D11 IndexBuffer implementation. + +#ifndef LIBGLESV2_RENDERER_INDEXBUFFER11_H_ +#define LIBGLESV2_RENDERER_INDEXBUFFER11_H_ + +#include "libGLESv2/renderer/IndexBuffer.h" + +namespace rx +{ +class Renderer11; + +class IndexBuffer11 : public IndexBuffer +{ + public: + explicit IndexBuffer11(Renderer11 *const renderer); + virtual ~IndexBuffer11(); + + virtual bool initialize(unsigned int bufferSize, GLenum indexType, bool dynamic); + + static IndexBuffer11 *makeIndexBuffer11(IndexBuffer *indexBuffer); + + virtual bool mapBuffer(unsigned int offset, unsigned int size, void** outMappedMemory); + virtual bool unmapBuffer(); + + virtual GLenum getIndexType() const; + virtual unsigned int getBufferSize() const; + virtual bool setSize(unsigned int bufferSize, GLenum indexType); + + virtual bool discard(); + + DXGI_FORMAT getIndexFormat() const; + ID3D11Buffer *getBuffer() const; + + private: + DISALLOW_COPY_AND_ASSIGN(IndexBuffer11); + + rx::Renderer11 *const mRenderer; + + ID3D11Buffer *mBuffer; + unsigned int mBufferSize; + GLenum mIndexType; + bool mDynamicUsage; +}; + +} + +#endif // LIBGLESV2_RENDERER_INDEXBUFFER11_H_
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer9.cpp new file mode 100644 index 0000000000..c6d83c5dca --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer9.cpp @@ -0,0 +1,207 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Indexffer9.cpp: Defines the D3D9 IndexBuffer implementation. + +#include "libGLESv2/renderer/IndexBuffer9.h" +#include "libGLESv2/renderer/Renderer9.h" + +namespace rx +{ + +IndexBuffer9::IndexBuffer9(Renderer9 *const renderer) : mRenderer(renderer) +{ + mIndexBuffer = NULL; + mBufferSize = 0; + mIndexType = 0; + mDynamic = false; +} + +IndexBuffer9::~IndexBuffer9() +{ + if (mIndexBuffer) + { + mIndexBuffer->Release(); + mIndexBuffer = NULL; + } +} + +bool IndexBuffer9::initialize(unsigned int bufferSize, GLenum indexType, bool dynamic) +{ + if (mIndexBuffer) + { + mIndexBuffer->Release(); + mIndexBuffer = NULL; + } + + updateSerial(); + + if (bufferSize > 0) + { + D3DFORMAT format; + if (indexType == GL_UNSIGNED_SHORT || indexType == GL_UNSIGNED_BYTE) + { + format = D3DFMT_INDEX16; + } + else if (indexType == GL_UNSIGNED_INT) + { + if (mRenderer->get32BitIndexSupport()) + { + format = D3DFMT_INDEX32; + } + else + { + ERR("Attempted to create a 32-bit index buffer but renderer does not support 32-bit indices."); + return false; + } + } + else + { + ERR("Invalid index type %u.", indexType); + return false; + } + + DWORD usageFlags = D3DUSAGE_WRITEONLY; + if (dynamic) + { + usageFlags |= D3DUSAGE_DYNAMIC; + } + + HRESULT result = mRenderer->createIndexBuffer(bufferSize, usageFlags, format, &mIndexBuffer); + if (FAILED(result)) + { + ERR("Failed to create an index buffer of size %u, result: 0x%08x.", mBufferSize, result); + return false; + } + } + + mBufferSize = bufferSize; + mIndexType = indexType; + mDynamic = dynamic; + + return true; +} + +IndexBuffer9 *IndexBuffer9::makeIndexBuffer9(IndexBuffer *indexBuffer) +{ + ASSERT(HAS_DYNAMIC_TYPE(IndexBuffer9*, indexBuffer)); + return static_cast<IndexBuffer9*>(indexBuffer); +} + +bool IndexBuffer9::mapBuffer(unsigned int offset, unsigned int size, void** outMappedMemory) +{ + if (mIndexBuffer) + { + DWORD lockFlags = mDynamic ? D3DLOCK_NOOVERWRITE : 0; + + void *mapPtr = NULL; + HRESULT result = mIndexBuffer->Lock(offset, size, &mapPtr, lockFlags); + if (FAILED(result)) + { + ERR("Index buffer lock failed with error 0x%08x", result); + return false; + } + + *outMappedMemory = mapPtr; + return true; + } + else + { + ERR("Index buffer not initialized."); + return false; + } +} + +bool IndexBuffer9::unmapBuffer() +{ + if (mIndexBuffer) + { + HRESULT result = mIndexBuffer->Unlock(); + if (FAILED(result)) + { + ERR("Index buffer unlock failed with error 0x%08x", result); + return false; + } + + return true; + } + else + { + ERR("Index buffer not initialized."); + return false; + } +} + +GLenum IndexBuffer9::getIndexType() const +{ + return mIndexType; +} + +unsigned int IndexBuffer9::getBufferSize() const +{ + return mBufferSize; +} + +bool IndexBuffer9::setSize(unsigned int bufferSize, GLenum indexType) +{ + if (bufferSize > mBufferSize || indexType != mIndexType) + { + return initialize(bufferSize, indexType, mDynamic); + } + else + { + return true; + } +} + +bool IndexBuffer9::discard() +{ + if (mIndexBuffer) + { + void *dummy; + HRESULT result; + + result = mIndexBuffer->Lock(0, 1, &dummy, D3DLOCK_DISCARD); + if (FAILED(result)) + { + ERR("Discard lock failed with error 0x%08x", result); + return false; + } + + result = mIndexBuffer->Unlock(); + if (FAILED(result)) + { + ERR("Discard unlock failed with error 0x%08x", result); + return false; + } + + return true; + } + else + { + ERR("Index buffer not initialized."); + return false; + } +} + +D3DFORMAT IndexBuffer9::getIndexFormat() const +{ + switch (mIndexType) + { + case GL_UNSIGNED_BYTE: return D3DFMT_INDEX16; + case GL_UNSIGNED_SHORT: return D3DFMT_INDEX16; + case GL_UNSIGNED_INT: return D3DFMT_INDEX32; + default: UNREACHABLE(); return D3DFMT_UNKNOWN; + } +} + +IDirect3DIndexBuffer9 * IndexBuffer9::getBuffer() const +{ + return mIndexBuffer; +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer9.h b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer9.h new file mode 100644 index 0000000000..6801867532 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/IndexBuffer9.h @@ -0,0 +1,53 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Indexffer9.h: Defines the D3D9 IndexBuffer implementation. + +#ifndef LIBGLESV2_RENDERER_INDEXBUFFER9_H_ +#define LIBGLESV2_RENDERER_INDEXBUFFER9_H_ + +#include "libGLESv2/renderer/IndexBuffer.h" + +namespace rx +{ +class Renderer9; + +class IndexBuffer9 : public IndexBuffer +{ + public: + explicit IndexBuffer9(Renderer9 *const renderer); + virtual ~IndexBuffer9(); + + virtual bool initialize(unsigned int bufferSize, GLenum indexType, bool dynamic); + + static IndexBuffer9 *makeIndexBuffer9(IndexBuffer *indexBuffer); + + virtual bool mapBuffer(unsigned int offset, unsigned int size, void** outMappedMemory); + virtual bool unmapBuffer(); + + virtual GLenum getIndexType() const; + virtual unsigned int getBufferSize() const; + virtual bool setSize(unsigned int bufferSize, GLenum indexType); + + virtual bool discard(); + + D3DFORMAT getIndexFormat() const; + IDirect3DIndexBuffer9 *getBuffer() const; + + private: + DISALLOW_COPY_AND_ASSIGN(IndexBuffer9); + + rx::Renderer9 *const mRenderer; + + IDirect3DIndexBuffer9 *mIndexBuffer; + unsigned int mBufferSize; + GLenum mIndexType; + bool mDynamic; +}; + +} + +#endif // LIBGLESV2_RENDERER_INDEXBUFFER9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/IndexDataManager.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/IndexDataManager.cpp new file mode 100644 index 0000000000..84b79b4109 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/IndexDataManager.cpp @@ -0,0 +1,316 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// IndexDataManager.cpp: Defines the IndexDataManager, a class that +// runs the Buffer translation process for index buffers. + +#include "libGLESv2/renderer/IndexDataManager.h" +#include "libGLESv2/renderer/BufferStorage.h" + +#include "libGLESv2/Buffer.h" +#include "libGLESv2/main.h" +#include "libGLESv2/renderer/IndexBuffer.h" + +namespace rx +{ + +IndexDataManager::IndexDataManager(Renderer *renderer) : mRenderer(renderer) +{ + mStreamingBufferShort = new StreamingIndexBufferInterface(mRenderer); + if (!mStreamingBufferShort->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_SHORT)) + { + delete mStreamingBufferShort; + mStreamingBufferShort = NULL; + } + + mStreamingBufferInt = new StreamingIndexBufferInterface(mRenderer); + if (!mStreamingBufferInt->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_INT)) + { + delete mStreamingBufferInt; + mStreamingBufferInt = NULL; + } + + if (!mStreamingBufferShort) + { + // Make sure both buffers are deleted. + delete mStreamingBufferInt; + mStreamingBufferInt = NULL; + + ERR("Failed to allocate the streaming index buffer(s)."); + } + + mCountingBuffer = NULL; +} + +IndexDataManager::~IndexDataManager() +{ + delete mStreamingBufferShort; + delete mStreamingBufferInt; + delete mCountingBuffer; +} + +static unsigned int indexTypeSize(GLenum type) +{ + switch (type) + { + case GL_UNSIGNED_INT: return sizeof(GLuint); + case GL_UNSIGNED_SHORT: return sizeof(GLushort); + case GL_UNSIGNED_BYTE: return sizeof(GLubyte); + default: UNREACHABLE(); return sizeof(GLushort); + } +} + +static void convertIndices(GLenum type, const void *input, GLsizei count, void *output) +{ + if (type == GL_UNSIGNED_BYTE) + { + const GLubyte *in = static_cast<const GLubyte*>(input); + GLushort *out = static_cast<GLushort*>(output); + + for (GLsizei i = 0; i < count; i++) + { + out[i] = in[i]; + } + } + else if (type == GL_UNSIGNED_INT) + { + memcpy(output, input, count * sizeof(GLuint)); + } + else if (type == GL_UNSIGNED_SHORT) + { + memcpy(output, input, count * sizeof(GLushort)); + } + else UNREACHABLE(); +} + +template <class IndexType> +static void computeRange(const IndexType *indices, GLsizei count, GLuint *minIndex, GLuint *maxIndex) +{ + *minIndex = indices[0]; + *maxIndex = indices[0]; + + for (GLsizei i = 0; i < count; i++) + { + if (*minIndex > indices[i]) *minIndex = indices[i]; + if (*maxIndex < indices[i]) *maxIndex = indices[i]; + } +} + +static void computeRange(GLenum type, const GLvoid *indices, GLsizei count, GLuint *minIndex, GLuint *maxIndex) +{ + if (type == GL_UNSIGNED_BYTE) + { + computeRange(static_cast<const GLubyte*>(indices), count, minIndex, maxIndex); + } + else if (type == GL_UNSIGNED_INT) + { + computeRange(static_cast<const GLuint*>(indices), count, minIndex, maxIndex); + } + else if (type == GL_UNSIGNED_SHORT) + { + computeRange(static_cast<const GLushort*>(indices), count, minIndex, maxIndex); + } + else UNREACHABLE(); +} + +GLenum IndexDataManager::prepareIndexData(GLenum type, GLsizei count, gl::Buffer *buffer, const GLvoid *indices, TranslatedIndexData *translated) +{ + if (!mStreamingBufferShort) + { + return GL_OUT_OF_MEMORY; + } + + GLenum destinationIndexType = (type == GL_UNSIGNED_INT) ? GL_UNSIGNED_INT : GL_UNSIGNED_SHORT; + intptr_t offset = reinterpret_cast<intptr_t>(indices); + bool alignedOffset = false; + + BufferStorage *storage = NULL; + + if (buffer != NULL) + { + storage = buffer->getStorage(); + + switch (type) + { + case GL_UNSIGNED_BYTE: alignedOffset = (offset % sizeof(GLubyte) == 0); break; + case GL_UNSIGNED_SHORT: alignedOffset = (offset % sizeof(GLushort) == 0); break; + case GL_UNSIGNED_INT: alignedOffset = (offset % sizeof(GLuint) == 0); break; + default: UNREACHABLE(); alignedOffset = false; + } + + if (indexTypeSize(type) * count + offset > storage->getSize()) + { + return GL_INVALID_OPERATION; + } + + indices = static_cast<const GLubyte*>(storage->getData()) + offset; + } + + StreamingIndexBufferInterface *streamingBuffer = (type == GL_UNSIGNED_INT) ? mStreamingBufferInt : mStreamingBufferShort; + + StaticIndexBufferInterface *staticBuffer = buffer ? buffer->getStaticIndexBuffer() : NULL; + IndexBufferInterface *indexBuffer = streamingBuffer; + bool directStorage = alignedOffset && storage && storage->supportsDirectBinding() && + destinationIndexType == type; + UINT streamOffset = 0; + + if (directStorage) + { + indexBuffer = streamingBuffer; + streamOffset = offset; + storage->markBufferUsage(); + computeRange(type, indices, count, &translated->minIndex, &translated->maxIndex); + } + else if (staticBuffer && staticBuffer->getBufferSize() != 0 && staticBuffer->getIndexType() == type && alignedOffset) + { + indexBuffer = staticBuffer; + streamOffset = staticBuffer->lookupRange(offset, count, &translated->minIndex, &translated->maxIndex); + + if (streamOffset == -1) + { + streamOffset = (offset / indexTypeSize(type)) * indexTypeSize(destinationIndexType); + computeRange(type, indices, count, &translated->minIndex, &translated->maxIndex); + staticBuffer->addRange(offset, count, translated->minIndex, translated->maxIndex, streamOffset); + } + } + else + { + int convertCount = count; + + if (staticBuffer) + { + if (staticBuffer->getBufferSize() == 0 && alignedOffset) + { + indexBuffer = staticBuffer; + convertCount = storage->getSize() / indexTypeSize(type); + } + else + { + buffer->invalidateStaticData(); + staticBuffer = NULL; + } + } + + if (!indexBuffer) + { + ERR("No valid index buffer."); + return GL_INVALID_OPERATION; + } + + unsigned int bufferSizeRequired = convertCount * indexTypeSize(destinationIndexType); + indexBuffer->reserveBufferSpace(bufferSizeRequired, type); + + void* output = NULL; + streamOffset = indexBuffer->mapBuffer(bufferSizeRequired, &output); + if (streamOffset == -1 || output == NULL) + { + ERR("Failed to map index buffer."); + return GL_OUT_OF_MEMORY; + } + + convertIndices(type, staticBuffer ? storage->getData() : indices, convertCount, output); + + if (!indexBuffer->unmapBuffer()) + { + ERR("Failed to unmap index buffer."); + return GL_OUT_OF_MEMORY; + } + + computeRange(type, indices, count, &translated->minIndex, &translated->maxIndex); + + if (staticBuffer) + { + streamOffset = (offset / indexTypeSize(type)) * indexTypeSize(destinationIndexType); + staticBuffer->addRange(offset, count, translated->minIndex, translated->maxIndex, streamOffset); + } + } + + translated->storage = directStorage ? storage : NULL; + translated->indexBuffer = indexBuffer->getIndexBuffer(); + translated->serial = directStorage ? storage->getSerial() : indexBuffer->getSerial(); + translated->startIndex = streamOffset / indexTypeSize(destinationIndexType); + translated->startOffset = streamOffset; + + if (buffer) + { + buffer->promoteStaticUsage(count * indexTypeSize(type)); + } + + return GL_NO_ERROR; +} + +StaticIndexBufferInterface *IndexDataManager::getCountingIndices(GLsizei count) +{ + if (count <= 65536) // 16-bit indices + { + const unsigned int spaceNeeded = count * sizeof(unsigned short); + + if (!mCountingBuffer || mCountingBuffer->getBufferSize() < spaceNeeded) + { + delete mCountingBuffer; + mCountingBuffer = new StaticIndexBufferInterface(mRenderer); + mCountingBuffer->reserveBufferSpace(spaceNeeded, GL_UNSIGNED_SHORT); + + void* mappedMemory = NULL; + if (mCountingBuffer->mapBuffer(spaceNeeded, &mappedMemory) == -1 || mappedMemory == NULL) + { + ERR("Failed to map counting buffer."); + return NULL; + } + + unsigned short *data = reinterpret_cast<unsigned short*>(mappedMemory); + for(int i = 0; i < count; i++) + { + data[i] = i; + } + + if (!mCountingBuffer->unmapBuffer()) + { + ERR("Failed to unmap counting buffer."); + return NULL; + } + } + } + else if (mStreamingBufferInt) // 32-bit indices supported + { + const unsigned int spaceNeeded = count * sizeof(unsigned int); + + if (!mCountingBuffer || mCountingBuffer->getBufferSize() < spaceNeeded) + { + delete mCountingBuffer; + mCountingBuffer = new StaticIndexBufferInterface(mRenderer); + mCountingBuffer->reserveBufferSpace(spaceNeeded, GL_UNSIGNED_INT); + + void* mappedMemory = NULL; + if (mCountingBuffer->mapBuffer(spaceNeeded, &mappedMemory) == -1 || mappedMemory == NULL) + { + ERR("Failed to map counting buffer."); + return NULL; + } + + unsigned int *data = reinterpret_cast<unsigned int*>(mappedMemory); + for(int i = 0; i < count; i++) + { + data[i] = i; + } + + if (!mCountingBuffer->unmapBuffer()) + { + ERR("Failed to unmap counting buffer."); + return NULL; + } + } + } + else + { + return NULL; + } + + return mCountingBuffer; +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/IndexDataManager.h b/src/3rdparty/angle/src/libGLESv2/renderer/IndexDataManager.h new file mode 100644 index 0000000000..0e77c81d1b --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/IndexDataManager.h @@ -0,0 +1,66 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// IndexDataManager.h: Defines the IndexDataManager, a class that +// runs the Buffer translation process for index buffers. + +#ifndef LIBGLESV2_INDEXDATAMANAGER_H_ +#define LIBGLESV2_INDEXDATAMANAGER_H_ + +#include "common/angleutils.h" + +namespace +{ + enum { INITIAL_INDEX_BUFFER_SIZE = 4096 * sizeof(GLuint) }; +} + +namespace gl +{ +class Buffer; +} + +namespace rx +{ +class StaticIndexBufferInterface; +class StreamingIndexBufferInterface; +class IndexBuffer; +class BufferStorage; +class Renderer; + +struct TranslatedIndexData +{ + unsigned int minIndex; + unsigned int maxIndex; + unsigned int startIndex; + unsigned int startOffset; // In bytes + + IndexBuffer *indexBuffer; + BufferStorage *storage; + unsigned int serial; +}; + +class IndexDataManager +{ + public: + explicit IndexDataManager(Renderer *renderer); + virtual ~IndexDataManager(); + + GLenum prepareIndexData(GLenum type, GLsizei count, gl::Buffer *arrayElementBuffer, const GLvoid *indices, TranslatedIndexData *translated); + StaticIndexBufferInterface *getCountingIndices(GLsizei count); + + private: + DISALLOW_COPY_AND_ASSIGN(IndexDataManager); + + Renderer *const mRenderer; + + StreamingIndexBufferInterface *mStreamingBufferShort; + StreamingIndexBufferInterface *mStreamingBufferInt; + StaticIndexBufferInterface *mCountingBuffer; +}; + +} + +#endif // LIBGLESV2_INDEXDATAMANAGER_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/InputLayoutCache.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/InputLayoutCache.cpp new file mode 100644 index 0000000000..0402bb35ac --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/InputLayoutCache.cpp @@ -0,0 +1,166 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// InputLayoutCache.cpp: Defines InputLayoutCache, a class that builds and caches +// D3D11 input layouts. + +#include "libGLESv2/renderer/InputLayoutCache.h" +#include "libGLESv2/renderer/VertexBuffer11.h" +#include "libGLESv2/renderer/BufferStorage11.h" +#include "libGLESv2/renderer/ShaderExecutable11.h" +#include "libGLESv2/ProgramBinary.h" +#include "libGLESv2/Context.h" +#include "libGLESv2/renderer/VertexDataManager.h" + +#include "third_party/murmurhash/MurmurHash3.h" + +namespace rx +{ + +const unsigned int InputLayoutCache::kMaxInputLayouts = 1024; + +InputLayoutCache::InputLayoutCache() : mInputLayoutMap(kMaxInputLayouts, hashInputLayout, compareInputLayouts) +{ + mCounter = 0; + mDevice = NULL; + mDeviceContext = NULL; +} + +InputLayoutCache::~InputLayoutCache() +{ + clear(); +} + +void InputLayoutCache::initialize(ID3D11Device *device, ID3D11DeviceContext *context) +{ + clear(); + mDevice = device; + mDeviceContext = context; +} + +void InputLayoutCache::clear() +{ + for (InputLayoutMap::iterator i = mInputLayoutMap.begin(); i != mInputLayoutMap.end(); i++) + { + i->second.inputLayout->Release(); + } + mInputLayoutMap.clear(); +} + +GLenum InputLayoutCache::applyVertexBuffers(TranslatedAttribute attributes[gl::MAX_VERTEX_ATTRIBS], + gl::ProgramBinary *programBinary) +{ + int sortedSemanticIndices[gl::MAX_VERTEX_ATTRIBS]; + programBinary->sortAttributesByLayout(attributes, sortedSemanticIndices); + + if (!mDevice || !mDeviceContext) + { + ERR("InputLayoutCache is not initialized."); + return GL_INVALID_OPERATION; + } + + InputLayoutKey ilKey = { 0 }; + + ID3D11Buffer *vertexBuffers[gl::MAX_VERTEX_ATTRIBS] = { NULL }; + UINT vertexStrides[gl::MAX_VERTEX_ATTRIBS] = { 0 }; + UINT vertexOffsets[gl::MAX_VERTEX_ATTRIBS] = { 0 }; + + static const char* semanticName = "TEXCOORD"; + + for (unsigned int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + if (attributes[i].active) + { + VertexBuffer11 *vertexBuffer = VertexBuffer11::makeVertexBuffer11(attributes[i].vertexBuffer); + BufferStorage11 *bufferStorage = attributes[i].storage ? BufferStorage11::makeBufferStorage11(attributes[i].storage) : NULL; + + D3D11_INPUT_CLASSIFICATION inputClass = attributes[i].divisor > 0 ? D3D11_INPUT_PER_INSTANCE_DATA : D3D11_INPUT_PER_VERTEX_DATA; + + // Record the type of the associated vertex shader vector in our key + // This will prevent mismatched vertex shaders from using the same input layout + GLint attributeSize; + programBinary->getActiveAttribute(ilKey.elementCount, 0, NULL, &attributeSize, &ilKey.glslElementType[ilKey.elementCount], NULL); + + ilKey.elements[ilKey.elementCount].SemanticName = semanticName; + ilKey.elements[ilKey.elementCount].SemanticIndex = sortedSemanticIndices[i]; + ilKey.elements[ilKey.elementCount].Format = attributes[i].attribute->mArrayEnabled ? vertexBuffer->getDXGIFormat(*attributes[i].attribute) : DXGI_FORMAT_R32G32B32A32_FLOAT; + ilKey.elements[ilKey.elementCount].InputSlot = i; + ilKey.elements[ilKey.elementCount].AlignedByteOffset = 0; + ilKey.elements[ilKey.elementCount].InputSlotClass = inputClass; + ilKey.elements[ilKey.elementCount].InstanceDataStepRate = attributes[i].divisor; + ilKey.elementCount++; + + vertexBuffers[i] = bufferStorage ? bufferStorage->getBuffer() : vertexBuffer->getBuffer(); + vertexStrides[i] = attributes[i].stride; + vertexOffsets[i] = attributes[i].offset; + } + } + + ID3D11InputLayout *inputLayout = NULL; + + InputLayoutMap::iterator i = mInputLayoutMap.find(ilKey); + if (i != mInputLayoutMap.end()) + { + inputLayout = i->second.inputLayout; + i->second.lastUsedTime = mCounter++; + } + else + { + ShaderExecutable11 *shader = ShaderExecutable11::makeShaderExecutable11(programBinary->getVertexExecutable()); + + HRESULT result = mDevice->CreateInputLayout(ilKey.elements, ilKey.elementCount, shader->getFunction(), shader->getLength(), &inputLayout); + if (FAILED(result)) + { + ERR("Failed to crate input layout, result: 0x%08x", result); + return GL_INVALID_OPERATION; + } + + if (mInputLayoutMap.size() >= kMaxInputLayouts) + { + TRACE("Overflowed the limit of %u input layouts, removing the least recently used " + "to make room.", kMaxInputLayouts); + + InputLayoutMap::iterator leastRecentlyUsed = mInputLayoutMap.begin(); + for (InputLayoutMap::iterator i = mInputLayoutMap.begin(); i != mInputLayoutMap.end(); i++) + { + if (i->second.lastUsedTime < leastRecentlyUsed->second.lastUsedTime) + { + leastRecentlyUsed = i; + } + } + leastRecentlyUsed->second.inputLayout->Release(); + mInputLayoutMap.erase(leastRecentlyUsed); + } + + InputLayoutCounterPair inputCounterPair; + inputCounterPair.inputLayout = inputLayout; + inputCounterPair.lastUsedTime = mCounter++; + + mInputLayoutMap.insert(std::make_pair(ilKey, inputCounterPair)); + } + + mDeviceContext->IASetInputLayout(inputLayout); + mDeviceContext->IASetVertexBuffers(0, gl::MAX_VERTEX_ATTRIBS, vertexBuffers, vertexStrides, vertexOffsets); + + return GL_NO_ERROR; +} + +std::size_t InputLayoutCache::hashInputLayout(const InputLayoutKey &inputLayout) +{ + static const unsigned int seed = 0xDEADBEEF; + + std::size_t hash = 0; + MurmurHash3_x86_32(&inputLayout, sizeof(InputLayoutKey), seed, &hash); + return hash; +} + +bool InputLayoutCache::compareInputLayouts(const InputLayoutKey &a, const InputLayoutKey &b) +{ + return memcmp(&a, &b, sizeof(InputLayoutKey)) == 0; +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/InputLayoutCache.h b/src/3rdparty/angle/src/libGLESv2/renderer/InputLayoutCache.h new file mode 100644 index 0000000000..d95f39fae4 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/InputLayoutCache.h @@ -0,0 +1,74 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// InputLayoutCache.h: Defines InputLayoutCache, a class that builds and caches +// D3D11 input layouts. + +#ifndef LIBGLESV2_RENDERER_INPUTLAYOUTCACHE_H_ +#define LIBGLESV2_RENDERER_INPUTLAYOUTCACHE_H_ + +#include "libGLESv2/Constants.h" +#include "common/angleutils.h" + +namespace gl +{ +class ProgramBinary; +} + +namespace rx +{ +struct TranslatedAttribute; + +class InputLayoutCache +{ + public: + InputLayoutCache(); + virtual ~InputLayoutCache(); + + void initialize(ID3D11Device *device, ID3D11DeviceContext *context); + void clear(); + + GLenum applyVertexBuffers(TranslatedAttribute attributes[gl::MAX_VERTEX_ATTRIBS], + gl::ProgramBinary *programBinary); + + private: + DISALLOW_COPY_AND_ASSIGN(InputLayoutCache); + + struct InputLayoutKey + { + unsigned int elementCount; + D3D11_INPUT_ELEMENT_DESC elements[gl::MAX_VERTEX_ATTRIBS]; + GLenum glslElementType[gl::MAX_VERTEX_ATTRIBS]; + }; + + struct InputLayoutCounterPair + { + ID3D11InputLayout *inputLayout; + unsigned long long lastUsedTime; + }; + + static std::size_t hashInputLayout(const InputLayoutKey &inputLayout); + static bool compareInputLayouts(const InputLayoutKey &a, const InputLayoutKey &b); + + typedef std::size_t (*InputLayoutHashFunction)(const InputLayoutKey &); + typedef bool (*InputLayoutEqualityFunction)(const InputLayoutKey &, const InputLayoutKey &); + typedef std::unordered_map<InputLayoutKey, + InputLayoutCounterPair, + InputLayoutHashFunction, + InputLayoutEqualityFunction> InputLayoutMap; + InputLayoutMap mInputLayoutMap; + + static const unsigned int kMaxInputLayouts; + + unsigned long long mCounter; + + ID3D11Device *mDevice; + ID3D11DeviceContext *mDeviceContext; +}; + +} + +#endif // LIBGLESV2_RENDERER_INPUTLAYOUTCACHE_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Query11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Query11.cpp new file mode 100644 index 0000000000..13210fc929 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Query11.cpp @@ -0,0 +1,122 @@ +#include "precompiled.h" +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Query11.cpp: Defines the rx::Query11 class which implements rx::QueryImpl. + +#include "libGLESv2/renderer/Query11.h" +#include "libGLESv2/renderer/Renderer11.h" +#include "libGLESv2/main.h" + +namespace rx +{ + +Query11::Query11(rx::Renderer11 *renderer, GLenum type) : QueryImpl(type) +{ + mRenderer = renderer; + mQuery = NULL; +} + +Query11::~Query11() +{ + if (mQuery) + { + mQuery->Release(); + mQuery = NULL; + } +} + +void Query11::begin() +{ + if (mQuery == NULL) + { + D3D11_QUERY_DESC queryDesc; + queryDesc.Query = D3D11_QUERY_OCCLUSION; + queryDesc.MiscFlags = 0; + + if (FAILED(mRenderer->getDevice()->CreateQuery(&queryDesc, &mQuery))) + { + return gl::error(GL_OUT_OF_MEMORY); + } + } + + mRenderer->getDeviceContext()->Begin(mQuery); +} + +void Query11::end() +{ + if (mQuery == NULL) + { + return gl::error(GL_INVALID_OPERATION); + } + + mRenderer->getDeviceContext()->End(mQuery); + + mStatus = GL_FALSE; + mResult = GL_FALSE; +} + +GLuint Query11::getResult() +{ + if (mQuery != NULL) + { + while (!testQuery()) + { + Sleep(0); + // explicitly check for device loss, some drivers seem to return S_FALSE + // if the device is lost + if (mRenderer->testDeviceLost(true)) + { + return gl::error(GL_OUT_OF_MEMORY, 0); + } + } + } + + return mResult; +} + +GLboolean Query11::isResultAvailable() +{ + if (mQuery != NULL) + { + testQuery(); + } + + return mStatus; +} + +GLboolean Query11::testQuery() +{ + if (mQuery != NULL && mStatus != GL_TRUE) + { + UINT64 numPixels = 0; + HRESULT result = mRenderer->getDeviceContext()->GetData(mQuery, &numPixels, sizeof(UINT64), 0); + if (result == S_OK) + { + mStatus = GL_TRUE; + + switch (getType()) + { + case GL_ANY_SAMPLES_PASSED_EXT: + case GL_ANY_SAMPLES_PASSED_CONSERVATIVE_EXT: + mResult = (numPixels > 0) ? GL_TRUE : GL_FALSE; + break; + default: + UNREACHABLE(); + } + } + else if (mRenderer->testDeviceLost(true)) + { + return gl::error(GL_OUT_OF_MEMORY, GL_TRUE); + } + + return mStatus; + } + + return GL_TRUE; // prevent blocking when query is null +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Query11.h b/src/3rdparty/angle/src/libGLESv2/renderer/Query11.h new file mode 100644 index 0000000000..0a03de77ca --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Query11.h @@ -0,0 +1,40 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Query11.h: Defines the rx::Query11 class which implements rx::QueryImpl. + +#ifndef LIBGLESV2_RENDERER_QUERY11_H_ +#define LIBGLESV2_RENDERER_QUERY11_H_ + +#include "libGLESv2/renderer/QueryImpl.h" + +namespace rx +{ +class Renderer11; + +class Query11 : public QueryImpl +{ + public: + Query11(rx::Renderer11 *renderer, GLenum type); + virtual ~Query11(); + + void begin(); + void end(); + GLuint getResult(); + GLboolean isResultAvailable(); + + private: + DISALLOW_COPY_AND_ASSIGN(Query11); + + GLboolean testQuery(); + + rx::Renderer11 *mRenderer; + ID3D11Query *mQuery; +}; + +} + +#endif // LIBGLESV2_RENDERER_QUERY11_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Query9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Query9.cpp new file mode 100644 index 0000000000..ef694267dd --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Query9.cpp @@ -0,0 +1,125 @@ +#include "precompiled.h" +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Query9.cpp: Defines the rx::Query9 class which implements rx::QueryImpl. + + +#include "libGLESv2/renderer/Query9.h" +#include "libGLESv2/main.h" +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/renderer/Renderer9.h" + +namespace rx +{ + +Query9::Query9(rx::Renderer9 *renderer, GLenum type) : QueryImpl(type) +{ + mRenderer = renderer; + mQuery = NULL; +} + +Query9::~Query9() +{ + if (mQuery) + { + mQuery->Release(); + mQuery = NULL; + } +} + +void Query9::begin() +{ + if (mQuery == NULL) + { + if (FAILED(mRenderer->getDevice()->CreateQuery(D3DQUERYTYPE_OCCLUSION, &mQuery))) + { + return gl::error(GL_OUT_OF_MEMORY); + } + } + + HRESULT result = mQuery->Issue(D3DISSUE_BEGIN); + ASSERT(SUCCEEDED(result)); +} + +void Query9::end() +{ + if (mQuery == NULL) + { + return gl::error(GL_INVALID_OPERATION); + } + + HRESULT result = mQuery->Issue(D3DISSUE_END); + ASSERT(SUCCEEDED(result)); + + mStatus = GL_FALSE; + mResult = GL_FALSE; +} + +GLuint Query9::getResult() +{ + if (mQuery != NULL) + { + while (!testQuery()) + { + Sleep(0); + // explicitly check for device loss + // some drivers seem to return S_FALSE even if the device is lost + // instead of D3DERR_DEVICELOST like they should + if (mRenderer->testDeviceLost(true)) + { + return gl::error(GL_OUT_OF_MEMORY, 0); + } + } + } + + return mResult; +} + +GLboolean Query9::isResultAvailable() +{ + if (mQuery != NULL) + { + testQuery(); + } + + return mStatus; +} + +GLboolean Query9::testQuery() +{ + if (mQuery != NULL && mStatus != GL_TRUE) + { + DWORD numPixels = 0; + + HRESULT hres = mQuery->GetData(&numPixels, sizeof(DWORD), D3DGETDATA_FLUSH); + if (hres == S_OK) + { + mStatus = GL_TRUE; + + switch (getType()) + { + case GL_ANY_SAMPLES_PASSED_EXT: + case GL_ANY_SAMPLES_PASSED_CONSERVATIVE_EXT: + mResult = (numPixels > 0) ? GL_TRUE : GL_FALSE; + break; + default: + ASSERT(false); + } + } + else if (d3d9::isDeviceLostError(hres)) + { + mRenderer->notifyDeviceLost(); + return gl::error(GL_OUT_OF_MEMORY, GL_TRUE); + } + + return mStatus; + } + + return GL_TRUE; // prevent blocking when query is null +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Query9.h b/src/3rdparty/angle/src/libGLESv2/renderer/Query9.h new file mode 100644 index 0000000000..47eef89336 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Query9.h @@ -0,0 +1,40 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Query9.h: Defines the rx::Query9 class which implements rx::QueryImpl. + +#ifndef LIBGLESV2_RENDERER_QUERY9_H_ +#define LIBGLESV2_RENDERER_QUERY9_H_ + +#include "libGLESv2/renderer/QueryImpl.h" + +namespace rx +{ +class Renderer9; + +class Query9 : public QueryImpl +{ + public: + Query9(rx::Renderer9 *renderer, GLenum type); + virtual ~Query9(); + + void begin(); + void end(); + GLuint getResult(); + GLboolean isResultAvailable(); + + private: + DISALLOW_COPY_AND_ASSIGN(Query9); + + GLboolean testQuery(); + + rx::Renderer9 *mRenderer; + IDirect3DQuery9 *mQuery; +}; + +} + +#endif // LIBGLESV2_RENDERER_QUERY9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/QueryImpl.h b/src/3rdparty/angle/src/libGLESv2/renderer/QueryImpl.h new file mode 100644 index 0000000000..a874047b0c --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/QueryImpl.h @@ -0,0 +1,42 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// QueryImpl.h: Defines the abstract rx::QueryImpl class. + +#ifndef LIBGLESV2_RENDERER_QUERYIMPL_H_ +#define LIBGLESV2_RENDERER_QUERYIMPL_H_ + +#include "common/angleutils.h" + +namespace rx +{ + +class QueryImpl +{ + public: + explicit QueryImpl(GLenum type) : mType(type), mStatus(GL_FALSE), mResult(0) { } + virtual ~QueryImpl() { } + + virtual void begin() = 0; + virtual void end() = 0; + virtual GLuint getResult() = 0; + virtual GLboolean isResultAvailable() = 0; + + GLenum getType() const { return mType; } + + protected: + GLuint mResult; + GLboolean mStatus; + + private: + DISALLOW_COPY_AND_ASSIGN(QueryImpl); + + GLenum mType; +}; + +} + +#endif // LIBGLESV2_RENDERER_QUERYIMPL_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/RenderStateCache.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/RenderStateCache.cpp new file mode 100644 index 0000000000..f60a88f97a --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/RenderStateCache.cpp @@ -0,0 +1,406 @@ +#include "precompiled.h" +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderStateCache.cpp: Defines rx::RenderStateCache, a cache of Direct3D render +// state objects. + +#include "libGLESv2/renderer/RenderStateCache.h" +#include "libGLESv2/renderer/renderer11_utils.h" + +#include "common/debug.h" +#include "third_party/murmurhash/MurmurHash3.h" + +namespace rx +{ + +// MSDN's documentation of ID3D11Device::CreateBlendState, ID3D11Device::CreateRasterizerState, +// ID3D11Device::CreateDepthStencilState and ID3D11Device::CreateSamplerState claims the maximum +// number of unique states of each type an application can create is 4096 +const unsigned int RenderStateCache::kMaxBlendStates = 4096; +const unsigned int RenderStateCache::kMaxRasterizerStates = 4096; +const unsigned int RenderStateCache::kMaxDepthStencilStates = 4096; +const unsigned int RenderStateCache::kMaxSamplerStates = 4096; + +RenderStateCache::RenderStateCache() : mDevice(NULL), mCounter(0), + mBlendStateCache(kMaxBlendStates, hashBlendState, compareBlendStates), + mRasterizerStateCache(kMaxRasterizerStates, hashRasterizerState, compareRasterizerStates), + mDepthStencilStateCache(kMaxDepthStencilStates, hashDepthStencilState, compareDepthStencilStates), + mSamplerStateCache(kMaxSamplerStates, hashSamplerState, compareSamplerStates) +{ +} + +RenderStateCache::~RenderStateCache() +{ + clear(); +} + +void RenderStateCache::initialize(ID3D11Device *device) +{ + clear(); + mDevice = device; +} + +void RenderStateCache::clear() +{ + for (BlendStateMap::iterator i = mBlendStateCache.begin(); i != mBlendStateCache.end(); i++) + { + i->second.first->Release(); + } + mBlendStateCache.clear(); + + for (RasterizerStateMap::iterator i = mRasterizerStateCache.begin(); i != mRasterizerStateCache.end(); i++) + { + i->second.first->Release(); + } + mRasterizerStateCache.clear(); + + for (DepthStencilStateMap::iterator i = mDepthStencilStateCache.begin(); i != mDepthStencilStateCache.end(); i++) + { + i->second.first->Release(); + } + mDepthStencilStateCache.clear(); + + for (SamplerStateMap::iterator i = mSamplerStateCache.begin(); i != mSamplerStateCache.end(); i++) + { + i->second.first->Release(); + } + mSamplerStateCache.clear(); +} + +std::size_t RenderStateCache::hashBlendState(const gl::BlendState &blendState) +{ + static const unsigned int seed = 0xABCDEF98; + + std::size_t hash = 0; + MurmurHash3_x86_32(&blendState, sizeof(gl::BlendState), seed, &hash); + return hash; +} + +bool RenderStateCache::compareBlendStates(const gl::BlendState &a, const gl::BlendState &b) +{ + return memcmp(&a, &b, sizeof(gl::BlendState)) == 0; +} + +ID3D11BlendState *RenderStateCache::getBlendState(const gl::BlendState &blendState) +{ + if (!mDevice) + { + ERR("RenderStateCache is not initialized."); + return NULL; + } + + BlendStateMap::iterator i = mBlendStateCache.find(blendState); + if (i != mBlendStateCache.end()) + { + BlendStateCounterPair &state = i->second; + state.second = mCounter++; + return state.first; + } + else + { + if (mBlendStateCache.size() >= kMaxBlendStates) + { + TRACE("Overflowed the limit of %u blend states, removing the least recently used " + "to make room.", kMaxBlendStates); + + BlendStateMap::iterator leastRecentlyUsed = mBlendStateCache.begin(); + for (BlendStateMap::iterator i = mBlendStateCache.begin(); i != mBlendStateCache.end(); i++) + { + if (i->second.second < leastRecentlyUsed->second.second) + { + leastRecentlyUsed = i; + } + } + leastRecentlyUsed->second.first->Release(); + mBlendStateCache.erase(leastRecentlyUsed); + } + + // Create a new blend state and insert it into the cache + D3D11_BLEND_DESC blendDesc = { 0 }; + blendDesc.AlphaToCoverageEnable = blendState.sampleAlphaToCoverage; + blendDesc.IndependentBlendEnable = FALSE; + + for (unsigned int i = 0; i < D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT; i++) + { + D3D11_RENDER_TARGET_BLEND_DESC &rtBlend = blendDesc.RenderTarget[i]; + + rtBlend.BlendEnable = blendState.blend; + if (blendState.blend) + { + rtBlend.SrcBlend = gl_d3d11::ConvertBlendFunc(blendState.sourceBlendRGB, false); + rtBlend.DestBlend = gl_d3d11::ConvertBlendFunc(blendState.destBlendRGB, false); + rtBlend.BlendOp = gl_d3d11::ConvertBlendOp(blendState.blendEquationRGB); + + rtBlend.SrcBlendAlpha = gl_d3d11::ConvertBlendFunc(blendState.sourceBlendAlpha, true); + rtBlend.DestBlendAlpha = gl_d3d11::ConvertBlendFunc(blendState.destBlendAlpha, true); + rtBlend.BlendOpAlpha = gl_d3d11::ConvertBlendOp(blendState.blendEquationAlpha); + } + + rtBlend.RenderTargetWriteMask = gl_d3d11::ConvertColorMask(blendState.colorMaskRed, + blendState.colorMaskGreen, + blendState.colorMaskBlue, + blendState.colorMaskAlpha); + } + + ID3D11BlendState *dx11BlendState = NULL; + HRESULT result = mDevice->CreateBlendState(&blendDesc, &dx11BlendState); + if (FAILED(result) || !dx11BlendState) + { + ERR("Unable to create a ID3D11BlendState, HRESULT: 0x%X.", result); + return NULL; + } + + mBlendStateCache.insert(std::make_pair(blendState, std::make_pair(dx11BlendState, mCounter++))); + + return dx11BlendState; + } +} + +std::size_t RenderStateCache::hashRasterizerState(const RasterizerStateKey &rasterState) +{ + static const unsigned int seed = 0xABCDEF98; + + std::size_t hash = 0; + MurmurHash3_x86_32(&rasterState, sizeof(RasterizerStateKey), seed, &hash); + return hash; +} + +bool RenderStateCache::compareRasterizerStates(const RasterizerStateKey &a, const RasterizerStateKey &b) +{ + return memcmp(&a, &b, sizeof(RasterizerStateKey)) == 0; +} + +ID3D11RasterizerState *RenderStateCache::getRasterizerState(const gl::RasterizerState &rasterState, + bool scissorEnabled, unsigned int depthSize) +{ + if (!mDevice) + { + ERR("RenderStateCache is not initialized."); + return NULL; + } + + RasterizerStateKey key; + key.rasterizerState = rasterState; + key.scissorEnabled = scissorEnabled; + key.depthSize = depthSize; + + RasterizerStateMap::iterator i = mRasterizerStateCache.find(key); + if (i != mRasterizerStateCache.end()) + { + RasterizerStateCounterPair &state = i->second; + state.second = mCounter++; + return state.first; + } + else + { + if (mRasterizerStateCache.size() >= kMaxRasterizerStates) + { + TRACE("Overflowed the limit of %u rasterizer states, removing the least recently used " + "to make room.", kMaxRasterizerStates); + + RasterizerStateMap::iterator leastRecentlyUsed = mRasterizerStateCache.begin(); + for (RasterizerStateMap::iterator i = mRasterizerStateCache.begin(); i != mRasterizerStateCache.end(); i++) + { + if (i->second.second < leastRecentlyUsed->second.second) + { + leastRecentlyUsed = i; + } + } + leastRecentlyUsed->second.first->Release(); + mRasterizerStateCache.erase(leastRecentlyUsed); + } + + D3D11_CULL_MODE cullMode = gl_d3d11::ConvertCullMode(rasterState.cullFace, rasterState.cullMode); + + // Disable culling if drawing points + if (rasterState.pointDrawMode) + { + cullMode = D3D11_CULL_NONE; + } + + D3D11_RASTERIZER_DESC rasterDesc; + rasterDesc.FillMode = D3D11_FILL_SOLID; + rasterDesc.CullMode = cullMode; + rasterDesc.FrontCounterClockwise = (rasterState.frontFace == GL_CCW) ? FALSE: TRUE; + rasterDesc.DepthBias = ldexp(rasterState.polygonOffsetUnits, -static_cast<int>(depthSize)); + rasterDesc.DepthBiasClamp = 0.0f; // MSDN documentation of DepthBiasClamp implies a value of zero will preform no clamping, must be tested though. + rasterDesc.SlopeScaledDepthBias = rasterState.polygonOffsetFactor; + rasterDesc.DepthClipEnable = TRUE; + rasterDesc.ScissorEnable = scissorEnabled ? TRUE : FALSE; + rasterDesc.MultisampleEnable = TRUE; + rasterDesc.AntialiasedLineEnable = FALSE; + + ID3D11RasterizerState *dx11RasterizerState = NULL; + HRESULT result = mDevice->CreateRasterizerState(&rasterDesc, &dx11RasterizerState); + if (FAILED(result) || !dx11RasterizerState) + { + ERR("Unable to create a ID3D11RasterizerState, HRESULT: 0x%X.", result); + return NULL; + } + + mRasterizerStateCache.insert(std::make_pair(key, std::make_pair(dx11RasterizerState, mCounter++))); + + return dx11RasterizerState; + } +} + +std::size_t RenderStateCache::hashDepthStencilState(const gl::DepthStencilState &dsState) +{ + static const unsigned int seed = 0xABCDEF98; + + std::size_t hash = 0; + MurmurHash3_x86_32(&dsState, sizeof(gl::DepthStencilState), seed, &hash); + return hash; +} + +bool RenderStateCache::compareDepthStencilStates(const gl::DepthStencilState &a, const gl::DepthStencilState &b) +{ + return memcmp(&a, &b, sizeof(gl::DepthStencilState)) == 0; +} + +ID3D11DepthStencilState *RenderStateCache::getDepthStencilState(const gl::DepthStencilState &dsState) +{ + if (!mDevice) + { + ERR("RenderStateCache is not initialized."); + return NULL; + } + + DepthStencilStateMap::iterator i = mDepthStencilStateCache.find(dsState); + if (i != mDepthStencilStateCache.end()) + { + DepthStencilStateCounterPair &state = i->second; + state.second = mCounter++; + return state.first; + } + else + { + if (mDepthStencilStateCache.size() >= kMaxDepthStencilStates) + { + TRACE("Overflowed the limit of %u depth stencil states, removing the least recently used " + "to make room.", kMaxDepthStencilStates); + + DepthStencilStateMap::iterator leastRecentlyUsed = mDepthStencilStateCache.begin(); + for (DepthStencilStateMap::iterator i = mDepthStencilStateCache.begin(); i != mDepthStencilStateCache.end(); i++) + { + if (i->second.second < leastRecentlyUsed->second.second) + { + leastRecentlyUsed = i; + } + } + leastRecentlyUsed->second.first->Release(); + mDepthStencilStateCache.erase(leastRecentlyUsed); + } + + D3D11_DEPTH_STENCIL_DESC dsDesc = { 0 }; + dsDesc.DepthEnable = dsState.depthTest ? TRUE : FALSE; + dsDesc.DepthWriteMask = gl_d3d11::ConvertDepthMask(dsState.depthMask); + dsDesc.DepthFunc = gl_d3d11::ConvertComparison(dsState.depthFunc); + dsDesc.StencilEnable = dsState.stencilTest ? TRUE : FALSE; + dsDesc.StencilReadMask = gl_d3d11::ConvertStencilMask(dsState.stencilMask); + dsDesc.StencilWriteMask = gl_d3d11::ConvertStencilMask(dsState.stencilWritemask); + dsDesc.FrontFace.StencilFailOp = gl_d3d11::ConvertStencilOp(dsState.stencilFail); + dsDesc.FrontFace.StencilDepthFailOp = gl_d3d11::ConvertStencilOp(dsState.stencilPassDepthFail); + dsDesc.FrontFace.StencilPassOp = gl_d3d11::ConvertStencilOp(dsState.stencilPassDepthPass); + dsDesc.FrontFace.StencilFunc = gl_d3d11::ConvertComparison(dsState.stencilFunc); + dsDesc.BackFace.StencilFailOp = gl_d3d11::ConvertStencilOp(dsState.stencilBackFail); + dsDesc.BackFace.StencilDepthFailOp = gl_d3d11::ConvertStencilOp(dsState.stencilBackPassDepthFail); + dsDesc.BackFace.StencilPassOp = gl_d3d11::ConvertStencilOp(dsState.stencilBackPassDepthPass); + dsDesc.BackFace.StencilFunc = gl_d3d11::ConvertComparison(dsState.stencilBackFunc); + + ID3D11DepthStencilState *dx11DepthStencilState = NULL; + HRESULT result = mDevice->CreateDepthStencilState(&dsDesc, &dx11DepthStencilState); + if (FAILED(result) || !dx11DepthStencilState) + { + ERR("Unable to create a ID3D11DepthStencilState, HRESULT: 0x%X.", result); + return NULL; + } + + mDepthStencilStateCache.insert(std::make_pair(dsState, std::make_pair(dx11DepthStencilState, mCounter++))); + + return dx11DepthStencilState; + } +} + +std::size_t RenderStateCache::hashSamplerState(const gl::SamplerState &samplerState) +{ + static const unsigned int seed = 0xABCDEF98; + + std::size_t hash = 0; + MurmurHash3_x86_32(&samplerState, sizeof(gl::SamplerState), seed, &hash); + return hash; +} + +bool RenderStateCache::compareSamplerStates(const gl::SamplerState &a, const gl::SamplerState &b) +{ + return memcmp(&a, &b, sizeof(gl::SamplerState)) == 0; +} + +ID3D11SamplerState *RenderStateCache::getSamplerState(const gl::SamplerState &samplerState) +{ + if (!mDevice) + { + ERR("RenderStateCache is not initialized."); + return NULL; + } + + SamplerStateMap::iterator i = mSamplerStateCache.find(samplerState); + if (i != mSamplerStateCache.end()) + { + SamplerStateCounterPair &state = i->second; + state.second = mCounter++; + return state.first; + } + else + { + if (mSamplerStateCache.size() >= kMaxSamplerStates) + { + TRACE("Overflowed the limit of %u sampler states, removing the least recently used " + "to make room.", kMaxSamplerStates); + + SamplerStateMap::iterator leastRecentlyUsed = mSamplerStateCache.begin(); + for (SamplerStateMap::iterator i = mSamplerStateCache.begin(); i != mSamplerStateCache.end(); i++) + { + if (i->second.second < leastRecentlyUsed->second.second) + { + leastRecentlyUsed = i; + } + } + leastRecentlyUsed->second.first->Release(); + mSamplerStateCache.erase(leastRecentlyUsed); + } + + D3D11_SAMPLER_DESC samplerDesc; + samplerDesc.Filter = gl_d3d11::ConvertFilter(samplerState.minFilter, samplerState.magFilter, samplerState.maxAnisotropy); + samplerDesc.AddressU = gl_d3d11::ConvertTextureWrap(samplerState.wrapS); + samplerDesc.AddressV = gl_d3d11::ConvertTextureWrap(samplerState.wrapT); + samplerDesc.AddressW = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.MipLODBias = static_cast<float>(samplerState.lodOffset); + samplerDesc.MaxAnisotropy = samplerState.maxAnisotropy; + samplerDesc.ComparisonFunc = D3D11_COMPARISON_NEVER; + samplerDesc.BorderColor[0] = 0.0f; + samplerDesc.BorderColor[1] = 0.0f; + samplerDesc.BorderColor[2] = 0.0f; + samplerDesc.BorderColor[3] = 0.0f; + samplerDesc.MinLOD = gl_d3d11::ConvertMinLOD(samplerState.minFilter, samplerState.lodOffset); + samplerDesc.MaxLOD = gl_d3d11::ConvertMaxLOD(samplerState.minFilter, samplerState.lodOffset); + + ID3D11SamplerState *dx11SamplerState = NULL; + HRESULT result = mDevice->CreateSamplerState(&samplerDesc, &dx11SamplerState); + if (FAILED(result) || !dx11SamplerState) + { + ERR("Unable to create a ID3D11DepthStencilState, HRESULT: 0x%X.", result); + return NULL; + } + + mSamplerStateCache.insert(std::make_pair(samplerState, std::make_pair(dx11SamplerState, mCounter++))); + + return dx11SamplerState; + } +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/RenderStateCache.h b/src/3rdparty/angle/src/libGLESv2/renderer/RenderStateCache.h new file mode 100644 index 0000000000..f8b5111de4 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/RenderStateCache.h @@ -0,0 +1,101 @@ +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderStateCache.h: Defines rx::RenderStateCache, a cache of Direct3D render +// state objects. + +#ifndef LIBGLESV2_RENDERER_RENDERSTATECACHE_H_ +#define LIBGLESV2_RENDERER_RENDERSTATECACHE_H_ + +#include "libGLESv2/angletypes.h" +#include "common/angleutils.h" + +namespace rx +{ + +class RenderStateCache +{ + public: + RenderStateCache(); + virtual ~RenderStateCache(); + + void initialize(ID3D11Device *device); + void clear(); + + // Increments refcount on the returned blend state, Release() must be called. + ID3D11BlendState *getBlendState(const gl::BlendState &blendState); + ID3D11RasterizerState *getRasterizerState(const gl::RasterizerState &rasterState, + bool scissorEnabled, unsigned int depthSize); + ID3D11DepthStencilState *getDepthStencilState(const gl::DepthStencilState &dsState); + ID3D11SamplerState *getSamplerState(const gl::SamplerState &samplerState); + + private: + DISALLOW_COPY_AND_ASSIGN(RenderStateCache); + + unsigned long long mCounter; + + // Blend state cache + static std::size_t hashBlendState(const gl::BlendState &blendState); + static bool compareBlendStates(const gl::BlendState &a, const gl::BlendState &b); + static const unsigned int kMaxBlendStates; + + typedef std::size_t (*BlendStateHashFunction)(const gl::BlendState &); + typedef bool (*BlendStateEqualityFunction)(const gl::BlendState &, const gl::BlendState &); + typedef std::pair<ID3D11BlendState*, unsigned long long> BlendStateCounterPair; + typedef std::unordered_map<gl::BlendState, BlendStateCounterPair, BlendStateHashFunction, BlendStateEqualityFunction> BlendStateMap; + BlendStateMap mBlendStateCache; + + // Rasterizer state cache + struct RasterizerStateKey + { + gl::RasterizerState rasterizerState; + bool scissorEnabled; + unsigned int depthSize; + }; + static std::size_t hashRasterizerState(const RasterizerStateKey &rasterState); + static bool compareRasterizerStates(const RasterizerStateKey &a, const RasterizerStateKey &b); + static const unsigned int kMaxRasterizerStates; + + typedef std::size_t (*RasterizerStateHashFunction)(const RasterizerStateKey &); + typedef bool (*RasterizerStateEqualityFunction)(const RasterizerStateKey &, const RasterizerStateKey &); + typedef std::pair<ID3D11RasterizerState*, unsigned long long> RasterizerStateCounterPair; + typedef std::unordered_map<RasterizerStateKey, RasterizerStateCounterPair, RasterizerStateHashFunction, RasterizerStateEqualityFunction> RasterizerStateMap; + RasterizerStateMap mRasterizerStateCache; + + // Depth stencil state cache + static std::size_t hashDepthStencilState(const gl::DepthStencilState &dsState); + static bool compareDepthStencilStates(const gl::DepthStencilState &a, const gl::DepthStencilState &b); + static const unsigned int kMaxDepthStencilStates; + + typedef std::size_t (*DepthStencilStateHashFunction)(const gl::DepthStencilState &); + typedef bool (*DepthStencilStateEqualityFunction)(const gl::DepthStencilState &, const gl::DepthStencilState &); + typedef std::pair<ID3D11DepthStencilState*, unsigned long long> DepthStencilStateCounterPair; + typedef std::unordered_map<gl::DepthStencilState, + DepthStencilStateCounterPair, + DepthStencilStateHashFunction, + DepthStencilStateEqualityFunction> DepthStencilStateMap; + DepthStencilStateMap mDepthStencilStateCache; + + // Sample state cache + static std::size_t hashSamplerState(const gl::SamplerState &samplerState); + static bool compareSamplerStates(const gl::SamplerState &a, const gl::SamplerState &b); + static const unsigned int kMaxSamplerStates; + + typedef std::size_t (*SamplerStateHashFunction)(const gl::SamplerState &); + typedef bool (*SamplerStateEqualityFunction)(const gl::SamplerState &, const gl::SamplerState &); + typedef std::pair<ID3D11SamplerState*, unsigned long long> SamplerStateCounterPair; + typedef std::unordered_map<gl::SamplerState, + SamplerStateCounterPair, + SamplerStateHashFunction, + SamplerStateEqualityFunction> SamplerStateMap; + SamplerStateMap mSamplerStateCache; + + ID3D11Device *mDevice; +}; + +} + +#endif // LIBGLESV2_RENDERER_RENDERSTATECACHE_H_
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget.h b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget.h new file mode 100644 index 0000000000..80de39f4f7 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget.h @@ -0,0 +1,56 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderTarget.h: Defines an abstract wrapper class to manage IDirect3DSurface9 +// and ID3D11View objects belonging to renderbuffers. + +#ifndef LIBGLESV2_RENDERER_RENDERTARGET_H_ +#define LIBGLESV2_RENDERER_RENDERTARGET_H_ + +#include "common/angleutils.h" + +namespace rx +{ +class RenderTarget +{ + public: + RenderTarget() + { + mWidth = 0; + mHeight = 0; + mInternalFormat = GL_NONE; + mActualFormat = GL_NONE; + mSamples = 0; + } + + virtual ~RenderTarget() {}; + + GLsizei getWidth() { return mWidth; } + GLsizei getHeight() { return mHeight; } + GLenum getInternalFormat() { return mInternalFormat; } + GLenum getActualFormat() { return mActualFormat; } + GLsizei getSamples() { return mSamples; } + + struct Desc { + GLsizei width; + GLsizei height; + GLenum format; + }; + + protected: + GLsizei mWidth; + GLsizei mHeight; + GLenum mInternalFormat; + GLenum mActualFormat; + GLsizei mSamples; + + private: + DISALLOW_COPY_AND_ASSIGN(RenderTarget); +}; + +} + +#endif // LIBGLESV2_RENDERTARGET_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget11.cpp new file mode 100644 index 0000000000..cf226de17b --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget11.cpp @@ -0,0 +1,378 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderTarget11.cpp: Implements a DX11-specific wrapper for ID3D11View pointers +// retained by Renderbuffers. + +#include "libGLESv2/renderer/RenderTarget11.h" +#include "libGLESv2/renderer/Renderer11.h" + +#include "libGLESv2/renderer/renderer11_utils.h" +#include "libGLESv2/main.h" + +namespace rx +{ + +static unsigned int getRTVSubresourceIndex(ID3D11Texture2D *texture, ID3D11RenderTargetView *view) +{ + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + view->GetDesc(&rtvDesc); + + D3D11_TEXTURE2D_DESC texDesc; + texture->GetDesc(&texDesc); + + unsigned int mipSlice = 0; + unsigned int arraySlice = 0; + unsigned int mipLevels = texDesc.MipLevels; + + switch (rtvDesc.ViewDimension) + { + case D3D11_RTV_DIMENSION_TEXTURE1D: + mipSlice = rtvDesc.Texture1D.MipSlice; + arraySlice = 0; + break; + + case D3D11_RTV_DIMENSION_TEXTURE1DARRAY: + mipSlice = rtvDesc.Texture1DArray.MipSlice; + arraySlice = rtvDesc.Texture1DArray.FirstArraySlice; + break; + + case D3D11_RTV_DIMENSION_TEXTURE2D: + mipSlice = rtvDesc.Texture2D.MipSlice; + arraySlice = 0; + break; + + case D3D11_RTV_DIMENSION_TEXTURE2DARRAY: + mipSlice = rtvDesc.Texture2DArray.MipSlice; + arraySlice = rtvDesc.Texture2DArray.FirstArraySlice; + break; + + case D3D11_RTV_DIMENSION_TEXTURE2DMS: + mipSlice = 0; + arraySlice = 0; + break; + + case D3D11_RTV_DIMENSION_TEXTURE2DMSARRAY: + mipSlice = 0; + arraySlice = rtvDesc.Texture2DMSArray.FirstArraySlice; + break; + + case D3D11_RTV_DIMENSION_TEXTURE3D: + mipSlice = rtvDesc.Texture3D.MipSlice; + arraySlice = 0; + break; + + case D3D11_RTV_DIMENSION_UNKNOWN: + case D3D11_RTV_DIMENSION_BUFFER: + UNIMPLEMENTED(); + break; + + default: + UNREACHABLE(); + break; + } + + return D3D11CalcSubresource(mipSlice, arraySlice, mipLevels); +} + +static unsigned int getDSVSubresourceIndex(ID3D11Texture2D *texture, ID3D11DepthStencilView *view) +{ + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; + view->GetDesc(&dsvDesc); + + D3D11_TEXTURE2D_DESC texDesc; + texture->GetDesc(&texDesc); + + unsigned int mipSlice = 0; + unsigned int arraySlice = 0; + unsigned int mipLevels = texDesc.MipLevels; + + switch (dsvDesc.ViewDimension) + { + case D3D11_DSV_DIMENSION_TEXTURE1D: + mipSlice = dsvDesc.Texture1D.MipSlice; + arraySlice = 0; + break; + + case D3D11_DSV_DIMENSION_TEXTURE1DARRAY: + mipSlice = dsvDesc.Texture1DArray.MipSlice; + arraySlice = dsvDesc.Texture1DArray.FirstArraySlice; + break; + + case D3D11_DSV_DIMENSION_TEXTURE2D: + mipSlice = dsvDesc.Texture2D.MipSlice; + arraySlice = 0; + break; + + case D3D11_DSV_DIMENSION_TEXTURE2DARRAY: + mipSlice = dsvDesc.Texture2DArray.MipSlice; + arraySlice = dsvDesc.Texture2DArray.FirstArraySlice; + break; + + case D3D11_DSV_DIMENSION_TEXTURE2DMS: + mipSlice = 0; + arraySlice = 0; + break; + + case D3D11_DSV_DIMENSION_TEXTURE2DMSARRAY: + mipSlice = 0; + arraySlice = dsvDesc.Texture2DMSArray.FirstArraySlice; + break; + + case D3D11_RTV_DIMENSION_UNKNOWN: + UNIMPLEMENTED(); + break; + + default: + UNREACHABLE(); + break; + } + + return D3D11CalcSubresource(mipSlice, arraySlice, mipLevels); +} + +RenderTarget11::RenderTarget11(Renderer *renderer, ID3D11RenderTargetView *rtv, ID3D11Texture2D *tex, ID3D11ShaderResourceView *srv, GLsizei width, GLsizei height) +{ + mRenderer = Renderer11::makeRenderer11(renderer); + mTexture = tex; + mRenderTarget = rtv; + mDepthStencil = NULL; + mShaderResource = srv; + mSubresourceIndex = 0; + + if (mRenderTarget && mTexture) + { + D3D11_RENDER_TARGET_VIEW_DESC desc; + mRenderTarget->GetDesc(&desc); + + D3D11_TEXTURE2D_DESC texDesc; + mTexture->GetDesc(&texDesc); + + mSubresourceIndex = getRTVSubresourceIndex(mTexture, mRenderTarget); + mWidth = width; + mHeight = height; + mSamples = (texDesc.SampleDesc.Count > 1) ? texDesc.SampleDesc.Count : 0; + + mInternalFormat = d3d11_gl::ConvertTextureInternalFormat(desc.Format); + mActualFormat = d3d11_gl::ConvertTextureInternalFormat(desc.Format); + } +} + +RenderTarget11::RenderTarget11(Renderer *renderer, ID3D11DepthStencilView *dsv, ID3D11Texture2D *tex, ID3D11ShaderResourceView *srv, GLsizei width, GLsizei height) +{ + mRenderer = Renderer11::makeRenderer11(renderer); + mTexture = tex; + mRenderTarget = NULL; + mDepthStencil = dsv; + mShaderResource = srv; + mSubresourceIndex = 0; + + if (mDepthStencil && mTexture) + { + D3D11_DEPTH_STENCIL_VIEW_DESC desc; + mDepthStencil->GetDesc(&desc); + + D3D11_TEXTURE2D_DESC texDesc; + mTexture->GetDesc(&texDesc); + + mSubresourceIndex = getDSVSubresourceIndex(mTexture, mDepthStencil); + mWidth = width; + mHeight = height; + mSamples = (texDesc.SampleDesc.Count > 1) ? texDesc.SampleDesc.Count : 0; + + mInternalFormat = d3d11_gl::ConvertTextureInternalFormat(desc.Format); + mActualFormat = d3d11_gl::ConvertTextureInternalFormat(desc.Format); + } +} + +RenderTarget11::RenderTarget11(Renderer *renderer, GLsizei width, GLsizei height, GLenum format, GLsizei samples, bool depth) +{ + mRenderer = Renderer11::makeRenderer11(renderer); + mTexture = NULL; + mRenderTarget = NULL; + mDepthStencil = NULL; + mShaderResource = NULL; + + DXGI_FORMAT requestedFormat = gl_d3d11::ConvertRenderbufferFormat(format); + + int supportedSamples = mRenderer->getNearestSupportedSamples(requestedFormat, samples); + if (supportedSamples < 0) + { + gl::error(GL_OUT_OF_MEMORY); + return; + } + + if (width > 0 && height > 0) + { + // Create texture resource + D3D11_TEXTURE2D_DESC desc; + desc.Width = width; + desc.Height = height; + desc.MipLevels = 1; + desc.ArraySize = 1; + desc.Format = requestedFormat; + desc.SampleDesc.Count = (supportedSamples == 0) ? 1 : supportedSamples; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + desc.BindFlags = (depth ? D3D11_BIND_DEPTH_STENCIL : (D3D11_BIND_RENDER_TARGET | D3D11_BIND_SHADER_RESOURCE)); + + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result = device->CreateTexture2D(&desc, NULL, &mTexture); + + if (result == E_OUTOFMEMORY) + { + gl::error(GL_OUT_OF_MEMORY); + return; + } + ASSERT(SUCCEEDED(result)); + + if (depth) + { + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; + dsvDesc.Format = requestedFormat; + dsvDesc.ViewDimension = (supportedSamples == 0) ? D3D11_DSV_DIMENSION_TEXTURE2D : D3D11_DSV_DIMENSION_TEXTURE2DMS; + dsvDesc.Texture2D.MipSlice = 0; + dsvDesc.Flags = 0; + result = device->CreateDepthStencilView(mTexture, &dsvDesc, &mDepthStencil); + + if (result == E_OUTOFMEMORY) + { + mTexture->Release(); + mTexture = NULL; + gl::error(GL_OUT_OF_MEMORY); + } + ASSERT(SUCCEEDED(result)); + } + else + { + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = requestedFormat; + rtvDesc.ViewDimension = (supportedSamples == 0) ? D3D11_RTV_DIMENSION_TEXTURE2D : D3D11_RTV_DIMENSION_TEXTURE2DMS; + rtvDesc.Texture2D.MipSlice = 0; + result = device->CreateRenderTargetView(mTexture, &rtvDesc, &mRenderTarget); + + if (result == E_OUTOFMEMORY) + { + mTexture->Release(); + mTexture = NULL; + gl::error(GL_OUT_OF_MEMORY); + return; + } + ASSERT(SUCCEEDED(result)); + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = requestedFormat; + srvDesc.ViewDimension = (supportedSamples == 0) ? D3D11_SRV_DIMENSION_TEXTURE2D : D3D11_SRV_DIMENSION_TEXTURE2DMS; + srvDesc.Texture2D.MostDetailedMip = 0; + srvDesc.Texture2D.MipLevels = 1; + result = device->CreateShaderResourceView(mTexture, &srvDesc, &mShaderResource); + + if (result == E_OUTOFMEMORY) + { + mTexture->Release(); + mTexture = NULL; + mRenderTarget->Release(); + mRenderTarget = NULL; + gl::error(GL_OUT_OF_MEMORY); + return; + } + ASSERT(SUCCEEDED(result)); + } + } + + mWidth = width; + mHeight = height; + mInternalFormat = format; + mSamples = supportedSamples; + mActualFormat = d3d11_gl::ConvertTextureInternalFormat(requestedFormat); + mSubresourceIndex = D3D11CalcSubresource(0, 0, 1); +} + +RenderTarget11::~RenderTarget11() +{ + if (mTexture) + { + mTexture->Release(); + mTexture = NULL; + } + + if (mRenderTarget) + { + mRenderTarget->Release(); + mRenderTarget = NULL; + } + + if (mDepthStencil) + { + mDepthStencil->Release(); + mDepthStencil = NULL; + } + + if (mShaderResource) + { + mShaderResource->Release(); + mShaderResource = NULL; + } +} + +RenderTarget11 *RenderTarget11::makeRenderTarget11(RenderTarget *target) +{ + ASSERT(HAS_DYNAMIC_TYPE(rx::RenderTarget11*, target)); + return static_cast<rx::RenderTarget11*>(target); +} + +ID3D11Texture2D *RenderTarget11::getTexture() const +{ + if (mTexture) + { + mTexture->AddRef(); + } + + return mTexture; +} + +// Adds reference, caller must call Release +ID3D11RenderTargetView *RenderTarget11::getRenderTargetView() const +{ + if (mRenderTarget) + { + mRenderTarget->AddRef(); + } + + return mRenderTarget; +} + +// Adds reference, caller must call Release +ID3D11DepthStencilView *RenderTarget11::getDepthStencilView() const +{ + if (mDepthStencil) + { + mDepthStencil->AddRef(); + } + + return mDepthStencil; +} + +// Adds reference, caller must call Release +ID3D11ShaderResourceView *RenderTarget11::getShaderResourceView() const +{ + if (mShaderResource) + { + mShaderResource->AddRef(); + } + + return mShaderResource; +} + +unsigned int RenderTarget11::getSubresourceIndex() const +{ + return mSubresourceIndex; +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget11.h b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget11.h new file mode 100644 index 0000000000..dc697cf0e3 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget11.h @@ -0,0 +1,58 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderTarget11.h: Defines a DX11-specific wrapper for ID3D11View pointers +// retained by Renderbuffers. + +#ifndef LIBGLESV2_RENDERER_RENDERTARGET11_H_ +#define LIBGLESV2_RENDERER_RENDERTARGET11_H_ + +#include "libGLESv2/renderer/RenderTarget.h" + +namespace rx +{ +class Renderer; +class Renderer11; + +class RenderTarget11 : public RenderTarget +{ + public: + RenderTarget11(Renderer *renderer, ID3D11RenderTargetView *rtv, ID3D11Texture2D *tex, ID3D11ShaderResourceView *srv, GLsizei width, GLsizei height); + RenderTarget11(Renderer *renderer, ID3D11DepthStencilView *dsv, ID3D11Texture2D *tex, ID3D11ShaderResourceView *srv, GLsizei width, GLsizei height); + RenderTarget11(Renderer *renderer, GLsizei width, GLsizei height, GLenum format, GLsizei samples, bool depth); + virtual ~RenderTarget11(); + + static RenderTarget11 *makeRenderTarget11(RenderTarget *renderTarget); + + // Adds reference, caller must call Release + ID3D11Texture2D *getTexture() const; + + // Adds reference, caller must call Release + ID3D11RenderTargetView *getRenderTargetView() const; + + // Adds reference, caller must call Release + ID3D11DepthStencilView *getDepthStencilView() const; + + // Adds reference, caller must call Release + ID3D11ShaderResourceView *getShaderResourceView() const; + + unsigned int getSubresourceIndex() const; + + private: + DISALLOW_COPY_AND_ASSIGN(RenderTarget11); + + unsigned int mSubresourceIndex; + ID3D11Texture2D *mTexture; + ID3D11RenderTargetView *mRenderTarget; + ID3D11DepthStencilView *mDepthStencil; + ID3D11ShaderResourceView *mShaderResource; + + Renderer11 *mRenderer; +}; + +} + +#endif LIBGLESV2_RENDERER_RENDERTARGET11_H_
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget9.cpp new file mode 100644 index 0000000000..a84c709059 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget9.cpp @@ -0,0 +1,113 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderTarget9.cpp: Implements a D3D9-specific wrapper for IDirect3DSurface9 +// pointers retained by renderbuffers. + +#include "libGLESv2/renderer/RenderTarget9.h" +#include "libGLESv2/renderer/Renderer9.h" + +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/main.h" + +namespace rx +{ + +RenderTarget9::RenderTarget9(Renderer *renderer, IDirect3DSurface9 *surface) +{ + mRenderer = Renderer9::makeRenderer9(renderer); + mRenderTarget = surface; + + if (mRenderTarget) + { + D3DSURFACE_DESC description; + mRenderTarget->GetDesc(&description); + + mWidth = description.Width; + mHeight = description.Height; + + mInternalFormat = d3d9_gl::GetEquivalentFormat(description.Format); + mActualFormat = d3d9_gl::GetEquivalentFormat(description.Format); + mSamples = d3d9_gl::GetSamplesFromMultisampleType(description.MultiSampleType); + } +} + +RenderTarget9::RenderTarget9(Renderer *renderer, GLsizei width, GLsizei height, GLenum format, GLsizei samples) +{ + mRenderer = Renderer9::makeRenderer9(renderer); + mRenderTarget = NULL; + + D3DFORMAT requestedFormat = gl_d3d9::ConvertRenderbufferFormat(format); + int supportedSamples = mRenderer->getNearestSupportedSamples(requestedFormat, samples); + + if (supportedSamples == -1) + { + gl::error(GL_OUT_OF_MEMORY); + + return; + } + + HRESULT result = D3DERR_INVALIDCALL; + + if (width > 0 && height > 0) + { + if (requestedFormat == D3DFMT_D24S8) + { + result = mRenderer->getDevice()->CreateDepthStencilSurface(width, height, requestedFormat, + gl_d3d9::GetMultisampleTypeFromSamples(supportedSamples), + 0, FALSE, &mRenderTarget, NULL); + } + else + { + result = mRenderer->getDevice()->CreateRenderTarget(width, height, requestedFormat, + gl_d3d9::GetMultisampleTypeFromSamples(supportedSamples), + 0, FALSE, &mRenderTarget, NULL); + } + + if (result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY) + { + gl::error(GL_OUT_OF_MEMORY); + + return; + } + + ASSERT(SUCCEEDED(result)); + } + + mWidth = width; + mHeight = height; + mInternalFormat = format; + mSamples = supportedSamples; + mActualFormat = d3d9_gl::GetEquivalentFormat(requestedFormat); +} + +RenderTarget9::~RenderTarget9() +{ + if (mRenderTarget) + { + mRenderTarget->Release(); + } +} + +RenderTarget9 *RenderTarget9::makeRenderTarget9(RenderTarget *target) +{ + ASSERT(HAS_DYNAMIC_TYPE(rx::RenderTarget9*, target)); + return static_cast<rx::RenderTarget9*>(target); +} + +IDirect3DSurface9 *RenderTarget9::getSurface() +{ + // Caller is responsible for releasing the returned surface reference. + if (mRenderTarget) + { + mRenderTarget->AddRef(); + } + + return mRenderTarget; +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget9.h b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget9.h new file mode 100644 index 0000000000..faf8ad1c6d --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/RenderTarget9.h @@ -0,0 +1,40 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderTarget9.h: Defines a D3D9-specific wrapper for IDirect3DSurface9 pointers +// retained by Renderbuffers. + +#ifndef LIBGLESV2_RENDERER_RENDERTARGET9_H_ +#define LIBGLESV2_RENDERER_RENDERTARGET9_H_ + +#include "libGLESv2/renderer/RenderTarget.h" + +namespace rx +{ +class Renderer; +class Renderer9; + +class RenderTarget9 : public RenderTarget +{ + public: + RenderTarget9(Renderer *renderer, IDirect3DSurface9 *surface); + RenderTarget9(Renderer *renderer, GLsizei width, GLsizei height, GLenum format, GLsizei samples); + virtual ~RenderTarget9(); + + static RenderTarget9 *makeRenderTarget9(RenderTarget *renderTarget); + IDirect3DSurface9 *getSurface(); + + private: + DISALLOW_COPY_AND_ASSIGN(RenderTarget9); + + IDirect3DSurface9 *mRenderTarget; + + Renderer9 *mRenderer; +}; + +} + +#endif // LIBGLESV2_RENDERER_RENDERTARGET9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Renderer.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer.cpp new file mode 100644 index 0000000000..41cdb8b278 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer.cpp @@ -0,0 +1,201 @@ +#include "precompiled.h" +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Renderer.cpp: Implements EGL dependencies for creating and destroying Renderer instances. + +#include "libGLESv2/main.h" +#include "libGLESv2/Program.h" +#include "libGLESv2/renderer/Renderer.h" +#if defined(ANGLE_ENABLE_D3D11) +# include "libGLESv2/renderer/Renderer11.h" +# define D3DERR_OUTOFVIDEOMEMORY MAKE_HRESULT( 1, 0x876, 380 ) +#else +# include "libGLESv2/renderer/Renderer9.h" +#endif +#include "libGLESv2/utilities.h" + +#if !defined(ANGLE_COMPILE_OPTIMIZATION_LEVEL) +#define ANGLE_COMPILE_OPTIMIZATION_LEVEL D3DCOMPILE_OPTIMIZATION_LEVEL3 +#endif + +namespace rx +{ + +Renderer::Renderer(egl::Display *display) : mDisplay(display) +{ + mD3dCompilerModule = NULL; + mD3DCompileFunc = NULL; +} + +Renderer::~Renderer() +{ + if (mD3dCompilerModule) + { + FreeLibrary(mD3dCompilerModule); + mD3dCompilerModule = NULL; + } +} + +bool Renderer::initializeCompiler() +{ +#if defined(ANGLE_PRELOADED_D3DCOMPILER_MODULE_NAMES) + // Find a D3DCompiler module that had already been loaded based on a predefined list of versions. + static TCHAR* d3dCompilerNames[] = ANGLE_PRELOADED_D3DCOMPILER_MODULE_NAMES; + + for (size_t i = 0; i < ArraySize(d3dCompilerNames); ++i) + { + if (GetModuleHandleEx(0, d3dCompilerNames[i], &mD3dCompilerModule)) + { + break; + } + } +#else + // Load the version of the D3DCompiler DLL associated with the Direct3D version ANGLE was built with. + mD3dCompilerModule = LoadLibrary(D3DCOMPILER_DLL); +#endif // ANGLE_PRELOADED_D3DCOMPILER_MODULE_NAMES + + if (!mD3dCompilerModule) + { + ERR("No D3D compiler module found - aborting!\n"); + return false; + } + + mD3DCompileFunc = reinterpret_cast<pCompileFunc>(GetProcAddress(mD3dCompilerModule, "D3DCompile")); + ASSERT(mD3DCompileFunc); + + return mD3DCompileFunc != NULL; +} + +// Compiles HLSL code into executable binaries +ShaderBlob *Renderer::compileToBinary(gl::InfoLog &infoLog, const char *hlsl, const char *profile, UINT optimizationFlags, bool alternateFlags) +{ + if (!hlsl) + { + return NULL; + } + + HRESULT result = S_OK; + UINT flags = 0; + std::string sourceText; + if (gl::perfActive()) + { + flags |= D3DCOMPILE_DEBUG; + +#ifdef NDEBUG + flags |= optimizationFlags; +#else + flags |= D3DCOMPILE_SKIP_OPTIMIZATION; +#endif + + std::string sourcePath = getTempPath(); + sourceText = std::string("#line 2 \"") + sourcePath + std::string("\"\n\n") + std::string(hlsl); + writeFile(sourcePath.c_str(), sourceText.c_str(), sourceText.size()); + } + else + { + flags |= optimizationFlags; + sourceText = hlsl; + } + + // Sometimes D3DCompile will fail with the default compilation flags for complicated shaders when it would otherwise pass with alternative options. + // Try the default flags first and if compilation fails, try some alternatives. + const static UINT extraFlags[] = + { + 0, + D3DCOMPILE_AVOID_FLOW_CONTROL, + D3DCOMPILE_PREFER_FLOW_CONTROL + }; + + const static char * const extraFlagNames[] = + { + "default", + "avoid flow control", + "prefer flow control" + }; + + int attempts = alternateFlags ? ArraySize(extraFlags) : 1; + pD3DCompile compileFunc = reinterpret_cast<pD3DCompile>(mD3DCompileFunc); + for (int i = 0; i < attempts; ++i) + { + ID3DBlob *errorMessage = NULL; + ID3DBlob *binary = NULL; + + result = compileFunc(hlsl, strlen(hlsl), gl::g_fakepath, NULL, NULL, + "main", profile, flags | extraFlags[i], 0, &binary, &errorMessage); + if (errorMessage) + { + const char *message = (const char*)errorMessage->GetBufferPointer(); + + infoLog.appendSanitized(message); + TRACE("\n%s", hlsl); + TRACE("\n%s", message); + + errorMessage->Release(); + errorMessage = NULL; + } + + if (SUCCEEDED(result)) + { + return (ShaderBlob*)binary; + } + else + { + if (result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY) + { + return gl::error(GL_OUT_OF_MEMORY, (ShaderBlob*) NULL); + } + + infoLog.append("Warning: D3D shader compilation failed with "); + infoLog.append(extraFlagNames[i]); + infoLog.append(" flags."); + if (i + 1 < attempts) + { + infoLog.append(" Retrying with "); + infoLog.append(extraFlagNames[i + 1]); + infoLog.append(".\n"); + } + } + } + + return NULL; +} + +} + +extern "C" +{ + +rx::Renderer *glCreateRenderer(egl::Display *display, HDC hDc, bool softwareDevice) +{ + rx::Renderer *renderer = NULL; + EGLint status = EGL_BAD_ALLOC; + +#if defined(ANGLE_ENABLE_D3D11) + renderer = new rx::Renderer11(display, hDc); +#else + renderer = new rx::Renderer9(display, hDc, softwareDevice); +#endif + + if (renderer) + { + status = renderer->initialize(); + } + + if (status == EGL_SUCCESS) + { + return renderer; + } + + return NULL; +} + +void glDestroyRenderer(rx::Renderer *renderer) +{ + delete renderer; +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Renderer.h b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer.h new file mode 100644 index 0000000000..656cb0f1bf --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer.h @@ -0,0 +1,238 @@ +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Renderer.h: Defines a back-end specific class that hides the details of the +// implementation-specific renderer. + +#ifndef LIBGLESV2_RENDERER_RENDERER_H_ +#define LIBGLESV2_RENDERER_RENDERER_H_ + +#include "libGLESv2/Uniform.h" +#include "libGLESv2/angletypes.h" + +const int versionWindowsVista = MAKEWORD(0x00, 0x06); +const int versionWindows7 = MAKEWORD(0x01, 0x06); + +// Return the version of the operating system in a format suitable for ordering +// comparison. +inline int getComparableOSVersion() +{ + DWORD version = GetVersion(); + int majorVersion = LOBYTE(LOWORD(version)); + int minorVersion = HIBYTE(LOWORD(version)); + return MAKEWORD(minorVersion, majorVersion); +} + +namespace egl +{ +class Display; +} + +namespace gl +{ +class InfoLog; +class ProgramBinary; +class VertexAttribute; +class Buffer; +class Texture; +class Framebuffer; +} + +namespace rx +{ +class TextureStorageInterface2D; +class TextureStorageInterfaceCube; +class VertexBuffer; +class IndexBuffer; +class QueryImpl; +class FenceImpl; +class BufferStorage; +class Blit; +struct TranslatedIndexData; +class ShaderExecutable; +class SwapChain; +class RenderTarget; +class Image; +class TextureStorage; + +typedef void * ShaderBlob; +typedef void (*pCompileFunc)(); + +struct ConfigDesc +{ + GLenum renderTargetFormat; + GLenum depthStencilFormat; + GLint multiSample; + bool fastConfig; +}; + +struct dx_VertexConstants +{ + float depthRange[4]; + float viewAdjust[4]; +}; + +struct dx_PixelConstants +{ + float depthRange[4]; + float viewCoords[4]; + float depthFront[4]; +}; + +enum ShaderType +{ + SHADER_VERTEX, + SHADER_PIXEL, + SHADER_GEOMETRY +}; + +class Renderer +{ + public: + explicit Renderer(egl::Display *display); + virtual ~Renderer(); + + virtual EGLint initialize() = 0; + virtual bool resetDevice() = 0; + + virtual int generateConfigs(ConfigDesc **configDescList) = 0; + virtual void deleteConfigs(ConfigDesc *configDescList) = 0; + + virtual void sync(bool block) = 0; + + virtual SwapChain *createSwapChain(HWND window, HANDLE shareHandle, GLenum backBufferFormat, GLenum depthBufferFormat) = 0; + + virtual void setSamplerState(gl::SamplerType type, int index, const gl::SamplerState &sampler) = 0; + virtual void setTexture(gl::SamplerType type, int index, gl::Texture *texture) = 0; + + virtual void setRasterizerState(const gl::RasterizerState &rasterState) = 0; + virtual void setBlendState(const gl::BlendState &blendState, const gl::Color &blendColor, + unsigned int sampleMask) = 0; + virtual void setDepthStencilState(const gl::DepthStencilState &depthStencilState, int stencilRef, + int stencilBackRef, bool frontFaceCCW) = 0; + + virtual void setScissorRectangle(const gl::Rectangle &scissor, bool enabled) = 0; + virtual bool setViewport(const gl::Rectangle &viewport, float zNear, float zFar, GLenum drawMode, GLenum frontFace, + bool ignoreViewport) = 0; + + virtual bool applyRenderTarget(gl::Framebuffer *frameBuffer) = 0; + virtual void applyShaders(gl::ProgramBinary *programBinary) = 0; + virtual void applyUniforms(gl::ProgramBinary *programBinary, gl::UniformArray *uniformArray) = 0; + virtual bool applyPrimitiveType(GLenum primitiveType, GLsizei elementCount) = 0; + virtual GLenum applyVertexBuffer(gl::ProgramBinary *programBinary, gl::VertexAttribute vertexAttributes[], GLint first, GLsizei count, GLsizei instances) = 0; + virtual GLenum applyIndexBuffer(const GLvoid *indices, gl::Buffer *elementArrayBuffer, GLsizei count, GLenum mode, GLenum type, TranslatedIndexData *indexInfo) = 0; + + virtual void drawArrays(GLenum mode, GLsizei count, GLsizei instances) = 0; + virtual void drawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices, gl::Buffer *elementArrayBuffer, const TranslatedIndexData &indexInfo, GLsizei instances) = 0; + + virtual void clear(const gl::ClearParameters &clearParams, gl::Framebuffer *frameBuffer) = 0; + + virtual void markAllStateDirty() = 0; + + // lost device + virtual void notifyDeviceLost() = 0; + virtual bool isDeviceLost() = 0; + virtual bool testDeviceLost(bool notify) = 0; + virtual bool testDeviceResettable() = 0; + + // Renderer capabilities + virtual DWORD getAdapterVendor() const = 0; + virtual std::string getRendererDescription() const = 0; + virtual GUID getAdapterIdentifier() const = 0; + + virtual bool getBGRATextureSupport() const = 0; + virtual bool getDXT1TextureSupport() = 0; + virtual bool getDXT3TextureSupport() = 0; + virtual bool getDXT5TextureSupport() = 0; + virtual bool getEventQuerySupport() = 0; + virtual bool getFloat32TextureSupport(bool *filtering, bool *renderable) = 0; + virtual bool getFloat16TextureSupport(bool *filtering, bool *renderable) = 0; + virtual bool getLuminanceTextureSupport() = 0; + virtual bool getLuminanceAlphaTextureSupport() = 0; + bool getVertexTextureSupport() const { return getMaxVertexTextureImageUnits() > 0; } + virtual unsigned int getMaxVertexTextureImageUnits() const = 0; + virtual unsigned int getMaxCombinedTextureImageUnits() const = 0; + virtual unsigned int getReservedVertexUniformVectors() const = 0; + virtual unsigned int getReservedFragmentUniformVectors() const = 0; + virtual unsigned int getMaxVertexUniformVectors() const = 0; + virtual unsigned int getMaxFragmentUniformVectors() const = 0; + virtual unsigned int getMaxVaryingVectors() const = 0; + virtual bool getNonPower2TextureSupport() const = 0; + virtual bool getDepthTextureSupport() const = 0; + virtual bool getOcclusionQuerySupport() const = 0; + virtual bool getInstancingSupport() const = 0; + virtual bool getTextureFilterAnisotropySupport() const = 0; + virtual float getTextureMaxAnisotropy() const = 0; + virtual bool getShareHandleSupport() const = 0; + virtual bool getDerivativeInstructionSupport() const = 0; + virtual bool getPostSubBufferSupport() const = 0; + + virtual int getMajorShaderModel() const = 0; + virtual float getMaxPointSize() const = 0; + virtual int getMaxViewportDimension() const = 0; + virtual int getMaxTextureWidth() const = 0; + virtual int getMaxTextureHeight() const = 0; + virtual bool get32BitIndexSupport() const = 0; + virtual int getMinSwapInterval() const = 0; + virtual int getMaxSwapInterval() const = 0; + + virtual GLsizei getMaxSupportedSamples() const = 0; + + virtual unsigned int getMaxRenderTargets() const = 0; + + // Pixel operations + virtual bool copyToRenderTarget(TextureStorageInterface2D *dest, TextureStorageInterface2D *source) = 0; + virtual bool copyToRenderTarget(TextureStorageInterfaceCube *dest, TextureStorageInterfaceCube *source) = 0; + + virtual bool copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterface2D *storage, GLint level) = 0; + virtual bool copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterfaceCube *storage, GLenum target, GLint level) = 0; + + virtual bool blitRect(gl::Framebuffer *readTarget, const gl::Rectangle &readRect, gl::Framebuffer *drawTarget, const gl::Rectangle &drawRect, + bool blitRenderTarget, bool blitDepthStencil) = 0; + virtual void readPixels(gl::Framebuffer *framebuffer, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, + GLsizei outputPitch, bool packReverseRowOrder, GLint packAlignment, void* pixels) = 0; + + // RenderTarget creation + virtual RenderTarget *createRenderTarget(SwapChain *swapChain, bool depth) = 0; + virtual RenderTarget *createRenderTarget(int width, int height, GLenum format, GLsizei samples, bool depth) = 0; + + // Shader operations + virtual ShaderExecutable *loadExecutable(const void *function, size_t length, rx::ShaderType type) = 0; + virtual ShaderExecutable *compileToExecutable(gl::InfoLog &infoLog, const char *shaderHLSL, rx::ShaderType type) = 0; + + // Image operations + virtual Image *createImage() = 0; + virtual void generateMipmap(Image *dest, Image *source) = 0; + virtual TextureStorage *createTextureStorage2D(SwapChain *swapChain) = 0; + virtual TextureStorage *createTextureStorage2D(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height) = 0; + virtual TextureStorage *createTextureStorageCube(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size) = 0; + + // Buffer creation + virtual VertexBuffer *createVertexBuffer() = 0; + virtual IndexBuffer *createIndexBuffer() = 0; + virtual BufferStorage *createBufferStorage() = 0; + + // Query and Fence creation + virtual QueryImpl *createQuery(GLenum type) = 0; + virtual FenceImpl *createFence() = 0; + + protected: + bool initializeCompiler(); + ShaderBlob *compileToBinary(gl::InfoLog &infoLog, const char *hlsl, const char *profile, UINT optimizationFlags, bool alternateFlags); + + egl::Display *mDisplay; + + private: + DISALLOW_COPY_AND_ASSIGN(Renderer); + + HMODULE mD3dCompilerModule; + pCompileFunc mD3DCompileFunc; +}; + +} +#endif // LIBGLESV2_RENDERER_RENDERER_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Renderer11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer11.cpp new file mode 100644 index 0000000000..cf083963e1 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer11.cpp @@ -0,0 +1,3531 @@ +#include "precompiled.h" +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Renderer11.cpp: Implements a back-end specific class for the D3D11 renderer. + +#include "libGLESv2/main.h" +#include "libGLESv2/utilities.h" +#include "libGLESv2/Buffer.h" +#include "libGLESv2/ProgramBinary.h" +#include "libGLESv2/Framebuffer.h" +#include "libGLESv2/RenderBuffer.h" +#include "libGLESv2/renderer/Renderer11.h" +#include "libGLESv2/renderer/RenderTarget11.h" +#include "libGLESv2/renderer/renderer11_utils.h" +#include "libGLESv2/renderer/ShaderExecutable11.h" +#include "libGLESv2/renderer/SwapChain11.h" +#include "libGLESv2/renderer/Image11.h" +#include "libGLESv2/renderer/VertexBuffer11.h" +#include "libGLESv2/renderer/IndexBuffer11.h" +#include "libGLESv2/renderer/BufferStorage11.h" +#include "libGLESv2/renderer/VertexDataManager.h" +#include "libGLESv2/renderer/IndexDataManager.h" +#include "libGLESv2/renderer/TextureStorage11.h" +#include "libGLESv2/renderer/Query11.h" +#include "libGLESv2/renderer/Fence11.h" + +#include "libGLESv2/renderer/shaders/compiled/passthrough11vs.h" +#include "libGLESv2/renderer/shaders/compiled/passthroughrgba11ps.h" +#include "libGLESv2/renderer/shaders/compiled/passthroughrgb11ps.h" +#include "libGLESv2/renderer/shaders/compiled/passthroughlum11ps.h" +#include "libGLESv2/renderer/shaders/compiled/passthroughlumalpha11ps.h" + +#include "libGLESv2/renderer/shaders/compiled/clear11vs.h" +#include "libGLESv2/renderer/shaders/compiled/clear11ps.h" + +#include "libEGL/Display.h" + +#ifdef _DEBUG +// this flag enables suppressing some spurious warnings that pop up in certain WebGL samples +// and conformance tests. to enable all warnings, remove this define. +#define ANGLE_SUPPRESS_D3D11_HAZARD_WARNINGS 1 +#endif + +namespace rx +{ +static const DXGI_FORMAT RenderTargetFormats[] = + { + DXGI_FORMAT_B8G8R8A8_UNORM, + DXGI_FORMAT_R8G8B8A8_UNORM + }; + +static const DXGI_FORMAT DepthStencilFormats[] = + { + DXGI_FORMAT_UNKNOWN, + DXGI_FORMAT_D24_UNORM_S8_UINT, + DXGI_FORMAT_D16_UNORM + }; + +enum +{ + MAX_TEXTURE_IMAGE_UNITS_VTF_SM4 = 16 +}; + +Renderer11::Renderer11(egl::Display *display, HDC hDc) : Renderer(display), mDc(hDc) +{ + mVertexDataManager = NULL; + mIndexDataManager = NULL; + + mLineLoopIB = NULL; + mTriangleFanIB = NULL; + + mCopyResourcesInitialized = false; + mCopyVB = NULL; + mCopySampler = NULL; + mCopyIL = NULL; + mCopyVS = NULL; + mCopyRGBAPS = NULL; + mCopyRGBPS = NULL; + mCopyLumPS = NULL; + mCopyLumAlphaPS = NULL; + + mClearResourcesInitialized = false; + mClearVB = NULL; + mClearIL = NULL; + mClearVS = NULL; + mClearPS = NULL; + mClearScissorRS = NULL; + mClearNoScissorRS = NULL; + + mSyncQuery = NULL; + + mD3d11Module = NULL; + mDxgiModule = NULL; + + mDeviceLost = false; + + mMaxSupportedSamples = 0; + + mDevice = NULL; + mDeviceContext = NULL; + mDxgiAdapter = NULL; + mDxgiFactory = NULL; + + mDriverConstantBufferVS = NULL; + mDriverConstantBufferPS = NULL; + + mBGRATextureSupport = false; + + mIsGeometryShaderActive = false; +} + +Renderer11::~Renderer11() +{ + release(); +} + +Renderer11 *Renderer11::makeRenderer11(Renderer *renderer) +{ + ASSERT(HAS_DYNAMIC_TYPE(rx::Renderer11*, renderer)); + return static_cast<rx::Renderer11*>(renderer); +} + +#ifndef __d3d11_1_h__ +#define D3D11_MESSAGE_ID_DEVICE_DRAW_RENDERTARGETVIEW_NOT_SET ((D3D11_MESSAGE_ID)3146081) +#endif + +EGLint Renderer11::initialize() +{ + if (!initializeCompiler()) + { + return EGL_NOT_INITIALIZED; + } + + mDxgiModule = LoadLibrary(TEXT("dxgi.dll")); + mD3d11Module = LoadLibrary(TEXT("d3d11.dll")); + + if (mD3d11Module == NULL || mDxgiModule == NULL) + { + ERR("Could not load D3D11 or DXGI library - aborting!\n"); + return EGL_NOT_INITIALIZED; + } + + // create the D3D11 device + ASSERT(mDevice == NULL); + PFN_D3D11_CREATE_DEVICE D3D11CreateDevice = (PFN_D3D11_CREATE_DEVICE)GetProcAddress(mD3d11Module, "D3D11CreateDevice"); + + if (D3D11CreateDevice == NULL) + { + ERR("Could not retrieve D3D11CreateDevice address - aborting!\n"); + return EGL_NOT_INITIALIZED; + } + + D3D_FEATURE_LEVEL featureLevels[] = + { + D3D_FEATURE_LEVEL_11_0, + D3D_FEATURE_LEVEL_10_1, + D3D_FEATURE_LEVEL_10_0, + }; + + HRESULT result = D3D11CreateDevice(NULL, + D3D_DRIVER_TYPE_HARDWARE, + NULL, + #if defined(_DEBUG) + D3D11_CREATE_DEVICE_DEBUG, + #else + 0, + #endif + featureLevels, + ArraySize(featureLevels), + D3D11_SDK_VERSION, + &mDevice, + &mFeatureLevel, + &mDeviceContext); + + if (!mDevice || FAILED(result)) + { + ERR("Could not create D3D11 device - aborting!\n"); + return EGL_NOT_INITIALIZED; // Cleanup done by destructor through glDestroyRenderer + } + + IDXGIDevice *dxgiDevice = NULL; + result = mDevice->QueryInterface(__uuidof(IDXGIDevice), (void**)&dxgiDevice); + + if (FAILED(result)) + { + ERR("Could not query DXGI device - aborting!\n"); + return EGL_NOT_INITIALIZED; + } + + result = dxgiDevice->GetParent(__uuidof(IDXGIAdapter), (void**)&mDxgiAdapter); + + if (FAILED(result)) + { + ERR("Could not retrieve DXGI adapter - aborting!\n"); + return EGL_NOT_INITIALIZED; + } + + dxgiDevice->Release(); + + mDxgiAdapter->GetDesc(&mAdapterDescription); + memset(mDescription, 0, sizeof(mDescription)); + wcstombs(mDescription, mAdapterDescription.Description, sizeof(mDescription) - 1); + + result = mDxgiAdapter->GetParent(__uuidof(IDXGIFactory), (void**)&mDxgiFactory); + + if (!mDxgiFactory || FAILED(result)) + { + ERR("Could not create DXGI factory - aborting!\n"); + return EGL_NOT_INITIALIZED; + } + + // Disable some spurious D3D11 debug warnings to prevent them from flooding the output log +#if defined(ANGLE_SUPPRESS_D3D11_HAZARD_WARNINGS) && defined(_DEBUG) + ID3D11InfoQueue *infoQueue; + result = mDevice->QueryInterface(__uuidof(ID3D11InfoQueue), (void **)&infoQueue); + + if (SUCCEEDED(result)) + { + D3D11_MESSAGE_ID hideMessages[] = + { + D3D11_MESSAGE_ID_DEVICE_OMSETRENDERTARGETS_HAZARD, + D3D11_MESSAGE_ID_DEVICE_PSSETSHADERRESOURCES_HAZARD, + D3D11_MESSAGE_ID_DEVICE_DRAW_RENDERTARGETVIEW_NOT_SET + }; + + D3D11_INFO_QUEUE_FILTER filter = {0}; + filter.DenyList.NumIDs = ArraySize(hideMessages); + filter.DenyList.pIDList = hideMessages; + + infoQueue->AddStorageFilterEntries(&filter); + + infoQueue->Release(); + } +#endif + + unsigned int maxSupportedSamples = 0; + unsigned int rtFormatCount = ArraySize(RenderTargetFormats); + unsigned int dsFormatCount = ArraySize(DepthStencilFormats); + for (unsigned int i = 0; i < rtFormatCount + dsFormatCount; ++i) + { + DXGI_FORMAT format = (i < rtFormatCount) ? RenderTargetFormats[i] : DepthStencilFormats[i - rtFormatCount]; + if (format != DXGI_FORMAT_UNKNOWN) + { + UINT formatSupport; + result = mDevice->CheckFormatSupport(format, &formatSupport); + if (SUCCEEDED(result) && (formatSupport & D3D11_FORMAT_SUPPORT_MULTISAMPLE_RENDERTARGET)) + { + MultisampleSupportInfo supportInfo; + + for (unsigned int j = 1; j <= D3D11_MAX_MULTISAMPLE_SAMPLE_COUNT; j++) + { + result = mDevice->CheckMultisampleQualityLevels(format, j, &supportInfo.qualityLevels[j - 1]); + if (SUCCEEDED(result) && supportInfo.qualityLevels[j - 1] > 0) + { + maxSupportedSamples = std::max(j, maxSupportedSamples); + } + else + { + supportInfo.qualityLevels[j - 1] = 0; + } + } + + mMultisampleSupportMap.insert(std::make_pair(format, supportInfo)); + } + } + } + mMaxSupportedSamples = maxSupportedSamples; + + initializeDevice(); + + // BGRA texture support is optional in feature levels 10 and 10_1 + UINT formatSupport; + result = mDevice->CheckFormatSupport(DXGI_FORMAT_B8G8R8A8_UNORM, &formatSupport); + if (FAILED(result)) + { + ERR("Error checking BGRA format support: 0x%08X", result); + } + else + { + const int flags = (D3D11_FORMAT_SUPPORT_TEXTURE2D | D3D11_FORMAT_SUPPORT_RENDER_TARGET); + mBGRATextureSupport = (formatSupport & flags) == flags; + } + + // Check floating point texture support + static const unsigned int requiredTextureFlags = D3D11_FORMAT_SUPPORT_TEXTURE2D | D3D11_FORMAT_SUPPORT_TEXTURECUBE; + static const unsigned int requiredRenderableFlags = D3D11_FORMAT_SUPPORT_RENDER_TARGET; + static const unsigned int requiredFilterFlags = D3D11_FORMAT_SUPPORT_SHADER_SAMPLE; + + DXGI_FORMAT float16Formats[] = + { + DXGI_FORMAT_R16_FLOAT, + DXGI_FORMAT_R16G16_FLOAT, + DXGI_FORMAT_R16G16B16A16_FLOAT, + }; + + DXGI_FORMAT float32Formats[] = + { + DXGI_FORMAT_R32_FLOAT, + DXGI_FORMAT_R32G32_FLOAT, + DXGI_FORMAT_R32G32B32_FLOAT, + DXGI_FORMAT_R32G32B32A32_FLOAT, + }; + + mFloat16TextureSupport = true; + mFloat16FilterSupport = true; + mFloat16RenderSupport = true; + for (unsigned int i = 0; i < ArraySize(float16Formats); i++) + { + if (SUCCEEDED(mDevice->CheckFormatSupport(float16Formats[i], &formatSupport))) + { + mFloat16TextureSupport = mFloat16TextureSupport && (formatSupport & requiredTextureFlags) == requiredTextureFlags; + mFloat16FilterSupport = mFloat16FilterSupport && (formatSupport & requiredFilterFlags) == requiredFilterFlags; + mFloat16RenderSupport = mFloat16RenderSupport && (formatSupport & requiredRenderableFlags) == requiredRenderableFlags; + } + else + { + mFloat16TextureSupport = false; + mFloat16RenderSupport = false; + mFloat16FilterSupport = false; + } + } + + mFloat32TextureSupport = true; + mFloat32FilterSupport = true; + mFloat32RenderSupport = true; + for (unsigned int i = 0; i < ArraySize(float32Formats); i++) + { + if (SUCCEEDED(mDevice->CheckFormatSupport(float32Formats[i], &formatSupport))) + { + mFloat32TextureSupport = mFloat32TextureSupport && (formatSupport & requiredTextureFlags) == requiredTextureFlags; + mFloat32FilterSupport = mFloat32FilterSupport && (formatSupport & requiredFilterFlags) == requiredFilterFlags; + mFloat32RenderSupport = mFloat32RenderSupport && (formatSupport & requiredRenderableFlags) == requiredRenderableFlags; + } + else + { + mFloat32TextureSupport = false; + mFloat32FilterSupport = false; + mFloat32RenderSupport = false; + } + } + + // Check compressed texture support + const unsigned int requiredCompressedTextureFlags = D3D11_FORMAT_SUPPORT_TEXTURE2D; + + if (SUCCEEDED(mDevice->CheckFormatSupport(DXGI_FORMAT_BC1_UNORM, &formatSupport))) + { + mDXT1TextureSupport = (formatSupport & requiredCompressedTextureFlags) == requiredCompressedTextureFlags; + } + else + { + mDXT1TextureSupport = false; + } + + if (SUCCEEDED(mDevice->CheckFormatSupport(DXGI_FORMAT_BC3_UNORM, &formatSupport))) + { + mDXT3TextureSupport = (formatSupport & requiredCompressedTextureFlags) == requiredCompressedTextureFlags; + } + else + { + mDXT3TextureSupport = false; + } + + if (SUCCEEDED(mDevice->CheckFormatSupport(DXGI_FORMAT_BC5_UNORM, &formatSupport))) + { + mDXT5TextureSupport = (formatSupport & requiredCompressedTextureFlags) == requiredCompressedTextureFlags; + } + else + { + mDXT5TextureSupport = false; + } + + // Check depth texture support + DXGI_FORMAT depthTextureFormats[] = + { + DXGI_FORMAT_D16_UNORM, + DXGI_FORMAT_D24_UNORM_S8_UINT, + }; + + static const unsigned int requiredDepthTextureFlags = D3D11_FORMAT_SUPPORT_DEPTH_STENCIL | + D3D11_FORMAT_SUPPORT_TEXTURE2D; + + mDepthTextureSupport = true; + for (unsigned int i = 0; i < ArraySize(depthTextureFormats); i++) + { + if (SUCCEEDED(mDevice->CheckFormatSupport(depthTextureFormats[i], &formatSupport))) + { + mDepthTextureSupport = mDepthTextureSupport && ((formatSupport & requiredDepthTextureFlags) == requiredDepthTextureFlags); + } + else + { + mDepthTextureSupport = false; + } + } + + return EGL_SUCCESS; +} + +// do any one-time device initialization +// NOTE: this is also needed after a device lost/reset +// to reset the scene status and ensure the default states are reset. +void Renderer11::initializeDevice() +{ + mStateCache.initialize(mDevice); + mInputLayoutCache.initialize(mDevice, mDeviceContext); + + ASSERT(!mVertexDataManager && !mIndexDataManager); + mVertexDataManager = new VertexDataManager(this); + mIndexDataManager = new IndexDataManager(this); + + markAllStateDirty(); +} + +int Renderer11::generateConfigs(ConfigDesc **configDescList) +{ + unsigned int numRenderFormats = ArraySize(RenderTargetFormats); + unsigned int numDepthFormats = ArraySize(DepthStencilFormats); + (*configDescList) = new ConfigDesc[numRenderFormats * numDepthFormats]; + int numConfigs = 0; + + for (unsigned int formatIndex = 0; formatIndex < numRenderFormats; formatIndex++) + { + for (unsigned int depthStencilIndex = 0; depthStencilIndex < numDepthFormats; depthStencilIndex++) + { + DXGI_FORMAT renderTargetFormat = RenderTargetFormats[formatIndex]; + + UINT formatSupport = 0; + HRESULT result = mDevice->CheckFormatSupport(renderTargetFormat, &formatSupport); + + if (SUCCEEDED(result) && (formatSupport & D3D11_FORMAT_SUPPORT_RENDER_TARGET)) + { + DXGI_FORMAT depthStencilFormat = DepthStencilFormats[depthStencilIndex]; + + bool depthStencilFormatOK = true; + + if (depthStencilFormat != DXGI_FORMAT_UNKNOWN) + { + UINT formatSupport = 0; + result = mDevice->CheckFormatSupport(depthStencilFormat, &formatSupport); + depthStencilFormatOK = SUCCEEDED(result) && (formatSupport & D3D11_FORMAT_SUPPORT_DEPTH_STENCIL); + } + + if (depthStencilFormatOK) + { + ConfigDesc newConfig; + newConfig.renderTargetFormat = d3d11_gl::ConvertBackBufferFormat(renderTargetFormat); + newConfig.depthStencilFormat = d3d11_gl::ConvertDepthStencilFormat(depthStencilFormat); + newConfig.multiSample = 0; // FIXME: enumerate multi-sampling + newConfig.fastConfig = true; // Assume all DX11 format conversions to be fast + + (*configDescList)[numConfigs++] = newConfig; + } + } + } + } + + return numConfigs; +} + +void Renderer11::deleteConfigs(ConfigDesc *configDescList) +{ + delete [] (configDescList); +} + +void Renderer11::sync(bool block) +{ + if (block) + { + HRESULT result; + + if (!mSyncQuery) + { + D3D11_QUERY_DESC queryDesc; + queryDesc.Query = D3D11_QUERY_EVENT; + queryDesc.MiscFlags = 0; + + result = mDevice->CreateQuery(&queryDesc, &mSyncQuery); + ASSERT(SUCCEEDED(result)); + } + + mDeviceContext->End(mSyncQuery); + mDeviceContext->Flush(); + + do + { + result = mDeviceContext->GetData(mSyncQuery, NULL, 0, D3D11_ASYNC_GETDATA_DONOTFLUSH); + + // Keep polling, but allow other threads to do something useful first + Sleep(0); + + if (testDeviceLost(true)) + { + return; + } + } + while (result == S_FALSE); + } + else + { + mDeviceContext->Flush(); + } +} + +SwapChain *Renderer11::createSwapChain(HWND window, HANDLE shareHandle, GLenum backBufferFormat, GLenum depthBufferFormat) +{ + return new rx::SwapChain11(this, window, shareHandle, backBufferFormat, depthBufferFormat); +} + +void Renderer11::setSamplerState(gl::SamplerType type, int index, const gl::SamplerState &samplerState) +{ + if (type == gl::SAMPLER_PIXEL) + { + if (index < 0 || index >= gl::MAX_TEXTURE_IMAGE_UNITS) + { + ERR("Pixel shader sampler index %i is not valid.", index); + return; + } + + if (mForceSetPixelSamplerStates[index] || memcmp(&samplerState, &mCurPixelSamplerStates[index], sizeof(gl::SamplerState)) != 0) + { + ID3D11SamplerState *dxSamplerState = mStateCache.getSamplerState(samplerState); + + if (!dxSamplerState) + { + ERR("NULL sampler state returned by RenderStateCache::getSamplerState, setting the default" + "sampler state for pixel shaders at slot %i.", index); + } + + mDeviceContext->PSSetSamplers(index, 1, &dxSamplerState); + + mCurPixelSamplerStates[index] = samplerState; + } + + mForceSetPixelSamplerStates[index] = false; + } + else if (type == gl::SAMPLER_VERTEX) + { + if (index < 0 || index >= (int)getMaxVertexTextureImageUnits()) + { + ERR("Vertex shader sampler index %i is not valid.", index); + return; + } + + if (mForceSetVertexSamplerStates[index] || memcmp(&samplerState, &mCurVertexSamplerStates[index], sizeof(gl::SamplerState)) != 0) + { + ID3D11SamplerState *dxSamplerState = mStateCache.getSamplerState(samplerState); + + if (!dxSamplerState) + { + ERR("NULL sampler state returned by RenderStateCache::getSamplerState, setting the default" + "sampler state for vertex shaders at slot %i.", index); + } + + mDeviceContext->VSSetSamplers(index, 1, &dxSamplerState); + + mCurVertexSamplerStates[index] = samplerState; + } + + mForceSetVertexSamplerStates[index] = false; + } + else UNREACHABLE(); +} + +void Renderer11::setTexture(gl::SamplerType type, int index, gl::Texture *texture) +{ + ID3D11ShaderResourceView *textureSRV = NULL; + unsigned int serial = 0; + bool forceSetTexture = false; + + if (texture) + { + TextureStorageInterface *texStorage = texture->getNativeTexture(); + if (texStorage) + { + TextureStorage11 *storage11 = TextureStorage11::makeTextureStorage11(texStorage->getStorageInstance()); + textureSRV = storage11->getSRV(); + } + + // If we get NULL back from getSRV here, something went wrong in the texture class and we're unexpectedly + // missing the shader resource view + ASSERT(textureSRV != NULL); + + serial = texture->getTextureSerial(); + forceSetTexture = texture->hasDirtyImages(); + } + + if (type == gl::SAMPLER_PIXEL) + { + if (index < 0 || index >= gl::MAX_TEXTURE_IMAGE_UNITS) + { + ERR("Pixel shader sampler index %i is not valid.", index); + return; + } + + if (forceSetTexture || mCurPixelTextureSerials[index] != serial) + { + mDeviceContext->PSSetShaderResources(index, 1, &textureSRV); + } + + mCurPixelTextureSerials[index] = serial; + } + else if (type == gl::SAMPLER_VERTEX) + { + if (index < 0 || index >= (int)getMaxVertexTextureImageUnits()) + { + ERR("Vertex shader sampler index %i is not valid.", index); + return; + } + + if (forceSetTexture || mCurVertexTextureSerials[index] != serial) + { + mDeviceContext->VSSetShaderResources(index, 1, &textureSRV); + } + + mCurVertexTextureSerials[index] = serial; + } + else UNREACHABLE(); +} + +void Renderer11::setRasterizerState(const gl::RasterizerState &rasterState) +{ + if (mForceSetRasterState || memcmp(&rasterState, &mCurRasterState, sizeof(gl::RasterizerState)) != 0) + { + ID3D11RasterizerState *dxRasterState = mStateCache.getRasterizerState(rasterState, mScissorEnabled, + mCurDepthSize); + if (!dxRasterState) + { + ERR("NULL rasterizer state returned by RenderStateCache::getRasterizerState, setting the default" + "rasterizer state."); + } + + mDeviceContext->RSSetState(dxRasterState); + + mCurRasterState = rasterState; + } + + mForceSetRasterState = false; +} + +void Renderer11::setBlendState(const gl::BlendState &blendState, const gl::Color &blendColor, + unsigned int sampleMask) +{ + if (mForceSetBlendState || + memcmp(&blendState, &mCurBlendState, sizeof(gl::BlendState)) != 0 || + memcmp(&blendColor, &mCurBlendColor, sizeof(gl::Color)) != 0 || + sampleMask != mCurSampleMask) + { + ID3D11BlendState *dxBlendState = mStateCache.getBlendState(blendState); + if (!dxBlendState) + { + ERR("NULL blend state returned by RenderStateCache::getBlendState, setting the default " + "blend state."); + } + + const float blendColors[] = { blendColor.red, blendColor.green, blendColor.blue, blendColor.alpha }; + mDeviceContext->OMSetBlendState(dxBlendState, blendColors, sampleMask); + + mCurBlendState = blendState; + mCurBlendColor = blendColor; + mCurSampleMask = sampleMask; + } + + mForceSetBlendState = false; +} + +void Renderer11::setDepthStencilState(const gl::DepthStencilState &depthStencilState, int stencilRef, + int stencilBackRef, bool frontFaceCCW) +{ + if (mForceSetDepthStencilState || + memcmp(&depthStencilState, &mCurDepthStencilState, sizeof(gl::DepthStencilState)) != 0 || + stencilRef != mCurStencilRef || stencilBackRef != mCurStencilBackRef) + { + if (depthStencilState.stencilWritemask != depthStencilState.stencilBackWritemask || + stencilRef != stencilBackRef || + depthStencilState.stencilMask != depthStencilState.stencilBackMask) + { + ERR("Separate front/back stencil writemasks, reference values, or stencil mask values are " + "invalid under WebGL."); + return gl::error(GL_INVALID_OPERATION); + } + + ID3D11DepthStencilState *dxDepthStencilState = mStateCache.getDepthStencilState(depthStencilState); + if (!dxDepthStencilState) + { + ERR("NULL depth stencil state returned by RenderStateCache::getDepthStencilState, " + "setting the default depth stencil state."); + } + + mDeviceContext->OMSetDepthStencilState(dxDepthStencilState, static_cast<UINT>(stencilRef)); + + mCurDepthStencilState = depthStencilState; + mCurStencilRef = stencilRef; + mCurStencilBackRef = stencilBackRef; + } + + mForceSetDepthStencilState = false; +} + +void Renderer11::setScissorRectangle(const gl::Rectangle &scissor, bool enabled) +{ + if (mForceSetScissor || memcmp(&scissor, &mCurScissor, sizeof(gl::Rectangle)) != 0 || + enabled != mScissorEnabled) + { + if (enabled) + { + D3D11_RECT rect; + rect.left = std::max(0, scissor.x); + rect.top = std::max(0, scissor.y); + rect.right = scissor.x + std::max(0, scissor.width); + rect.bottom = scissor.y + std::max(0, scissor.height); + + mDeviceContext->RSSetScissorRects(1, &rect); + } + + if (enabled != mScissorEnabled) + { + mForceSetRasterState = true; + } + + mCurScissor = scissor; + mScissorEnabled = enabled; + } + + mForceSetScissor = false; +} + +bool Renderer11::setViewport(const gl::Rectangle &viewport, float zNear, float zFar, GLenum drawMode, GLenum frontFace, + bool ignoreViewport) +{ + gl::Rectangle actualViewport = viewport; + float actualZNear = gl::clamp01(zNear); + float actualZFar = gl::clamp01(zFar); + if (ignoreViewport) + { + actualViewport.x = 0; + actualViewport.y = 0; + actualViewport.width = mRenderTargetDesc.width; + actualViewport.height = mRenderTargetDesc.height; + actualZNear = 0.0f; + actualZFar = 1.0f; + } + + // Get D3D viewport bounds, which depends on the feature level + const Range& viewportBounds = getViewportBounds(); + + // Clamp width and height first to the gl maximum, then clamp further if we extend past the D3D maximum bounds + D3D11_VIEWPORT dxViewport; + dxViewport.TopLeftX = gl::clamp(actualViewport.x, viewportBounds.start, viewportBounds.end); + dxViewport.TopLeftY = gl::clamp(actualViewport.y, viewportBounds.start, viewportBounds.end); + dxViewport.Width = gl::clamp(actualViewport.width, 0, getMaxViewportDimension()); + dxViewport.Height = gl::clamp(actualViewport.height, 0, getMaxViewportDimension()); + dxViewport.Width = std::min((int)dxViewport.Width, viewportBounds.end - static_cast<int>(dxViewport.TopLeftX)); + dxViewport.Height = std::min((int)dxViewport.Height, viewportBounds.end - static_cast<int>(dxViewport.TopLeftY)); + dxViewport.MinDepth = actualZNear; + dxViewport.MaxDepth = actualZFar; + + if (dxViewport.Width <= 0 || dxViewport.Height <= 0) + { + return false; // Nothing to render + } + + bool viewportChanged = mForceSetViewport || memcmp(&actualViewport, &mCurViewport, sizeof(gl::Rectangle)) != 0 || + actualZNear != mCurNear || actualZFar != mCurFar; + + if (viewportChanged) + { + mDeviceContext->RSSetViewports(1, &dxViewport); + + mCurViewport = actualViewport; + mCurNear = actualZNear; + mCurFar = actualZFar; + + mPixelConstants.viewCoords[0] = actualViewport.width * 0.5f; + mPixelConstants.viewCoords[1] = actualViewport.height * 0.5f; + mPixelConstants.viewCoords[2] = actualViewport.x + (actualViewport.width * 0.5f); + mPixelConstants.viewCoords[3] = actualViewport.y + (actualViewport.height * 0.5f); + + mPixelConstants.depthFront[0] = (actualZFar - actualZNear) * 0.5f; + mPixelConstants.depthFront[1] = (actualZNear + actualZFar) * 0.5f; + + mVertexConstants.depthRange[0] = actualZNear; + mVertexConstants.depthRange[1] = actualZFar; + mVertexConstants.depthRange[2] = actualZFar - actualZNear; + + mPixelConstants.depthRange[0] = actualZNear; + mPixelConstants.depthRange[1] = actualZFar; + mPixelConstants.depthRange[2] = actualZFar - actualZNear; + } + + mForceSetViewport = false; + return true; +} + +bool Renderer11::applyPrimitiveType(GLenum mode, GLsizei count) +{ + D3D11_PRIMITIVE_TOPOLOGY primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_UNDEFINED; + + switch (mode) + { + case GL_POINTS: primitiveTopology = D3D11_PRIMITIVE_TOPOLOGY_POINTLIST; break; + case GL_LINES: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINELIST; break; + case GL_LINE_LOOP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINESTRIP; break; + case GL_LINE_STRIP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINESTRIP; break; + case GL_TRIANGLES: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; break; + case GL_TRIANGLE_STRIP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP; break; + // emulate fans via rewriting index buffer + case GL_TRIANGLE_FAN: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; break; + default: + return gl::error(GL_INVALID_ENUM, false); + } + + mDeviceContext->IASetPrimitiveTopology(primitiveTopology); + + return count > 0; +} + +bool Renderer11::applyRenderTarget(gl::Framebuffer *framebuffer) +{ + // Get the color render buffer and serial + // Also extract the render target dimensions and view + unsigned int renderTargetWidth = 0; + unsigned int renderTargetHeight = 0; + GLenum renderTargetFormat = 0; + unsigned int renderTargetSerials[gl::IMPLEMENTATION_MAX_DRAW_BUFFERS] = {0}; + ID3D11RenderTargetView* framebufferRTVs[gl::IMPLEMENTATION_MAX_DRAW_BUFFERS] = {NULL}; + bool missingColorRenderTarget = true; + + for (unsigned int colorAttachment = 0; colorAttachment < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; colorAttachment++) + { + const GLenum drawBufferState = framebuffer->getDrawBufferState(colorAttachment); + + if (framebuffer->getColorbufferType(colorAttachment) != GL_NONE && drawBufferState != GL_NONE) + { + // the draw buffer must be either "none", "back" for the default buffer or the same index as this color (in order) + ASSERT(drawBufferState == GL_BACK || drawBufferState == (GL_COLOR_ATTACHMENT0_EXT + colorAttachment)); + + gl::Renderbuffer *colorbuffer = framebuffer->getColorbuffer(colorAttachment); + + if (!colorbuffer) + { + ERR("render target pointer unexpectedly null."); + return false; + } + + // check for zero-sized default framebuffer, which is a special case. + // in this case we do not wish to modify any state and just silently return false. + // this will not report any gl error but will cause the calling method to return. + if (colorbuffer->getWidth() == 0 || colorbuffer->getHeight() == 0) + { + return false; + } + + renderTargetSerials[colorAttachment] = colorbuffer->getSerial(); + + // Extract the render target dimensions and view + RenderTarget11 *renderTarget = RenderTarget11::makeRenderTarget11(colorbuffer->getRenderTarget()); + if (!renderTarget) + { + ERR("render target pointer unexpectedly null."); + return false; + } + + framebufferRTVs[colorAttachment] = renderTarget->getRenderTargetView(); + if (!framebufferRTVs[colorAttachment]) + { + ERR("render target view pointer unexpectedly null."); + return false; + } + + if (missingColorRenderTarget) + { + renderTargetWidth = colorbuffer->getWidth(); + renderTargetHeight = colorbuffer->getHeight(); + renderTargetFormat = colorbuffer->getActualFormat(); + missingColorRenderTarget = false; + } + } + } + + // Get the depth stencil render buffer and serials + gl::Renderbuffer *depthStencil = NULL; + unsigned int depthbufferSerial = 0; + unsigned int stencilbufferSerial = 0; + if (framebuffer->getDepthbufferType() != GL_NONE) + { + depthStencil = framebuffer->getDepthbuffer(); + if (!depthStencil) + { + ERR("Depth stencil pointer unexpectedly null."); + SafeRelease(framebufferRTVs); + return false; + } + + depthbufferSerial = depthStencil->getSerial(); + } + else if (framebuffer->getStencilbufferType() != GL_NONE) + { + depthStencil = framebuffer->getStencilbuffer(); + if (!depthStencil) + { + ERR("Depth stencil pointer unexpectedly null."); + SafeRelease(framebufferRTVs); + return false; + } + + stencilbufferSerial = depthStencil->getSerial(); + } + + // Extract the depth stencil sizes and view + unsigned int depthSize = 0; + unsigned int stencilSize = 0; + ID3D11DepthStencilView* framebufferDSV = NULL; + if (depthStencil) + { + RenderTarget11 *depthStencilRenderTarget = RenderTarget11::makeRenderTarget11(depthStencil->getDepthStencil()); + if (!depthStencilRenderTarget) + { + ERR("render target pointer unexpectedly null."); + SafeRelease(framebufferRTVs); + return false; + } + + framebufferDSV = depthStencilRenderTarget->getDepthStencilView(); + if (!framebufferDSV) + { + ERR("depth stencil view pointer unexpectedly null."); + SafeRelease(framebufferRTVs); + return false; + } + + // If there is no render buffer, the width, height and format values come from + // the depth stencil + if (missingColorRenderTarget) + { + renderTargetWidth = depthStencil->getWidth(); + renderTargetHeight = depthStencil->getHeight(); + renderTargetFormat = depthStencil->getActualFormat(); + } + + depthSize = depthStencil->getDepthSize(); + stencilSize = depthStencil->getStencilSize(); + } + + // Apply the render target and depth stencil + if (!mRenderTargetDescInitialized || !mDepthStencilInitialized || + memcmp(renderTargetSerials, mAppliedRenderTargetSerials, sizeof(renderTargetSerials)) != 0 || + depthbufferSerial != mAppliedDepthbufferSerial || + stencilbufferSerial != mAppliedStencilbufferSerial) + { + mDeviceContext->OMSetRenderTargets(getMaxRenderTargets(), framebufferRTVs, framebufferDSV); + + mRenderTargetDesc.width = renderTargetWidth; + mRenderTargetDesc.height = renderTargetHeight; + mRenderTargetDesc.format = renderTargetFormat; + mForceSetViewport = true; + mForceSetScissor = true; + + if (!mDepthStencilInitialized || depthSize != mCurDepthSize) + { + mCurDepthSize = depthSize; + mForceSetRasterState = true; + } + + mCurStencilSize = stencilSize; + + for (unsigned int rtIndex = 0; rtIndex < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; rtIndex++) + { + mAppliedRenderTargetSerials[rtIndex] = renderTargetSerials[rtIndex]; + } + mAppliedDepthbufferSerial = depthbufferSerial; + mAppliedStencilbufferSerial = stencilbufferSerial; + mRenderTargetDescInitialized = true; + mDepthStencilInitialized = true; + } + + SafeRelease(framebufferRTVs); + SafeRelease(framebufferDSV); + + return true; +} + +GLenum Renderer11::applyVertexBuffer(gl::ProgramBinary *programBinary, gl::VertexAttribute vertexAttributes[], GLint first, GLsizei count, GLsizei instances) +{ + TranslatedAttribute attributes[gl::MAX_VERTEX_ATTRIBS]; + GLenum err = mVertexDataManager->prepareVertexData(vertexAttributes, programBinary, first, count, attributes, instances); + if (err != GL_NO_ERROR) + { + return err; + } + + return mInputLayoutCache.applyVertexBuffers(attributes, programBinary); +} + +GLenum Renderer11::applyIndexBuffer(const GLvoid *indices, gl::Buffer *elementArrayBuffer, GLsizei count, GLenum mode, GLenum type, TranslatedIndexData *indexInfo) +{ + GLenum err = mIndexDataManager->prepareIndexData(type, count, elementArrayBuffer, indices, indexInfo); + + if (err == GL_NO_ERROR) + { + if (indexInfo->storage) + { + if (indexInfo->serial != mAppliedStorageIBSerial || indexInfo->startOffset != mAppliedIBOffset) + { + BufferStorage11 *storage = BufferStorage11::makeBufferStorage11(indexInfo->storage); + IndexBuffer11* indexBuffer = IndexBuffer11::makeIndexBuffer11(indexInfo->indexBuffer); + + mDeviceContext->IASetIndexBuffer(storage->getBuffer(), indexBuffer->getIndexFormat(), indexInfo->startOffset); + + mAppliedIBSerial = 0; + mAppliedStorageIBSerial = storage->getSerial(); + mAppliedIBOffset = indexInfo->startOffset; + } + } + else if (indexInfo->serial != mAppliedIBSerial || indexInfo->startOffset != mAppliedIBOffset) + { + IndexBuffer11* indexBuffer = IndexBuffer11::makeIndexBuffer11(indexInfo->indexBuffer); + + mDeviceContext->IASetIndexBuffer(indexBuffer->getBuffer(), indexBuffer->getIndexFormat(), indexInfo->startOffset); + + mAppliedIBSerial = indexInfo->serial; + mAppliedStorageIBSerial = 0; + mAppliedIBOffset = indexInfo->startOffset; + } + } + + return err; +} + +void Renderer11::drawArrays(GLenum mode, GLsizei count, GLsizei instances) +{ + if (mode == GL_LINE_LOOP) + { + drawLineLoop(count, GL_NONE, NULL, 0, NULL); + } + else if (mode == GL_TRIANGLE_FAN) + { + drawTriangleFan(count, GL_NONE, NULL, 0, NULL, instances); + } + else if (instances > 0) + { + mDeviceContext->DrawInstanced(count, instances, 0, 0); + } + else + { + mDeviceContext->Draw(count, 0); + } +} + +void Renderer11::drawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices, gl::Buffer *elementArrayBuffer, const TranslatedIndexData &indexInfo, GLsizei instances) +{ + if (mode == GL_LINE_LOOP) + { + drawLineLoop(count, type, indices, indexInfo.minIndex, elementArrayBuffer); + } + else if (mode == GL_TRIANGLE_FAN) + { + drawTriangleFan(count, type, indices, indexInfo.minIndex, elementArrayBuffer, instances); + } + else if (instances > 0) + { + mDeviceContext->DrawIndexedInstanced(count, instances, 0, -static_cast<int>(indexInfo.minIndex), 0); + } + else + { + mDeviceContext->DrawIndexed(count, 0, -static_cast<int>(indexInfo.minIndex)); + } +} + +void Renderer11::drawLineLoop(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer) +{ + // Get the raw indices for an indexed draw + if (type != GL_NONE && elementArrayBuffer) + { + gl::Buffer *indexBuffer = elementArrayBuffer; + BufferStorage *storage = indexBuffer->getStorage(); + intptr_t offset = reinterpret_cast<intptr_t>(indices); + indices = static_cast<const GLubyte*>(storage->getData()) + offset; + } + + if (!mLineLoopIB) + { + mLineLoopIB = new StreamingIndexBufferInterface(this); + if (!mLineLoopIB->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_INT)) + { + delete mLineLoopIB; + mLineLoopIB = NULL; + + ERR("Could not create a 32-bit looping index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + } + + const int spaceNeeded = (count + 1) * sizeof(unsigned int); + if (!mLineLoopIB->reserveBufferSpace(spaceNeeded, GL_UNSIGNED_INT)) + { + ERR("Could not reserve enough space in looping index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + + void* mappedMemory = NULL; + int offset = mLineLoopIB->mapBuffer(spaceNeeded, &mappedMemory); + if (offset == -1 || mappedMemory == NULL) + { + ERR("Could not map index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + + unsigned int *data = reinterpret_cast<unsigned int*>(mappedMemory); + unsigned int indexBufferOffset = static_cast<unsigned int>(offset); + + switch (type) + { + case GL_NONE: // Non-indexed draw + for (int i = 0; i < count; i++) + { + data[i] = i; + } + data[count] = 0; + break; + case GL_UNSIGNED_BYTE: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLubyte*>(indices)[i]; + } + data[count] = static_cast<const GLubyte*>(indices)[0]; + break; + case GL_UNSIGNED_SHORT: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLushort*>(indices)[i]; + } + data[count] = static_cast<const GLushort*>(indices)[0]; + break; + case GL_UNSIGNED_INT: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLuint*>(indices)[i]; + } + data[count] = static_cast<const GLuint*>(indices)[0]; + break; + default: UNREACHABLE(); + } + + if (!mLineLoopIB->unmapBuffer()) + { + ERR("Could not unmap index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + + if (mAppliedIBSerial != mLineLoopIB->getSerial() || mAppliedIBOffset != indexBufferOffset) + { + IndexBuffer11 *indexBuffer = IndexBuffer11::makeIndexBuffer11(mLineLoopIB->getIndexBuffer()); + + mDeviceContext->IASetIndexBuffer(indexBuffer->getBuffer(), indexBuffer->getIndexFormat(), indexBufferOffset); + mAppliedIBSerial = mLineLoopIB->getSerial(); + mAppliedStorageIBSerial = 0; + mAppliedIBOffset = indexBufferOffset; + } + + mDeviceContext->DrawIndexed(count + 1, 0, -minIndex); +} + +void Renderer11::drawTriangleFan(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer, int instances) +{ + // Get the raw indices for an indexed draw + if (type != GL_NONE && elementArrayBuffer) + { + gl::Buffer *indexBuffer = elementArrayBuffer; + BufferStorage *storage = indexBuffer->getStorage(); + intptr_t offset = reinterpret_cast<intptr_t>(indices); + indices = static_cast<const GLubyte*>(storage->getData()) + offset; + } + + if (!mTriangleFanIB) + { + mTriangleFanIB = new StreamingIndexBufferInterface(this); + if (!mTriangleFanIB->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_INT)) + { + delete mTriangleFanIB; + mTriangleFanIB = NULL; + + ERR("Could not create a scratch index buffer for GL_TRIANGLE_FAN."); + return gl::error(GL_OUT_OF_MEMORY); + } + } + + const int numTris = count - 2; + const int spaceNeeded = (numTris * 3) * sizeof(unsigned int); + if (!mTriangleFanIB->reserveBufferSpace(spaceNeeded, GL_UNSIGNED_INT)) + { + ERR("Could not reserve enough space in scratch index buffer for GL_TRIANGLE_FAN."); + return gl::error(GL_OUT_OF_MEMORY); + } + + void* mappedMemory = NULL; + int offset = mTriangleFanIB->mapBuffer(spaceNeeded, &mappedMemory); + if (offset == -1 || mappedMemory == NULL) + { + ERR("Could not map scratch index buffer for GL_TRIANGLE_FAN."); + return gl::error(GL_OUT_OF_MEMORY); + } + + unsigned int *data = reinterpret_cast<unsigned int*>(mappedMemory); + unsigned int indexBufferOffset = static_cast<unsigned int>(offset); + + switch (type) + { + case GL_NONE: // Non-indexed draw + for (int i = 0; i < numTris; i++) + { + data[i*3 + 0] = 0; + data[i*3 + 1] = i + 1; + data[i*3 + 2] = i + 2; + } + break; + case GL_UNSIGNED_BYTE: + for (int i = 0; i < numTris; i++) + { + data[i*3 + 0] = static_cast<const GLubyte*>(indices)[0]; + data[i*3 + 1] = static_cast<const GLubyte*>(indices)[i + 1]; + data[i*3 + 2] = static_cast<const GLubyte*>(indices)[i + 2]; + } + break; + case GL_UNSIGNED_SHORT: + for (int i = 0; i < numTris; i++) + { + data[i*3 + 0] = static_cast<const GLushort*>(indices)[0]; + data[i*3 + 1] = static_cast<const GLushort*>(indices)[i + 1]; + data[i*3 + 2] = static_cast<const GLushort*>(indices)[i + 2]; + } + break; + case GL_UNSIGNED_INT: + for (int i = 0; i < numTris; i++) + { + data[i*3 + 0] = static_cast<const GLuint*>(indices)[0]; + data[i*3 + 1] = static_cast<const GLuint*>(indices)[i + 1]; + data[i*3 + 2] = static_cast<const GLuint*>(indices)[i + 2]; + } + break; + default: UNREACHABLE(); + } + + if (!mTriangleFanIB->unmapBuffer()) + { + ERR("Could not unmap scratch index buffer for GL_TRIANGLE_FAN."); + return gl::error(GL_OUT_OF_MEMORY); + } + + if (mAppliedIBSerial != mTriangleFanIB->getSerial() || mAppliedIBOffset != indexBufferOffset) + { + IndexBuffer11 *indexBuffer = IndexBuffer11::makeIndexBuffer11(mTriangleFanIB->getIndexBuffer()); + + mDeviceContext->IASetIndexBuffer(indexBuffer->getBuffer(), indexBuffer->getIndexFormat(), indexBufferOffset); + mAppliedIBSerial = mTriangleFanIB->getSerial(); + mAppliedStorageIBSerial = 0; + mAppliedIBOffset = indexBufferOffset; + } + + if (instances > 0) + { + mDeviceContext->DrawIndexedInstanced(numTris * 3, instances, 0, -minIndex, 0); + } + else + { + mDeviceContext->DrawIndexed(numTris * 3, 0, -minIndex); + } +} + +void Renderer11::applyShaders(gl::ProgramBinary *programBinary) +{ + unsigned int programBinarySerial = programBinary->getSerial(); + const bool updateProgramState = (programBinarySerial != mAppliedProgramBinarySerial); + + if (updateProgramState) + { + ShaderExecutable *vertexExe = programBinary->getVertexExecutable(); + ShaderExecutable *pixelExe = programBinary->getPixelExecutable(); + + ID3D11VertexShader *vertexShader = NULL; + if (vertexExe) vertexShader = ShaderExecutable11::makeShaderExecutable11(vertexExe)->getVertexShader(); + + ID3D11PixelShader *pixelShader = NULL; + if (pixelExe) pixelShader = ShaderExecutable11::makeShaderExecutable11(pixelExe)->getPixelShader(); + + mDeviceContext->PSSetShader(pixelShader, NULL, 0); + mDeviceContext->VSSetShader(vertexShader, NULL, 0); + + programBinary->dirtyAllUniforms(); + + mAppliedProgramBinarySerial = programBinarySerial; + } + + // Only use the geometry shader currently for point sprite drawing + const bool usesGeometryShader = (programBinary->usesGeometryShader() && mCurRasterState.pointDrawMode); + + if (updateProgramState || usesGeometryShader != mIsGeometryShaderActive) + { + if (usesGeometryShader) + { + ShaderExecutable *geometryExe = programBinary->getGeometryExecutable(); + ID3D11GeometryShader *geometryShader = ShaderExecutable11::makeShaderExecutable11(geometryExe)->getGeometryShader(); + mDeviceContext->GSSetShader(geometryShader, NULL, 0); + } + else + { + mDeviceContext->GSSetShader(NULL, NULL, 0); + } + + mIsGeometryShaderActive = usesGeometryShader; + } +} + +void Renderer11::applyUniforms(gl::ProgramBinary *programBinary, gl::UniformArray *uniformArray) +{ + ShaderExecutable11 *vertexExecutable = ShaderExecutable11::makeShaderExecutable11(programBinary->getVertexExecutable()); + ShaderExecutable11 *pixelExecutable = ShaderExecutable11::makeShaderExecutable11(programBinary->getPixelExecutable()); + + unsigned int totalRegisterCountVS = 0; + unsigned int totalRegisterCountPS = 0; + + bool vertexUniformsDirty = false; + bool pixelUniformsDirty = false; + + for (gl::UniformArray::const_iterator uniform_iterator = uniformArray->begin(); uniform_iterator != uniformArray->end(); uniform_iterator++) + { + const gl::Uniform *uniform = *uniform_iterator; + + if (uniform->vsRegisterIndex >= 0) + { + totalRegisterCountVS += uniform->registerCount; + vertexUniformsDirty = vertexUniformsDirty || uniform->dirty; + } + + if (uniform->psRegisterIndex >= 0) + { + totalRegisterCountPS += uniform->registerCount; + pixelUniformsDirty = pixelUniformsDirty || uniform->dirty; + } + } + + ID3D11Buffer *vertexConstantBuffer = vertexExecutable->getConstantBuffer(mDevice, totalRegisterCountVS); + ID3D11Buffer *pixelConstantBuffer = pixelExecutable->getConstantBuffer(mDevice, totalRegisterCountPS); + + float (*mapVS)[4] = NULL; + float (*mapPS)[4] = NULL; + + if (totalRegisterCountVS > 0 && vertexUniformsDirty) + { + D3D11_MAPPED_SUBRESOURCE map = {0}; + HRESULT result = mDeviceContext->Map(vertexConstantBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &map); + ASSERT(SUCCEEDED(result)); + mapVS = (float(*)[4])map.pData; + } + + if (totalRegisterCountPS > 0 && pixelUniformsDirty) + { + D3D11_MAPPED_SUBRESOURCE map = {0}; + HRESULT result = mDeviceContext->Map(pixelConstantBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &map); + ASSERT(SUCCEEDED(result)); + mapPS = (float(*)[4])map.pData; + } + + for (gl::UniformArray::iterator uniform_iterator = uniformArray->begin(); uniform_iterator != uniformArray->end(); uniform_iterator++) + { + gl::Uniform *uniform = *uniform_iterator; + + if (uniform->type != GL_SAMPLER_2D && uniform->type != GL_SAMPLER_CUBE) + { + if (uniform->vsRegisterIndex >= 0 && mapVS) + { + memcpy(mapVS + uniform->vsRegisterIndex, uniform->data, uniform->registerCount * sizeof(float[4])); + } + + if (uniform->psRegisterIndex >= 0 && mapPS) + { + memcpy(mapPS + uniform->psRegisterIndex, uniform->data, uniform->registerCount * sizeof(float[4])); + } + } + + uniform->dirty = false; + } + + if (mapVS) + { + mDeviceContext->Unmap(vertexConstantBuffer, 0); + } + + if (mapPS) + { + mDeviceContext->Unmap(pixelConstantBuffer, 0); + } + + mDeviceContext->VSSetConstantBuffers(0, 1, &vertexConstantBuffer); + mDeviceContext->PSSetConstantBuffers(0, 1, &pixelConstantBuffer); + + // Driver uniforms + if (!mDriverConstantBufferVS) + { + D3D11_BUFFER_DESC constantBufferDescription = {0}; + constantBufferDescription.ByteWidth = sizeof(dx_VertexConstants); + constantBufferDescription.Usage = D3D11_USAGE_DEFAULT; + constantBufferDescription.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + constantBufferDescription.CPUAccessFlags = 0; + constantBufferDescription.MiscFlags = 0; + constantBufferDescription.StructureByteStride = 0; + + HRESULT result = mDevice->CreateBuffer(&constantBufferDescription, NULL, &mDriverConstantBufferVS); + ASSERT(SUCCEEDED(result)); + + mDeviceContext->VSSetConstantBuffers(1, 1, &mDriverConstantBufferVS); + } + + if (!mDriverConstantBufferPS) + { + D3D11_BUFFER_DESC constantBufferDescription = {0}; + constantBufferDescription.ByteWidth = sizeof(dx_PixelConstants); + constantBufferDescription.Usage = D3D11_USAGE_DEFAULT; + constantBufferDescription.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + constantBufferDescription.CPUAccessFlags = 0; + constantBufferDescription.MiscFlags = 0; + constantBufferDescription.StructureByteStride = 0; + + HRESULT result = mDevice->CreateBuffer(&constantBufferDescription, NULL, &mDriverConstantBufferPS); + ASSERT(SUCCEEDED(result)); + + mDeviceContext->PSSetConstantBuffers(1, 1, &mDriverConstantBufferPS); + } + + if (memcmp(&mVertexConstants, &mAppliedVertexConstants, sizeof(dx_VertexConstants)) != 0) + { + mDeviceContext->UpdateSubresource(mDriverConstantBufferVS, 0, NULL, &mVertexConstants, 16, 0); + memcpy(&mAppliedVertexConstants, &mVertexConstants, sizeof(dx_VertexConstants)); + } + + if (memcmp(&mPixelConstants, &mAppliedPixelConstants, sizeof(dx_PixelConstants)) != 0) + { + mDeviceContext->UpdateSubresource(mDriverConstantBufferPS, 0, NULL, &mPixelConstants, 16, 0); + memcpy(&mAppliedPixelConstants, &mPixelConstants, sizeof(dx_PixelConstants)); + } + + // needed for the point sprite geometry shader + mDeviceContext->GSSetConstantBuffers(0, 1, &mDriverConstantBufferPS); +} + +void Renderer11::clear(const gl::ClearParameters &clearParams, gl::Framebuffer *frameBuffer) +{ + bool alphaUnmasked = (gl::GetAlphaSize(mRenderTargetDesc.format) == 0) || clearParams.colorMaskAlpha; + bool needMaskedColorClear = (clearParams.mask & GL_COLOR_BUFFER_BIT) && + !(clearParams.colorMaskRed && clearParams.colorMaskGreen && + clearParams.colorMaskBlue && alphaUnmasked); + + unsigned int stencilUnmasked = 0x0; + if (frameBuffer->hasStencil()) + { + unsigned int stencilSize = gl::GetStencilSize(frameBuffer->getStencilbuffer()->getActualFormat()); + stencilUnmasked = (0x1 << stencilSize) - 1; + } + bool needMaskedStencilClear = (clearParams.mask & GL_STENCIL_BUFFER_BIT) && + (clearParams.stencilWriteMask & stencilUnmasked) != stencilUnmasked; + + bool needScissoredClear = mScissorEnabled && (mCurScissor.x > 0 || mCurScissor.y > 0 || + mCurScissor.x + mCurScissor.width < mRenderTargetDesc.width || + mCurScissor.y + mCurScissor.height < mRenderTargetDesc.height); + + if (needMaskedColorClear || needMaskedStencilClear || needScissoredClear) + { + maskedClear(clearParams); + } + else + { + if (clearParams.mask & GL_COLOR_BUFFER_BIT) + { + for (unsigned int colorAttachment = 0; colorAttachment < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; colorAttachment++) + { + if (frameBuffer->isEnabledColorAttachment(colorAttachment)) + { + gl::Renderbuffer *renderbufferObject = frameBuffer->getColorbuffer(colorAttachment); + if (renderbufferObject) + { + RenderTarget11 *renderTarget = RenderTarget11::makeRenderTarget11(renderbufferObject->getRenderTarget()); + if (!renderTarget) + { + ERR("render target pointer unexpectedly null."); + return; + } + + ID3D11RenderTargetView *framebufferRTV = renderTarget->getRenderTargetView(); + if (!framebufferRTV) + { + ERR("render target view pointer unexpectedly null."); + return; + } + + const float clearValues[4] = { clearParams.colorClearValue.red, + clearParams.colorClearValue.green, + clearParams.colorClearValue.blue, + clearParams.colorClearValue.alpha }; + mDeviceContext->ClearRenderTargetView(framebufferRTV, clearValues); + + framebufferRTV->Release(); + } + } + } + } + if (clearParams.mask & GL_DEPTH_BUFFER_BIT || clearParams.mask & GL_STENCIL_BUFFER_BIT) + { + gl::Renderbuffer *renderbufferObject = frameBuffer->getDepthOrStencilbuffer(); + if (renderbufferObject) + { + RenderTarget11 *renderTarget = RenderTarget11::makeRenderTarget11(renderbufferObject->getDepthStencil()); + if (!renderTarget) + { + ERR("render target pointer unexpectedly null."); + return; + } + + ID3D11DepthStencilView *framebufferDSV = renderTarget->getDepthStencilView(); + if (!framebufferDSV) + { + ERR("depth stencil view pointer unexpectedly null."); + return; + } + + UINT clearFlags = 0; + if (clearParams.mask & GL_DEPTH_BUFFER_BIT) + { + clearFlags |= D3D11_CLEAR_DEPTH; + } + if (clearParams.mask & GL_STENCIL_BUFFER_BIT) + { + clearFlags |= D3D11_CLEAR_STENCIL; + } + + float depthClear = gl::clamp01(clearParams.depthClearValue); + UINT8 stencilClear = clearParams.stencilClearValue & 0x000000FF; + + mDeviceContext->ClearDepthStencilView(framebufferDSV, clearFlags, depthClear, stencilClear); + + framebufferDSV->Release(); + } + } + } +} + +void Renderer11::maskedClear(const gl::ClearParameters &clearParams) +{ + HRESULT result; + + if (!mClearResourcesInitialized) + { + ASSERT(!mClearVB && !mClearVS && !mClearPS && !mClearScissorRS && !mClearNoScissorRS); + + D3D11_BUFFER_DESC vbDesc; + vbDesc.ByteWidth = sizeof(d3d11::PositionDepthColorVertex) * 4; + vbDesc.Usage = D3D11_USAGE_DYNAMIC; + vbDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + vbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + vbDesc.MiscFlags = 0; + vbDesc.StructureByteStride = 0; + + result = mDevice->CreateBuffer(&vbDesc, NULL, &mClearVB); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mClearVB, "Renderer11 masked clear vertex buffer"); + + D3D11_INPUT_ELEMENT_DESC quadLayout[] = + { + { "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + { "COLOR", 0, DXGI_FORMAT_R32G32B32A32_FLOAT, 0, 12, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + }; + + result = mDevice->CreateInputLayout(quadLayout, 2, g_VS_Clear, sizeof(g_VS_Clear), &mClearIL); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mClearIL, "Renderer11 masked clear input layout"); + + result = mDevice->CreateVertexShader(g_VS_Clear, sizeof(g_VS_Clear), NULL, &mClearVS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mClearVS, "Renderer11 masked clear vertex shader"); + + result = mDevice->CreatePixelShader(g_PS_Clear, sizeof(g_PS_Clear), NULL, &mClearPS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mClearPS, "Renderer11 masked clear pixel shader"); + + D3D11_RASTERIZER_DESC rsScissorDesc; + rsScissorDesc.FillMode = D3D11_FILL_SOLID; + rsScissorDesc.CullMode = D3D11_CULL_NONE; + rsScissorDesc.FrontCounterClockwise = FALSE; + rsScissorDesc.DepthBias = 0; + rsScissorDesc.DepthBiasClamp = 0.0f; + rsScissorDesc.SlopeScaledDepthBias = 0.0f; + rsScissorDesc.DepthClipEnable = FALSE; + rsScissorDesc.ScissorEnable = TRUE; + rsScissorDesc.MultisampleEnable = FALSE; + rsScissorDesc.AntialiasedLineEnable = FALSE; + + result = mDevice->CreateRasterizerState(&rsScissorDesc, &mClearScissorRS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mClearScissorRS, "Renderer11 masked clear scissor rasterizer state"); + + D3D11_RASTERIZER_DESC rsNoScissorDesc; + rsNoScissorDesc.FillMode = D3D11_FILL_SOLID; + rsNoScissorDesc.CullMode = D3D11_CULL_NONE; + rsNoScissorDesc.FrontCounterClockwise = FALSE; + rsNoScissorDesc.DepthBias = 0; + rsNoScissorDesc.DepthBiasClamp = 0.0f; + rsNoScissorDesc.SlopeScaledDepthBias = 0.0f; + rsNoScissorDesc.DepthClipEnable = FALSE; + rsNoScissorDesc.ScissorEnable = FALSE; + rsNoScissorDesc.MultisampleEnable = FALSE; + rsNoScissorDesc.AntialiasedLineEnable = FALSE; + + result = mDevice->CreateRasterizerState(&rsNoScissorDesc, &mClearNoScissorRS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mClearNoScissorRS, "Renderer11 masked clear no scissor rasterizer state"); + + mClearResourcesInitialized = true; + } + + // Prepare the depth stencil state to write depth values if the depth should be cleared + // and stencil values if the stencil should be cleared + gl::DepthStencilState glDSState; + glDSState.depthTest = (clearParams.mask & GL_DEPTH_BUFFER_BIT) != 0; + glDSState.depthFunc = GL_ALWAYS; + glDSState.depthMask = (clearParams.mask & GL_DEPTH_BUFFER_BIT) != 0; + glDSState.stencilTest = (clearParams.mask & GL_STENCIL_BUFFER_BIT) != 0; + glDSState.stencilFunc = GL_ALWAYS; + glDSState.stencilMask = 0; + glDSState.stencilFail = GL_REPLACE; + glDSState.stencilPassDepthFail = GL_REPLACE; + glDSState.stencilPassDepthPass = GL_REPLACE; + glDSState.stencilWritemask = clearParams.stencilWriteMask; + glDSState.stencilBackFunc = GL_ALWAYS; + glDSState.stencilBackMask = 0; + glDSState.stencilBackFail = GL_REPLACE; + glDSState.stencilBackPassDepthFail = GL_REPLACE; + glDSState.stencilBackPassDepthPass = GL_REPLACE; + glDSState.stencilBackWritemask = clearParams.stencilWriteMask; + + int stencilClear = clearParams.stencilClearValue & 0x000000FF; + + ID3D11DepthStencilState *dsState = mStateCache.getDepthStencilState(glDSState); + + // Prepare the blend state to use a write mask if the color buffer should be cleared + gl::BlendState glBlendState; + glBlendState.blend = false; + glBlendState.sourceBlendRGB = GL_ONE; + glBlendState.destBlendRGB = GL_ZERO; + glBlendState.sourceBlendAlpha = GL_ONE; + glBlendState.destBlendAlpha = GL_ZERO; + glBlendState.blendEquationRGB = GL_FUNC_ADD; + glBlendState.blendEquationAlpha = GL_FUNC_ADD; + glBlendState.colorMaskRed = (clearParams.mask & GL_COLOR_BUFFER_BIT) ? clearParams.colorMaskRed : false; + glBlendState.colorMaskGreen = (clearParams.mask & GL_COLOR_BUFFER_BIT) ? clearParams.colorMaskGreen : false; + glBlendState.colorMaskBlue = (clearParams.mask & GL_COLOR_BUFFER_BIT) ? clearParams.colorMaskBlue : false; + glBlendState.colorMaskAlpha = (clearParams.mask & GL_COLOR_BUFFER_BIT) ? clearParams.colorMaskAlpha : false; + glBlendState.sampleAlphaToCoverage = false; + glBlendState.dither = false; + + static const float blendFactors[4] = { 1.0f, 1.0f, 1.0f, 1.0f }; + static const UINT sampleMask = 0xFFFFFFFF; + + ID3D11BlendState *blendState = mStateCache.getBlendState(glBlendState); + + // Set the vertices + D3D11_MAPPED_SUBRESOURCE mappedResource; + result = mDeviceContext->Map(mClearVB, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + ERR("Failed to map masked clear vertex buffer, HRESULT: 0x%X.", result); + return; + } + + d3d11::PositionDepthColorVertex *vertices = reinterpret_cast<d3d11::PositionDepthColorVertex*>(mappedResource.pData); + + float depthClear = gl::clamp01(clearParams.depthClearValue); + d3d11::SetPositionDepthColorVertex(&vertices[0], -1.0f, 1.0f, depthClear, clearParams.colorClearValue); + d3d11::SetPositionDepthColorVertex(&vertices[1], -1.0f, -1.0f, depthClear, clearParams.colorClearValue); + d3d11::SetPositionDepthColorVertex(&vertices[2], 1.0f, 1.0f, depthClear, clearParams.colorClearValue); + d3d11::SetPositionDepthColorVertex(&vertices[3], 1.0f, -1.0f, depthClear, clearParams.colorClearValue); + + mDeviceContext->Unmap(mClearVB, 0); + + // Apply state + mDeviceContext->OMSetBlendState(blendState, blendFactors, sampleMask); + mDeviceContext->OMSetDepthStencilState(dsState, stencilClear); + mDeviceContext->RSSetState(mScissorEnabled ? mClearScissorRS : mClearNoScissorRS); + + // Apply shaders + mDeviceContext->IASetInputLayout(mClearIL); + mDeviceContext->VSSetShader(mClearVS, NULL, 0); + mDeviceContext->PSSetShader(mClearPS, NULL, 0); + mDeviceContext->GSSetShader(NULL, NULL, 0); + + // Apply vertex buffer + static UINT stride = sizeof(d3d11::PositionDepthColorVertex); + static UINT startIdx = 0; + mDeviceContext->IASetVertexBuffers(0, 1, &mClearVB, &stride, &startIdx); + mDeviceContext->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP); + + // Draw the clear quad + mDeviceContext->Draw(4, 0); + + // Clean up + markAllStateDirty(); +} + +void Renderer11::markAllStateDirty() +{ + for (unsigned int rtIndex = 0; rtIndex < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; rtIndex++) + { + mAppliedRenderTargetSerials[rtIndex] = 0; + } + mAppliedDepthbufferSerial = 0; + mAppliedStencilbufferSerial = 0; + mDepthStencilInitialized = false; + mRenderTargetDescInitialized = false; + + for (int i = 0; i < gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS; i++) + { + mForceSetVertexSamplerStates[i] = true; + mCurVertexTextureSerials[i] = 0; + } + for (int i = 0; i < gl::MAX_TEXTURE_IMAGE_UNITS; i++) + { + mForceSetPixelSamplerStates[i] = true; + mCurPixelTextureSerials[i] = 0; + } + + mForceSetBlendState = true; + mForceSetRasterState = true; + mForceSetDepthStencilState = true; + mForceSetScissor = true; + mForceSetViewport = true; + + mAppliedIBSerial = 0; + mAppliedStorageIBSerial = 0; + mAppliedIBOffset = 0; + + mAppliedProgramBinarySerial = 0; + memset(&mAppliedVertexConstants, 0, sizeof(dx_VertexConstants)); + memset(&mAppliedPixelConstants, 0, sizeof(dx_PixelConstants)); +} + +void Renderer11::releaseDeviceResources() +{ + mStateCache.clear(); + mInputLayoutCache.clear(); + + delete mVertexDataManager; + mVertexDataManager = NULL; + + delete mIndexDataManager; + mIndexDataManager = NULL; + + delete mLineLoopIB; + mLineLoopIB = NULL; + + delete mTriangleFanIB; + mTriangleFanIB = NULL; + + if (mCopyVB) + { + mCopyVB->Release(); + mCopyVB = NULL; + } + + if (mCopySampler) + { + mCopySampler->Release(); + mCopySampler = NULL; + } + + if (mCopyIL) + { + mCopyIL->Release(); + mCopyIL = NULL; + } + + if (mCopyVS) + { + mCopyVS->Release(); + mCopyVS = NULL; + } + + if (mCopyRGBAPS) + { + mCopyRGBAPS->Release(); + mCopyRGBAPS = NULL; + } + + if (mCopyRGBPS) + { + mCopyRGBPS->Release(); + mCopyRGBPS = NULL; + } + + if (mCopyLumPS) + { + mCopyLumPS->Release(); + mCopyLumPS = NULL; + } + + if (mCopyLumAlphaPS) + { + mCopyLumAlphaPS->Release(); + mCopyLumAlphaPS = NULL; + } + + mCopyResourcesInitialized = false; + + if (mClearVB) + { + mClearVB->Release(); + mClearVB = NULL; + } + + if (mClearIL) + { + mClearIL->Release(); + mClearIL = NULL; + } + + if (mClearVS) + { + mClearVS->Release(); + mClearVS = NULL; + } + + if (mClearPS) + { + mClearPS->Release(); + mClearPS = NULL; + } + + if (mClearScissorRS) + { + mClearScissorRS->Release(); + mClearScissorRS = NULL; + } + + if (mClearNoScissorRS) + { + mClearNoScissorRS->Release(); + mClearNoScissorRS = NULL; + } + + mClearResourcesInitialized = false; + + if (mDriverConstantBufferVS) + { + mDriverConstantBufferVS->Release(); + mDriverConstantBufferVS = NULL; + } + + if (mDriverConstantBufferPS) + { + mDriverConstantBufferPS->Release(); + mDriverConstantBufferPS = NULL; + } + + if (mSyncQuery) + { + mSyncQuery->Release(); + mSyncQuery = NULL; + } +} + +void Renderer11::notifyDeviceLost() +{ + mDeviceLost = true; + mDisplay->notifyDeviceLost(); +} + +bool Renderer11::isDeviceLost() +{ + return mDeviceLost; +} + +// set notify to true to broadcast a message to all contexts of the device loss +bool Renderer11::testDeviceLost(bool notify) +{ + bool isLost = false; + + // GetRemovedReason is used to test if the device is removed + HRESULT result = mDevice->GetDeviceRemovedReason(); + isLost = d3d11::isDeviceLostError(result); + + if (isLost) + { + // Log error if this is a new device lost event + if (mDeviceLost == false) + { + ERR("The D3D11 device was removed: 0x%08X", result); + } + + // ensure we note the device loss -- + // we'll probably get this done again by notifyDeviceLost + // but best to remember it! + // Note that we don't want to clear the device loss status here + // -- this needs to be done by resetDevice + mDeviceLost = true; + if (notify) + { + notifyDeviceLost(); + } + } + + return isLost; +} + +bool Renderer11::testDeviceResettable() +{ + // determine if the device is resettable by creating a dummy device + PFN_D3D11_CREATE_DEVICE D3D11CreateDevice = (PFN_D3D11_CREATE_DEVICE)GetProcAddress(mD3d11Module, "D3D11CreateDevice"); + + if (D3D11CreateDevice == NULL) + { + return false; + } + + D3D_FEATURE_LEVEL featureLevels[] = + { + D3D_FEATURE_LEVEL_11_0, + D3D_FEATURE_LEVEL_10_1, + D3D_FEATURE_LEVEL_10_0, + }; + + ID3D11Device* dummyDevice; + D3D_FEATURE_LEVEL dummyFeatureLevel; + ID3D11DeviceContext* dummyContext; + + HRESULT result = D3D11CreateDevice(NULL, + D3D_DRIVER_TYPE_HARDWARE, + NULL, + #if defined(_DEBUG) + D3D11_CREATE_DEVICE_DEBUG, + #else + 0, + #endif + featureLevels, + ArraySize(featureLevels), + D3D11_SDK_VERSION, + &dummyDevice, + &dummyFeatureLevel, + &dummyContext); + + if (!mDevice || FAILED(result)) + { + return false; + } + + dummyContext->Release(); + dummyDevice->Release(); + + return true; +} + +void Renderer11::release() +{ + releaseDeviceResources(); + + if (mDxgiFactory) + { + mDxgiFactory->Release(); + mDxgiFactory = NULL; + } + + if (mDxgiAdapter) + { + mDxgiAdapter->Release(); + mDxgiAdapter = NULL; + } + + if (mDeviceContext) + { + mDeviceContext->ClearState(); + mDeviceContext->Flush(); + mDeviceContext->Release(); + mDeviceContext = NULL; + } + + if (mDevice) + { + mDevice->Release(); + mDevice = NULL; + } + + if (mD3d11Module) + { + FreeLibrary(mD3d11Module); + mD3d11Module = NULL; + } + + if (mDxgiModule) + { + FreeLibrary(mDxgiModule); + mDxgiModule = NULL; + } +} + +bool Renderer11::resetDevice() +{ + // recreate everything + release(); + EGLint result = initialize(); + + if (result != EGL_SUCCESS) + { + ERR("Could not reinitialize D3D11 device: %08X", result); + return false; + } + + mDeviceLost = false; + + return true; +} + +DWORD Renderer11::getAdapterVendor() const +{ + return mAdapterDescription.VendorId; +} + +std::string Renderer11::getRendererDescription() const +{ + std::ostringstream rendererString; + + rendererString << mDescription; + rendererString << " Direct3D11"; + + rendererString << " vs_" << getMajorShaderModel() << "_" << getMinorShaderModel(); + rendererString << " ps_" << getMajorShaderModel() << "_" << getMinorShaderModel(); + + return rendererString.str(); +} + +GUID Renderer11::getAdapterIdentifier() const +{ + // Use the adapter LUID as our adapter ID + // This number is local to a machine is only guaranteed to be unique between restarts + META_ASSERT(sizeof(LUID) <= sizeof(GUID)); + GUID adapterId = {0}; + memcpy(&adapterId, &mAdapterDescription.AdapterLuid, sizeof(LUID)); + return adapterId; +} + +bool Renderer11::getBGRATextureSupport() const +{ + return mBGRATextureSupport; +} + +bool Renderer11::getDXT1TextureSupport() +{ + return mDXT1TextureSupport; +} + +bool Renderer11::getDXT3TextureSupport() +{ + return mDXT3TextureSupport; +} + +bool Renderer11::getDXT5TextureSupport() +{ + return mDXT5TextureSupport; +} + +bool Renderer11::getDepthTextureSupport() const +{ + return mDepthTextureSupport; +} + +bool Renderer11::getFloat32TextureSupport(bool *filtering, bool *renderable) +{ + *renderable = mFloat32RenderSupport; + *filtering = mFloat32FilterSupport; + return mFloat32TextureSupport; +} + +bool Renderer11::getFloat16TextureSupport(bool *filtering, bool *renderable) +{ + *renderable = mFloat16RenderSupport; + *filtering = mFloat16FilterSupport; + return mFloat16TextureSupport; +} + +bool Renderer11::getLuminanceTextureSupport() +{ + return false; +} + +bool Renderer11::getLuminanceAlphaTextureSupport() +{ + return false; +} + +bool Renderer11::getTextureFilterAnisotropySupport() const +{ + return true; +} + +float Renderer11::getTextureMaxAnisotropy() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + return D3D11_MAX_MAXANISOTROPY; + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return D3D10_MAX_MAXANISOTROPY; + default: UNREACHABLE(); + return 0; + } +} + +bool Renderer11::getEventQuerySupport() +{ + return true; +} + +Range Renderer11::getViewportBounds() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + return Range(D3D11_VIEWPORT_BOUNDS_MIN, D3D11_VIEWPORT_BOUNDS_MAX); + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return Range(D3D10_VIEWPORT_BOUNDS_MIN, D3D10_VIEWPORT_BOUNDS_MAX); + default: UNREACHABLE(); + return Range(0, 0); + } +} + +unsigned int Renderer11::getMaxVertexTextureImageUnits() const +{ + META_ASSERT(MAX_TEXTURE_IMAGE_UNITS_VTF_SM4 <= gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS); + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return MAX_TEXTURE_IMAGE_UNITS_VTF_SM4; + default: UNREACHABLE(); + return 0; + } +} + +unsigned int Renderer11::getMaxCombinedTextureImageUnits() const +{ + return gl::MAX_TEXTURE_IMAGE_UNITS + getMaxVertexTextureImageUnits(); +} + +unsigned int Renderer11::getReservedVertexUniformVectors() const +{ + return 0; // Driver uniforms are stored in a separate constant buffer +} + +unsigned int Renderer11::getReservedFragmentUniformVectors() const +{ + return 0; // Driver uniforms are stored in a separate constant buffer +} + +unsigned int Renderer11::getMaxVertexUniformVectors() const +{ + META_ASSERT(MAX_VERTEX_UNIFORM_VECTORS_D3D11 <= D3D10_REQ_CONSTANT_BUFFER_ELEMENT_COUNT); + ASSERT(mFeatureLevel >= D3D_FEATURE_LEVEL_10_0); + return MAX_VERTEX_UNIFORM_VECTORS_D3D11; +} + +unsigned int Renderer11::getMaxFragmentUniformVectors() const +{ + META_ASSERT(MAX_FRAGMENT_UNIFORM_VECTORS_D3D11 <= D3D10_REQ_CONSTANT_BUFFER_ELEMENT_COUNT); + ASSERT(mFeatureLevel >= D3D_FEATURE_LEVEL_10_0); + return MAX_FRAGMENT_UNIFORM_VECTORS_D3D11; +} + +unsigned int Renderer11::getMaxVaryingVectors() const +{ + META_ASSERT(gl::IMPLEMENTATION_MAX_VARYING_VECTORS == D3D11_VS_OUTPUT_REGISTER_COUNT); + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + return D3D11_VS_OUTPUT_REGISTER_COUNT; + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return D3D10_VS_OUTPUT_REGISTER_COUNT; + default: UNREACHABLE(); + return 0; + } +} + +bool Renderer11::getNonPower2TextureSupport() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return true; + default: UNREACHABLE(); + return false; + } +} + +bool Renderer11::getOcclusionQuerySupport() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return true; + default: UNREACHABLE(); + return false; + } +} + +bool Renderer11::getInstancingSupport() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return true; + default: UNREACHABLE(); + return false; + } +} + +bool Renderer11::getShareHandleSupport() const +{ + // We only currently support share handles with BGRA surfaces, because + // chrome needs BGRA. Once chrome fixes this, we should always support them. + // PIX doesn't seem to support using share handles, so disable them. + return getBGRATextureSupport() && !gl::perfActive(); +} + +bool Renderer11::getDerivativeInstructionSupport() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return true; + default: UNREACHABLE(); + return false; + } +} + +bool Renderer11::getPostSubBufferSupport() const +{ + // D3D11 does not support present with dirty rectangles until D3D11.1 and DXGI 1.2. + return false; +} + +int Renderer11::getMajorShaderModel() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: return D3D11_SHADER_MAJOR_VERSION; // 5 + case D3D_FEATURE_LEVEL_10_1: return D3D10_1_SHADER_MAJOR_VERSION; // 4 + case D3D_FEATURE_LEVEL_10_0: return D3D10_SHADER_MAJOR_VERSION; // 4 + default: UNREACHABLE(); return 0; + } +} + +int Renderer11::getMinorShaderModel() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: return D3D11_SHADER_MINOR_VERSION; // 0 + case D3D_FEATURE_LEVEL_10_1: return D3D10_1_SHADER_MINOR_VERSION; // 1 + case D3D_FEATURE_LEVEL_10_0: return D3D10_SHADER_MINOR_VERSION; // 0 + default: UNREACHABLE(); return 0; + } +} + +float Renderer11::getMaxPointSize() const +{ + // choose a reasonable maximum. we enforce this in the shader. + // (nb: on a Radeon 2600xt, DX9 reports a 256 max point size) + return 1024.0f; +} + +int Renderer11::getMaxViewportDimension() const +{ + // Maximum viewport size must be at least as large as the largest render buffer (or larger). + // In our case return the maximum texture size, which is the maximum render buffer size. + META_ASSERT(D3D11_REQ_TEXTURE2D_U_OR_V_DIMENSION * 2 - 1 <= D3D11_VIEWPORT_BOUNDS_MAX); + META_ASSERT(D3D10_REQ_TEXTURE2D_U_OR_V_DIMENSION * 2 - 1 <= D3D10_VIEWPORT_BOUNDS_MAX); + + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + return D3D11_REQ_TEXTURE2D_U_OR_V_DIMENSION; // 16384 + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return D3D10_REQ_TEXTURE2D_U_OR_V_DIMENSION; // 8192 + default: UNREACHABLE(); + return 0; + } +} + +int Renderer11::getMaxTextureWidth() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_TEXTURE2D_U_OR_V_DIMENSION; // 16384 + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_TEXTURE2D_U_OR_V_DIMENSION; // 8192 + default: UNREACHABLE(); return 0; + } +} + +int Renderer11::getMaxTextureHeight() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_TEXTURE2D_U_OR_V_DIMENSION; // 16384 + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_TEXTURE2D_U_OR_V_DIMENSION; // 8192 + default: UNREACHABLE(); return 0; + } +} + +bool Renderer11::get32BitIndexSupport() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_DRAWINDEXED_INDEX_COUNT_2_TO_EXP >= 32; // true + default: UNREACHABLE(); return false; + } +} + +int Renderer11::getMinSwapInterval() const +{ + return 0; +} + +int Renderer11::getMaxSwapInterval() const +{ + return 4; +} + +int Renderer11::getMaxSupportedSamples() const +{ + return mMaxSupportedSamples; +} + +int Renderer11::getNearestSupportedSamples(DXGI_FORMAT format, unsigned int requested) const +{ + if (requested == 0) + { + return 0; + } + + MultisampleSupportMap::const_iterator iter = mMultisampleSupportMap.find(format); + if (iter != mMultisampleSupportMap.end()) + { + const MultisampleSupportInfo& info = iter->second; + for (unsigned int i = requested - 1; i < D3D11_MAX_MULTISAMPLE_SAMPLE_COUNT; i++) + { + if (info.qualityLevels[i] > 0) + { + return i + 1; + } + } + } + + return -1; +} + +unsigned int Renderer11::getMaxRenderTargets() const +{ + META_ASSERT(D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT <= gl::IMPLEMENTATION_MAX_DRAW_BUFFERS); + META_ASSERT(D3D10_SIMULTANEOUS_RENDER_TARGET_COUNT <= gl::IMPLEMENTATION_MAX_DRAW_BUFFERS); + + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: + return D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT; // 8 + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + return D3D10_SIMULTANEOUS_RENDER_TARGET_COUNT; // 8 + default: + UNREACHABLE(); + return 1; + } +} + +bool Renderer11::copyToRenderTarget(TextureStorageInterface2D *dest, TextureStorageInterface2D *source) +{ + if (source && dest) + { + TextureStorage11_2D *source11 = TextureStorage11_2D::makeTextureStorage11_2D(source->getStorageInstance()); + TextureStorage11_2D *dest11 = TextureStorage11_2D::makeTextureStorage11_2D(dest->getStorageInstance()); + + mDeviceContext->CopyResource(dest11->getBaseTexture(), source11->getBaseTexture()); + return true; + } + + return false; +} + +bool Renderer11::copyToRenderTarget(TextureStorageInterfaceCube *dest, TextureStorageInterfaceCube *source) +{ + if (source && dest) + { + TextureStorage11_Cube *source11 = TextureStorage11_Cube::makeTextureStorage11_Cube(source->getStorageInstance()); + TextureStorage11_Cube *dest11 = TextureStorage11_Cube::makeTextureStorage11_Cube(dest->getStorageInstance()); + + mDeviceContext->CopyResource(dest11->getBaseTexture(), source11->getBaseTexture()); + return true; + } + + return false; +} + +bool Renderer11::copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterface2D *storage, GLint level) +{ + gl::Renderbuffer *colorbuffer = framebuffer->getReadColorbuffer(); + if (!colorbuffer) + { + ERR("Failed to retrieve the color buffer from the frame buffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RenderTarget11 *sourceRenderTarget = RenderTarget11::makeRenderTarget11(colorbuffer->getRenderTarget()); + if (!sourceRenderTarget) + { + ERR("Failed to retrieve the render target from the frame buffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + ID3D11ShaderResourceView *source = sourceRenderTarget->getShaderResourceView(); + if (!source) + { + ERR("Failed to retrieve the render target view from the render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + TextureStorage11_2D *storage11 = TextureStorage11_2D::makeTextureStorage11_2D(storage->getStorageInstance()); + if (!storage11) + { + source->Release(); + ERR("Failed to retrieve the texture storage from the destination."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RenderTarget11 *destRenderTarget = RenderTarget11::makeRenderTarget11(storage11->getRenderTarget(level)); + if (!destRenderTarget) + { + source->Release(); + ERR("Failed to retrieve the render target from the destination storage."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + ID3D11RenderTargetView *dest = destRenderTarget->getRenderTargetView(); + if (!dest) + { + source->Release(); + ERR("Failed to retrieve the render target view from the destination render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + gl::Rectangle destRect; + destRect.x = xoffset; + destRect.y = yoffset; + destRect.width = sourceRect.width; + destRect.height = sourceRect.height; + + bool ret = copyTexture(source, sourceRect, sourceRenderTarget->getWidth(), sourceRenderTarget->getHeight(), + dest, destRect, destRenderTarget->getWidth(), destRenderTarget->getHeight(), destFormat); + + source->Release(); + dest->Release(); + + return ret; +} + +bool Renderer11::copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterfaceCube *storage, GLenum target, GLint level) +{ + gl::Renderbuffer *colorbuffer = framebuffer->getReadColorbuffer(); + if (!colorbuffer) + { + ERR("Failed to retrieve the color buffer from the frame buffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RenderTarget11 *sourceRenderTarget = RenderTarget11::makeRenderTarget11(colorbuffer->getRenderTarget()); + if (!sourceRenderTarget) + { + ERR("Failed to retrieve the render target from the frame buffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + ID3D11ShaderResourceView *source = sourceRenderTarget->getShaderResourceView(); + if (!source) + { + ERR("Failed to retrieve the render target view from the render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + TextureStorage11_Cube *storage11 = TextureStorage11_Cube::makeTextureStorage11_Cube(storage->getStorageInstance()); + if (!storage11) + { + source->Release(); + ERR("Failed to retrieve the texture storage from the destination."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RenderTarget11 *destRenderTarget = RenderTarget11::makeRenderTarget11(storage11->getRenderTarget(target, level)); + if (!destRenderTarget) + { + source->Release(); + ERR("Failed to retrieve the render target from the destination storage."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + ID3D11RenderTargetView *dest = destRenderTarget->getRenderTargetView(); + if (!dest) + { + source->Release(); + ERR("Failed to retrieve the render target view from the destination render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + gl::Rectangle destRect; + destRect.x = xoffset; + destRect.y = yoffset; + destRect.width = sourceRect.width; + destRect.height = sourceRect.height; + + bool ret = copyTexture(source, sourceRect, sourceRenderTarget->getWidth(), sourceRenderTarget->getHeight(), + dest, destRect, destRenderTarget->getWidth(), destRenderTarget->getHeight(), destFormat); + + source->Release(); + dest->Release(); + + return ret; +} + +bool Renderer11::copyTexture(ID3D11ShaderResourceView *source, const gl::Rectangle &sourceArea, unsigned int sourceWidth, unsigned int sourceHeight, + ID3D11RenderTargetView *dest, const gl::Rectangle &destArea, unsigned int destWidth, unsigned int destHeight, GLenum destFormat) +{ + HRESULT result; + + if (!mCopyResourcesInitialized) + { + ASSERT(!mCopyVB && !mCopySampler && !mCopyIL && !mCopyVS && !mCopyRGBAPS && !mCopyRGBPS && !mCopyLumPS && !mCopyLumAlphaPS); + + D3D11_BUFFER_DESC vbDesc; + vbDesc.ByteWidth = sizeof(d3d11::PositionTexCoordVertex) * 4; + vbDesc.Usage = D3D11_USAGE_DYNAMIC; + vbDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + vbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + vbDesc.MiscFlags = 0; + vbDesc.StructureByteStride = 0; + + result = mDevice->CreateBuffer(&vbDesc, NULL, &mCopyVB); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mCopyVB, "Renderer11 copy texture vertex buffer"); + + D3D11_SAMPLER_DESC samplerDesc; + samplerDesc.Filter = D3D11_FILTER_MIN_MAG_MIP_LINEAR; + samplerDesc.AddressU = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.AddressV = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.AddressW = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.MipLODBias = 0.0f; + samplerDesc.MaxAnisotropy = 0; + samplerDesc.ComparisonFunc = D3D11_COMPARISON_NEVER; + samplerDesc.BorderColor[0] = 0.0f; + samplerDesc.BorderColor[1] = 0.0f; + samplerDesc.BorderColor[2] = 0.0f; + samplerDesc.BorderColor[3] = 0.0f; + samplerDesc.MinLOD = 0.0f; + samplerDesc.MaxLOD = 0.0f; + + result = mDevice->CreateSamplerState(&samplerDesc, &mCopySampler); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mCopySampler, "Renderer11 copy sampler"); + + D3D11_INPUT_ELEMENT_DESC quadLayout[] = + { + { "POSITION", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + { "TEXCOORD", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 8, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + }; + + result = mDevice->CreateInputLayout(quadLayout, 2, g_VS_Passthrough, sizeof(g_VS_Passthrough), &mCopyIL); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mCopyIL, "Renderer11 copy texture input layout"); + + result = mDevice->CreateVertexShader(g_VS_Passthrough, sizeof(g_VS_Passthrough), NULL, &mCopyVS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mCopyVS, "Renderer11 copy texture vertex shader"); + + result = mDevice->CreatePixelShader(g_PS_PassthroughRGBA, sizeof(g_PS_PassthroughRGBA), NULL, &mCopyRGBAPS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mCopyRGBAPS, "Renderer11 copy texture RGBA pixel shader"); + + result = mDevice->CreatePixelShader(g_PS_PassthroughRGB, sizeof(g_PS_PassthroughRGB), NULL, &mCopyRGBPS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mCopyRGBPS, "Renderer11 copy texture RGB pixel shader"); + + result = mDevice->CreatePixelShader(g_PS_PassthroughLum, sizeof(g_PS_PassthroughLum), NULL, &mCopyLumPS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mCopyLumPS, "Renderer11 copy texture luminance pixel shader"); + + result = mDevice->CreatePixelShader(g_PS_PassthroughLumAlpha, sizeof(g_PS_PassthroughLumAlpha), NULL, &mCopyLumAlphaPS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mCopyLumAlphaPS, "Renderer11 copy texture luminance alpha pixel shader"); + + mCopyResourcesInitialized = true; + } + + // Verify the source and destination area sizes + if (sourceArea.x < 0 || sourceArea.x + sourceArea.width > static_cast<int>(sourceWidth) || + sourceArea.y < 0 || sourceArea.y + sourceArea.height > static_cast<int>(sourceHeight) || + destArea.x < 0 || destArea.x + destArea.width > static_cast<int>(destWidth) || + destArea.y < 0 || destArea.y + destArea.height > static_cast<int>(destHeight)) + { + return gl::error(GL_INVALID_VALUE, false); + } + + // Set vertices + D3D11_MAPPED_SUBRESOURCE mappedResource; + result = mDeviceContext->Map(mCopyVB, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + ERR("Failed to map vertex buffer for texture copy, HRESULT: 0x%X.", result); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + d3d11::PositionTexCoordVertex *vertices = static_cast<d3d11::PositionTexCoordVertex*>(mappedResource.pData); + + // Create a quad in homogeneous coordinates + float x1 = (destArea.x / float(destWidth)) * 2.0f - 1.0f; + float y1 = ((destHeight - destArea.y - destArea.height) / float(destHeight)) * 2.0f - 1.0f; + float x2 = ((destArea.x + destArea.width) / float(destWidth)) * 2.0f - 1.0f; + float y2 = ((destHeight - destArea.y) / float(destHeight)) * 2.0f - 1.0f; + + float u1 = sourceArea.x / float(sourceWidth); + float v1 = sourceArea.y / float(sourceHeight); + float u2 = (sourceArea.x + sourceArea.width) / float(sourceWidth); + float v2 = (sourceArea.y + sourceArea.height) / float(sourceHeight); + + d3d11::SetPositionTexCoordVertex(&vertices[0], x1, y1, u1, v2); + d3d11::SetPositionTexCoordVertex(&vertices[1], x1, y2, u1, v1); + d3d11::SetPositionTexCoordVertex(&vertices[2], x2, y1, u2, v2); + d3d11::SetPositionTexCoordVertex(&vertices[3], x2, y2, u2, v1); + + mDeviceContext->Unmap(mCopyVB, 0); + + static UINT stride = sizeof(d3d11::PositionTexCoordVertex); + static UINT startIdx = 0; + mDeviceContext->IASetVertexBuffers(0, 1, &mCopyVB, &stride, &startIdx); + + // Apply state + mDeviceContext->OMSetBlendState(NULL, NULL, 0xFFFFFFF); + mDeviceContext->OMSetDepthStencilState(NULL, 0xFFFFFFFF); + mDeviceContext->RSSetState(NULL); + + // Apply shaders + mDeviceContext->IASetInputLayout(mCopyIL); + mDeviceContext->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP); + mDeviceContext->VSSetShader(mCopyVS, NULL, 0); + + ID3D11PixelShader *ps = NULL; + switch(destFormat) + { + case GL_RGBA: ps = mCopyRGBAPS; break; + case GL_RGB: ps = mCopyRGBPS; break; + case GL_ALPHA: ps = mCopyRGBAPS; break; + case GL_BGRA_EXT: ps = mCopyRGBAPS; break; + case GL_LUMINANCE: ps = mCopyLumPS; break; + case GL_LUMINANCE_ALPHA: ps = mCopyLumAlphaPS; break; + default: UNREACHABLE(); ps = NULL; break; + } + + mDeviceContext->PSSetShader(ps, NULL, 0); + mDeviceContext->GSSetShader(NULL, NULL, 0); + + // Unset the currently bound shader resource to avoid conflicts + static ID3D11ShaderResourceView *const nullSRV = NULL; + mDeviceContext->PSSetShaderResources(0, 1, &nullSRV); + + // Apply render target + setOneTimeRenderTarget(dest); + + // Set the viewport + D3D11_VIEWPORT viewport; + viewport.TopLeftX = 0; + viewport.TopLeftY = 0; + viewport.Width = destWidth; + viewport.Height = destHeight; + viewport.MinDepth = 0.0f; + viewport.MaxDepth = 1.0f; + mDeviceContext->RSSetViewports(1, &viewport); + + // Apply textures + mDeviceContext->PSSetShaderResources(0, 1, &source); + mDeviceContext->PSSetSamplers(0, 1, &mCopySampler); + + // Draw the quad + mDeviceContext->Draw(4, 0); + + // Unbind textures and render targets and vertex buffer + mDeviceContext->PSSetShaderResources(0, 1, &nullSRV); + + unapplyRenderTargets(); + + UINT zero = 0; + ID3D11Buffer *const nullBuffer = NULL; + mDeviceContext->IASetVertexBuffers(0, 1, &nullBuffer, &zero, &zero); + + markAllStateDirty(); + + return true; +} + +void Renderer11::unapplyRenderTargets() +{ + setOneTimeRenderTarget(NULL); +} + +void Renderer11::setOneTimeRenderTarget(ID3D11RenderTargetView *renderTargetView) +{ + ID3D11RenderTargetView *rtvArray[gl::IMPLEMENTATION_MAX_DRAW_BUFFERS] = {NULL}; + + rtvArray[0] = renderTargetView; + + mDeviceContext->OMSetRenderTargets(getMaxRenderTargets(), rtvArray, NULL); + + // Do not preserve the serial for this one-time-use render target + for (unsigned int rtIndex = 0; rtIndex < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; rtIndex++) + { + mAppliedRenderTargetSerials[rtIndex] = 0; + } +} + +RenderTarget *Renderer11::createRenderTarget(SwapChain *swapChain, bool depth) +{ + SwapChain11 *swapChain11 = SwapChain11::makeSwapChain11(swapChain); + RenderTarget11 *renderTarget = NULL; + + if (depth) + { + // Note: depth stencil may be NULL for 0 sized surfaces + renderTarget = new RenderTarget11(this, swapChain11->getDepthStencil(), + swapChain11->getDepthStencilTexture(), NULL, + swapChain11->getWidth(), swapChain11->getHeight()); + } + else + { + // Note: render target may be NULL for 0 sized surfaces + renderTarget = new RenderTarget11(this, swapChain11->getRenderTarget(), + swapChain11->getOffscreenTexture(), + swapChain11->getRenderTargetShaderResource(), + swapChain11->getWidth(), swapChain11->getHeight()); + } + return renderTarget; +} + +RenderTarget *Renderer11::createRenderTarget(int width, int height, GLenum format, GLsizei samples, bool depth) +{ + RenderTarget11 *renderTarget = new RenderTarget11(this, width, height, format, samples, depth); + return renderTarget; +} + +ShaderExecutable *Renderer11::loadExecutable(const void *function, size_t length, rx::ShaderType type) +{ + ShaderExecutable11 *executable = NULL; + + switch (type) + { + case rx::SHADER_VERTEX: + { + ID3D11VertexShader *vshader = NULL; + HRESULT result = mDevice->CreateVertexShader(function, length, NULL, &vshader); + ASSERT(SUCCEEDED(result)); + + if (vshader) + { + executable = new ShaderExecutable11(function, length, vshader); + } + } + break; + case rx::SHADER_PIXEL: + { + ID3D11PixelShader *pshader = NULL; + HRESULT result = mDevice->CreatePixelShader(function, length, NULL, &pshader); + ASSERT(SUCCEEDED(result)); + + if (pshader) + { + executable = new ShaderExecutable11(function, length, pshader); + } + } + break; + case rx::SHADER_GEOMETRY: + { + ID3D11GeometryShader *gshader = NULL; + HRESULT result = mDevice->CreateGeometryShader(function, length, NULL, &gshader); + ASSERT(SUCCEEDED(result)); + + if (gshader) + { + executable = new ShaderExecutable11(function, length, gshader); + } + } + break; + default: + UNREACHABLE(); + break; + } + + return executable; +} + +ShaderExecutable *Renderer11::compileToExecutable(gl::InfoLog &infoLog, const char *shaderHLSL, rx::ShaderType type) +{ + const char *profile = NULL; + + switch (type) + { + case rx::SHADER_VERTEX: + profile = "vs_4_0"; + break; + case rx::SHADER_PIXEL: + profile = "ps_4_0"; + break; + case rx::SHADER_GEOMETRY: + profile = "gs_4_0"; + break; + default: + UNREACHABLE(); + return NULL; + } + + ID3DBlob *binary = (ID3DBlob*)compileToBinary(infoLog, shaderHLSL, profile, D3DCOMPILE_OPTIMIZATION_LEVEL0, false); + if (!binary) + return NULL; + + ShaderExecutable *executable = loadExecutable((DWORD *)binary->GetBufferPointer(), binary->GetBufferSize(), type); + binary->Release(); + + return executable; +} + +VertexBuffer *Renderer11::createVertexBuffer() +{ + return new VertexBuffer11(this); +} + +IndexBuffer *Renderer11::createIndexBuffer() +{ + return new IndexBuffer11(this); +} + +BufferStorage *Renderer11::createBufferStorage() +{ + return new BufferStorage11(this); +} + +QueryImpl *Renderer11::createQuery(GLenum type) +{ + return new Query11(this, type); +} + +FenceImpl *Renderer11::createFence() +{ + return new Fence11(this); +} + +bool Renderer11::getRenderTargetResource(gl::Renderbuffer *colorbuffer, unsigned int *subresourceIndex, ID3D11Texture2D **resource) +{ + ASSERT(colorbuffer != NULL); + + RenderTarget11 *renderTarget = RenderTarget11::makeRenderTarget11(colorbuffer->getRenderTarget()); + if (renderTarget) + { + *subresourceIndex = renderTarget->getSubresourceIndex(); + + ID3D11RenderTargetView *colorBufferRTV = renderTarget->getRenderTargetView(); + if (colorBufferRTV) + { + ID3D11Resource *textureResource = NULL; + colorBufferRTV->GetResource(&textureResource); + colorBufferRTV->Release(); + + if (textureResource) + { + HRESULT result = textureResource->QueryInterface(IID_ID3D11Texture2D, (void**)resource); + textureResource->Release(); + + if (SUCCEEDED(result)) + { + return true; + } + else + { + ERR("Failed to extract the ID3D11Texture2D from the render target resource, " + "HRESULT: 0x%X.", result); + } + } + } + } + + return false; +} + +bool Renderer11::blitRect(gl::Framebuffer *readTarget, const gl::Rectangle &readRect, gl::Framebuffer *drawTarget, const gl::Rectangle &drawRect, + bool blitRenderTarget, bool blitDepthStencil) +{ + if (blitRenderTarget) + { + gl::Renderbuffer *readBuffer = readTarget->getReadColorbuffer(); + + if (!readBuffer) + { + ERR("Failed to retrieve the read buffer from the read framebuffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RenderTarget *readRenderTarget = readBuffer->getRenderTarget(); + + for (unsigned int colorAttachment = 0; colorAttachment < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; colorAttachment++) + { + if (drawTarget->isEnabledColorAttachment(colorAttachment)) + { + gl::Renderbuffer *drawBuffer = drawTarget->getColorbuffer(colorAttachment); + + if (!drawBuffer) + { + ERR("Failed to retrieve the draw buffer from the draw framebuffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RenderTarget *drawRenderTarget = drawBuffer->getRenderTarget(); + + if (!blitRenderbufferRect(readRect, drawRect, readRenderTarget, drawRenderTarget, false)) + { + return false; + } + } + } + } + + if (blitDepthStencil) + { + gl::Renderbuffer *readBuffer = readTarget->getDepthOrStencilbuffer(); + gl::Renderbuffer *drawBuffer = drawTarget->getDepthOrStencilbuffer(); + + if (!readBuffer) + { + ERR("Failed to retrieve the read depth-stencil buffer from the read framebuffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + if (!drawBuffer) + { + ERR("Failed to retrieve the draw depth-stencil buffer from the draw framebuffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RenderTarget *readRenderTarget = readBuffer->getDepthStencil(); + RenderTarget *drawRenderTarget = drawBuffer->getDepthStencil(); + + if (!blitRenderbufferRect(readRect, drawRect, readRenderTarget, drawRenderTarget, true)) + { + return false; + } + } + + return true; +} + +void Renderer11::readPixels(gl::Framebuffer *framebuffer, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, + GLsizei outputPitch, bool packReverseRowOrder, GLint packAlignment, void* pixels) +{ + ID3D11Texture2D *colorBufferTexture = NULL; + unsigned int subresourceIndex = 0; + + gl::Renderbuffer *colorbuffer = framebuffer->getReadColorbuffer(); + + if (colorbuffer && getRenderTargetResource(colorbuffer, &subresourceIndex, &colorBufferTexture)) + { + gl::Rectangle area; + area.x = x; + area.y = y; + area.width = width; + area.height = height; + + readTextureData(colorBufferTexture, subresourceIndex, area, format, type, outputPitch, + packReverseRowOrder, packAlignment, pixels); + + colorBufferTexture->Release(); + colorBufferTexture = NULL; + } +} + +Image *Renderer11::createImage() +{ + return new Image11(); +} + +void Renderer11::generateMipmap(Image *dest, Image *src) +{ + Image11 *dest11 = Image11::makeImage11(dest); + Image11 *src11 = Image11::makeImage11(src); + Image11::generateMipmap(dest11, src11); +} + +TextureStorage *Renderer11::createTextureStorage2D(SwapChain *swapChain) +{ + SwapChain11 *swapChain11 = SwapChain11::makeSwapChain11(swapChain); + return new TextureStorage11_2D(this, swapChain11); +} + +TextureStorage *Renderer11::createTextureStorage2D(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height) +{ + return new TextureStorage11_2D(this, levels, internalformat, usage, forceRenderable, width, height); +} + +TextureStorage *Renderer11::createTextureStorageCube(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size) +{ + return new TextureStorage11_Cube(this, levels, internalformat, usage, forceRenderable, size); +} + +static inline unsigned int getFastPixelCopySize(DXGI_FORMAT sourceFormat, GLenum destFormat, GLenum destType) +{ + if (sourceFormat == DXGI_FORMAT_A8_UNORM && + destFormat == GL_ALPHA && + destType == GL_UNSIGNED_BYTE) + { + return 1; + } + else if (sourceFormat == DXGI_FORMAT_R8G8B8A8_UNORM && + destFormat == GL_RGBA && + destType == GL_UNSIGNED_BYTE) + { + return 4; + } + else if (sourceFormat == DXGI_FORMAT_B8G8R8A8_UNORM && + destFormat == GL_BGRA_EXT && + destType == GL_UNSIGNED_BYTE) + { + return 4; + } + else if (sourceFormat == DXGI_FORMAT_R16G16B16A16_FLOAT && + destFormat == GL_RGBA && + destType == GL_HALF_FLOAT_OES) + { + return 8; + } + else if (sourceFormat == DXGI_FORMAT_R32G32B32_FLOAT && + destFormat == GL_RGB && + destType == GL_FLOAT) + { + return 12; + } + else if (sourceFormat == DXGI_FORMAT_R32G32B32A32_FLOAT && + destFormat == GL_RGBA && + destType == GL_FLOAT) + { + return 16; + } + else + { + return 0; + } +} + +static inline void readPixelColor(const unsigned char *data, DXGI_FORMAT format, unsigned int x, + unsigned int y, int inputPitch, gl::Color *outColor) +{ + switch (format) + { + case DXGI_FORMAT_R8G8B8A8_UNORM: + { + unsigned int rgba = *reinterpret_cast<const unsigned int*>(data + 4 * x + y * inputPitch); + outColor->red = (rgba & 0x000000FF) * (1.0f / 0x000000FF); + outColor->green = (rgba & 0x0000FF00) * (1.0f / 0x0000FF00); + outColor->blue = (rgba & 0x00FF0000) * (1.0f / 0x00FF0000); + outColor->alpha = (rgba & 0xFF000000) * (1.0f / 0xFF000000); + } + break; + + case DXGI_FORMAT_A8_UNORM: + { + outColor->red = 0.0f; + outColor->green = 0.0f; + outColor->blue = 0.0f; + outColor->alpha = *(data + x + y * inputPitch) / 255.0f; + } + break; + + case DXGI_FORMAT_R32G32B32A32_FLOAT: + { + outColor->red = *(reinterpret_cast<const float*>(data + 16 * x + y * inputPitch) + 0); + outColor->green = *(reinterpret_cast<const float*>(data + 16 * x + y * inputPitch) + 1); + outColor->blue = *(reinterpret_cast<const float*>(data + 16 * x + y * inputPitch) + 2); + outColor->alpha = *(reinterpret_cast<const float*>(data + 16 * x + y * inputPitch) + 3); + } + break; + + case DXGI_FORMAT_R32G32B32_FLOAT: + { + outColor->red = *(reinterpret_cast<const float*>(data + 12 * x + y * inputPitch) + 0); + outColor->green = *(reinterpret_cast<const float*>(data + 12 * x + y * inputPitch) + 1); + outColor->blue = *(reinterpret_cast<const float*>(data + 12 * x + y * inputPitch) + 2); + outColor->alpha = 1.0f; + } + break; + + case DXGI_FORMAT_R16G16B16A16_FLOAT: + { + outColor->red = gl::float16ToFloat32(*(reinterpret_cast<const unsigned short*>(data + 8 * x + y * inputPitch) + 0)); + outColor->green = gl::float16ToFloat32(*(reinterpret_cast<const unsigned short*>(data + 8 * x + y * inputPitch) + 1)); + outColor->blue = gl::float16ToFloat32(*(reinterpret_cast<const unsigned short*>(data + 8 * x + y * inputPitch) + 2)); + outColor->alpha = gl::float16ToFloat32(*(reinterpret_cast<const unsigned short*>(data + 8 * x + y * inputPitch) + 3)); + } + break; + + case DXGI_FORMAT_B8G8R8A8_UNORM: + { + unsigned int bgra = *reinterpret_cast<const unsigned int*>(data + 4 * x + y * inputPitch); + outColor->red = (bgra & 0x00FF0000) * (1.0f / 0x00FF0000); + outColor->blue = (bgra & 0x000000FF) * (1.0f / 0x000000FF); + outColor->green = (bgra & 0x0000FF00) * (1.0f / 0x0000FF00); + outColor->alpha = (bgra & 0xFF000000) * (1.0f / 0xFF000000); + } + break; + + case DXGI_FORMAT_R8_UNORM: + { + outColor->red = *(data + x + y * inputPitch) / 255.0f; + outColor->green = 0.0f; + outColor->blue = 0.0f; + outColor->alpha = 1.0f; + } + break; + + case DXGI_FORMAT_R8G8_UNORM: + { + unsigned short rg = *reinterpret_cast<const unsigned short*>(data + 2 * x + y * inputPitch); + + outColor->red = (rg & 0xFF00) * (1.0f / 0xFF00); + outColor->green = (rg & 0x00FF) * (1.0f / 0x00FF); + outColor->blue = 0.0f; + outColor->alpha = 1.0f; + } + break; + + case DXGI_FORMAT_R16_FLOAT: + { + outColor->red = gl::float16ToFloat32(*reinterpret_cast<const unsigned short*>(data + 2 * x + y * inputPitch)); + outColor->green = 0.0f; + outColor->blue = 0.0f; + outColor->alpha = 1.0f; + } + break; + + case DXGI_FORMAT_R16G16_FLOAT: + { + outColor->red = gl::float16ToFloat32(*(reinterpret_cast<const unsigned short*>(data + 4 * x + y * inputPitch) + 0)); + outColor->green = gl::float16ToFloat32(*(reinterpret_cast<const unsigned short*>(data + 4 * x + y * inputPitch) + 1)); + outColor->blue = 0.0f; + outColor->alpha = 1.0f; + } + break; + + default: + ERR("ReadPixelColor not implemented for DXGI format %u.", format); + UNIMPLEMENTED(); + break; + } +} + +static inline void writePixelColor(const gl::Color &color, GLenum format, GLenum type, unsigned int x, + unsigned int y, int outputPitch, void *outData) +{ + unsigned char* byteData = reinterpret_cast<unsigned char*>(outData); + unsigned short* shortData = reinterpret_cast<unsigned short*>(outData); + + switch (format) + { + case GL_RGBA: + switch (type) + { + case GL_UNSIGNED_BYTE: + byteData[4 * x + y * outputPitch + 0] = static_cast<unsigned char>(255 * color.red + 0.5f); + byteData[4 * x + y * outputPitch + 1] = static_cast<unsigned char>(255 * color.green + 0.5f); + byteData[4 * x + y * outputPitch + 2] = static_cast<unsigned char>(255 * color.blue + 0.5f); + byteData[4 * x + y * outputPitch + 3] = static_cast<unsigned char>(255 * color.alpha + 0.5f); + break; + + default: + ERR("WritePixelColor not implemented for format GL_RGBA and type 0x%X.", type); + UNIMPLEMENTED(); + break; + } + break; + + case GL_BGRA_EXT: + switch (type) + { + case GL_UNSIGNED_BYTE: + byteData[4 * x + y * outputPitch + 0] = static_cast<unsigned char>(255 * color.blue + 0.5f); + byteData[4 * x + y * outputPitch + 1] = static_cast<unsigned char>(255 * color.green + 0.5f); + byteData[4 * x + y * outputPitch + 2] = static_cast<unsigned char>(255 * color.red + 0.5f); + byteData[4 * x + y * outputPitch + 3] = static_cast<unsigned char>(255 * color.alpha + 0.5f); + break; + + case GL_UNSIGNED_SHORT_4_4_4_4_REV_EXT: + // According to the desktop GL spec in the "Transfer of Pixel Rectangles" section + // this type is packed as follows: + // 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 + // -------------------------------------------------------------------------------- + // | 4th | 3rd | 2nd | 1st component | + // -------------------------------------------------------------------------------- + // in the case of BGRA_EXT, B is the first component, G the second, and so forth. + shortData[x + y * outputPitch / sizeof(unsigned short)] = + (static_cast<unsigned short>(15 * color.alpha + 0.5f) << 12) | + (static_cast<unsigned short>(15 * color.red + 0.5f) << 8) | + (static_cast<unsigned short>(15 * color.green + 0.5f) << 4) | + (static_cast<unsigned short>(15 * color.blue + 0.5f) << 0); + break; + + case GL_UNSIGNED_SHORT_1_5_5_5_REV_EXT: + // According to the desktop GL spec in the "Transfer of Pixel Rectangles" section + // this type is packed as follows: + // 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 + // -------------------------------------------------------------------------------- + // | 4th | 3rd | 2nd | 1st component | + // -------------------------------------------------------------------------------- + // in the case of BGRA_EXT, B is the first component, G the second, and so forth. + shortData[x + y * outputPitch / sizeof(unsigned short)] = + (static_cast<unsigned short>( color.alpha + 0.5f) << 15) | + (static_cast<unsigned short>(31 * color.red + 0.5f) << 10) | + (static_cast<unsigned short>(31 * color.green + 0.5f) << 5) | + (static_cast<unsigned short>(31 * color.blue + 0.5f) << 0); + break; + + default: + ERR("WritePixelColor not implemented for format GL_BGRA_EXT and type 0x%X.", type); + UNIMPLEMENTED(); + break; + } + break; + + case GL_RGB: + switch (type) + { + case GL_UNSIGNED_SHORT_5_6_5: + shortData[x + y * outputPitch / sizeof(unsigned short)] = + (static_cast<unsigned short>(31 * color.blue + 0.5f) << 0) | + (static_cast<unsigned short>(63 * color.green + 0.5f) << 5) | + (static_cast<unsigned short>(31 * color.red + 0.5f) << 11); + break; + + case GL_UNSIGNED_BYTE: + byteData[3 * x + y * outputPitch + 0] = static_cast<unsigned char>(255 * color.red + 0.5f); + byteData[3 * x + y * outputPitch + 1] = static_cast<unsigned char>(255 * color.green + 0.5f); + byteData[3 * x + y * outputPitch + 2] = static_cast<unsigned char>(255 * color.blue + 0.5f); + break; + + default: + ERR("WritePixelColor not implemented for format GL_RGB and type 0x%X.", type); + UNIMPLEMENTED(); + break; + } + break; + + default: + ERR("WritePixelColor not implemented for format 0x%X.", format); + UNIMPLEMENTED(); + break; + } +} + +void Renderer11::readTextureData(ID3D11Texture2D *texture, unsigned int subResource, const gl::Rectangle &area, + GLenum format, GLenum type, GLsizei outputPitch, bool packReverseRowOrder, + GLint packAlignment, void *pixels) +{ + D3D11_TEXTURE2D_DESC textureDesc; + texture->GetDesc(&textureDesc); + + D3D11_TEXTURE2D_DESC stagingDesc; + stagingDesc.Width = area.width; + stagingDesc.Height = area.height; + stagingDesc.MipLevels = 1; + stagingDesc.ArraySize = 1; + stagingDesc.Format = textureDesc.Format; + stagingDesc.SampleDesc.Count = 1; + stagingDesc.SampleDesc.Quality = 0; + stagingDesc.Usage = D3D11_USAGE_STAGING; + stagingDesc.BindFlags = 0; + stagingDesc.CPUAccessFlags = D3D11_CPU_ACCESS_READ; + stagingDesc.MiscFlags = 0; + + ID3D11Texture2D* stagingTex = NULL; + HRESULT result = mDevice->CreateTexture2D(&stagingDesc, NULL, &stagingTex); + if (FAILED(result)) + { + ERR("Failed to create staging texture for readPixels, HRESULT: 0x%X.", result); + return; + } + + ID3D11Texture2D* srcTex = NULL; + if (textureDesc.SampleDesc.Count > 1) + { + D3D11_TEXTURE2D_DESC resolveDesc; + resolveDesc.Width = textureDesc.Width; + resolveDesc.Height = textureDesc.Height; + resolveDesc.MipLevels = 1; + resolveDesc.ArraySize = 1; + resolveDesc.Format = textureDesc.Format; + resolveDesc.SampleDesc.Count = 1; + resolveDesc.SampleDesc.Quality = 0; + resolveDesc.Usage = D3D11_USAGE_DEFAULT; + resolveDesc.BindFlags = 0; + resolveDesc.CPUAccessFlags = 0; + resolveDesc.MiscFlags = 0; + + result = mDevice->CreateTexture2D(&resolveDesc, NULL, &srcTex); + if (FAILED(result)) + { + ERR("Failed to create resolve texture for readPixels, HRESULT: 0x%X.", result); + stagingTex->Release(); + return; + } + + mDeviceContext->ResolveSubresource(srcTex, 0, texture, subResource, textureDesc.Format); + subResource = 0; + } + else + { + srcTex = texture; + srcTex->AddRef(); + } + + D3D11_BOX srcBox; + srcBox.left = area.x; + srcBox.right = area.x + area.width; + srcBox.top = area.y; + srcBox.bottom = area.y + area.height; + srcBox.front = 0; + srcBox.back = 1; + + mDeviceContext->CopySubresourceRegion(stagingTex, 0, 0, 0, 0, srcTex, subResource, &srcBox); + + srcTex->Release(); + srcTex = NULL; + + D3D11_MAPPED_SUBRESOURCE mapping; + mDeviceContext->Map(stagingTex, 0, D3D11_MAP_READ, 0, &mapping); + + unsigned char *source; + int inputPitch; + if (packReverseRowOrder) + { + source = static_cast<unsigned char*>(mapping.pData) + mapping.RowPitch * (area.height - 1); + inputPitch = -static_cast<int>(mapping.RowPitch); + } + else + { + source = static_cast<unsigned char*>(mapping.pData); + inputPitch = static_cast<int>(mapping.RowPitch); + } + + unsigned int fastPixelSize = getFastPixelCopySize(textureDesc.Format, format, type); + if (fastPixelSize != 0) + { + unsigned char *dest = static_cast<unsigned char*>(pixels); + for (int j = 0; j < area.height; j++) + { + memcpy(dest + j * outputPitch, source + j * inputPitch, area.width * fastPixelSize); + } + } + else if (textureDesc.Format == DXGI_FORMAT_B8G8R8A8_UNORM && + format == GL_RGBA && + type == GL_UNSIGNED_BYTE) + { + // Fast path for swapping red with blue + unsigned char *dest = static_cast<unsigned char*>(pixels); + + for (int j = 0; j < area.height; j++) + { + for (int i = 0; i < area.width; i++) + { + unsigned int argb = *(unsigned int*)(source + 4 * i + j * inputPitch); + *(unsigned int*)(dest + 4 * i + j * outputPitch) = + (argb & 0xFF00FF00) | // Keep alpha and green + (argb & 0x00FF0000) >> 16 | // Move red to blue + (argb & 0x000000FF) << 16; // Move blue to red + } + } + } + else + { + gl::Color pixelColor; + for (int j = 0; j < area.height; j++) + { + for (int i = 0; i < area.width; i++) + { + readPixelColor(source, textureDesc.Format, i, j, inputPitch, &pixelColor); + writePixelColor(pixelColor, format, type, i, j, outputPitch, pixels); + } + } + } + + mDeviceContext->Unmap(stagingTex, 0); + + stagingTex->Release(); + stagingTex = NULL; +} + +bool Renderer11::blitRenderbufferRect(const gl::Rectangle &readRect, const gl::Rectangle &drawRect, RenderTarget *readRenderTarget, + RenderTarget *drawRenderTarget, bool wholeBufferCopy) +{ + ASSERT(readRect.width == drawRect.width && readRect.height == drawRect.height); + + RenderTarget11 *readRenderTarget11 = RenderTarget11::makeRenderTarget11(readRenderTarget); + if (!readRenderTarget) + { + ERR("Failed to retrieve the read render target from the read framebuffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + ID3D11Texture2D *readTexture = NULL; + unsigned int readSubresource = 0; + if (readRenderTarget->getSamples() > 0) + { + ID3D11Texture2D *unresolvedTexture = readRenderTarget11->getTexture(); + + readTexture = resolveMultisampledTexture(unresolvedTexture, readRenderTarget11->getSubresourceIndex()); + readSubresource = 0; + + unresolvedTexture->Release(); + } + else + { + readTexture = readRenderTarget11->getTexture(); + readSubresource = readRenderTarget11->getSubresourceIndex(); + } + + if (!readTexture) + { + ERR("Failed to retrieve the read render target view from the read render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RenderTarget11 *drawRenderTarget11 = RenderTarget11::makeRenderTarget11(drawRenderTarget); + if (!drawRenderTarget) + { + readTexture->Release(); + ERR("Failed to retrieve the draw render target from the draw framebuffer."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + ID3D11Texture2D *drawTexture = drawRenderTarget11->getTexture(); + unsigned int drawSubresource = drawRenderTarget11->getSubresourceIndex(); + + D3D11_BOX readBox; + readBox.left = readRect.x; + readBox.right = readRect.x + readRect.width; + readBox.top = readRect.y; + readBox.bottom = readRect.y + readRect.height; + readBox.front = 0; + readBox.back = 1; + + // D3D11 needs depth-stencil CopySubresourceRegions to have a NULL pSrcBox + // We also require complete framebuffer copies for depth-stencil blit. + D3D11_BOX *pSrcBox = wholeBufferCopy ? NULL : &readBox; + + mDeviceContext->CopySubresourceRegion(drawTexture, drawSubresource, drawRect.x, drawRect.y, 0, + readTexture, readSubresource, pSrcBox); + + readTexture->Release(); + drawTexture->Release(); + + return true; +} + +ID3D11Texture2D *Renderer11::resolveMultisampledTexture(ID3D11Texture2D *source, unsigned int subresource) +{ + D3D11_TEXTURE2D_DESC textureDesc; + source->GetDesc(&textureDesc); + + if (textureDesc.SampleDesc.Count > 1) + { + D3D11_TEXTURE2D_DESC resolveDesc; + resolveDesc.Width = textureDesc.Width; + resolveDesc.Height = textureDesc.Height; + resolveDesc.MipLevels = 1; + resolveDesc.ArraySize = 1; + resolveDesc.Format = textureDesc.Format; + resolveDesc.SampleDesc.Count = 1; + resolveDesc.SampleDesc.Quality = 0; + resolveDesc.Usage = textureDesc.Usage; + resolveDesc.BindFlags = textureDesc.BindFlags; + resolveDesc.CPUAccessFlags = 0; + resolveDesc.MiscFlags = 0; + + ID3D11Texture2D *resolveTexture = NULL; + HRESULT result = mDevice->CreateTexture2D(&resolveDesc, NULL, &resolveTexture); + if (FAILED(result)) + { + ERR("Failed to create a multisample resolve texture, HRESULT: 0x%X.", result); + return NULL; + } + + mDeviceContext->ResolveSubresource(resolveTexture, 0, source, subresource, textureDesc.Format); + return resolveTexture; + } + else + { + source->AddRef(); + return source; + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Renderer11.h b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer11.h new file mode 100644 index 0000000000..a3e42e9048 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer11.h @@ -0,0 +1,348 @@ +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Renderer11.h: Defines a back-end specific class for the D3D11 renderer. + +#ifndef LIBGLESV2_RENDERER_RENDERER11_H_ +#define LIBGLESV2_RENDERER_RENDERER11_H_ + +#include "common/angleutils.h" +#include "libGLESv2/angletypes.h" +#include "libGLESv2/mathutil.h" + +#include "libGLESv2/renderer/Renderer.h" +#include "libGLESv2/renderer/RenderStateCache.h" +#include "libGLESv2/renderer/InputLayoutCache.h" +#include "libGLESv2/renderer/RenderTarget.h" + +namespace gl +{ +class Renderbuffer; +} + +namespace rx +{ + +class VertexDataManager; +class IndexDataManager; +class StreamingIndexBufferInterface; + +enum +{ + MAX_VERTEX_UNIFORM_VECTORS_D3D11 = 1024, + MAX_FRAGMENT_UNIFORM_VECTORS_D3D11 = 1024 +}; + +class Renderer11 : public Renderer +{ + public: + Renderer11(egl::Display *display, HDC hDc); + virtual ~Renderer11(); + + static Renderer11 *makeRenderer11(Renderer *renderer); + + virtual EGLint initialize(); + virtual bool resetDevice(); + + virtual int generateConfigs(ConfigDesc **configDescList); + virtual void deleteConfigs(ConfigDesc *configDescList); + + virtual void sync(bool block); + + virtual SwapChain *createSwapChain(HWND window, HANDLE shareHandle, GLenum backBufferFormat, GLenum depthBufferFormat); + + virtual void setSamplerState(gl::SamplerType type, int index, const gl::SamplerState &sampler); + virtual void setTexture(gl::SamplerType type, int index, gl::Texture *texture); + + virtual void setRasterizerState(const gl::RasterizerState &rasterState); + virtual void setBlendState(const gl::BlendState &blendState, const gl::Color &blendColor, + unsigned int sampleMask); + virtual void setDepthStencilState(const gl::DepthStencilState &depthStencilState, int stencilRef, + int stencilBackRef, bool frontFaceCCW); + + virtual void setScissorRectangle(const gl::Rectangle &scissor, bool enabled); + virtual bool setViewport(const gl::Rectangle &viewport, float zNear, float zFar, GLenum drawMode, GLenum frontFace, + bool ignoreViewport); + + virtual bool applyPrimitiveType(GLenum mode, GLsizei count); + virtual bool applyRenderTarget(gl::Framebuffer *frameBuffer); + virtual void applyShaders(gl::ProgramBinary *programBinary); + virtual void applyUniforms(gl::ProgramBinary *programBinary, gl::UniformArray *uniformArray); + virtual GLenum applyVertexBuffer(gl::ProgramBinary *programBinary, gl::VertexAttribute vertexAttributes[], GLint first, GLsizei count, GLsizei instances); + virtual GLenum applyIndexBuffer(const GLvoid *indices, gl::Buffer *elementArrayBuffer, GLsizei count, GLenum mode, GLenum type, TranslatedIndexData *indexInfo); + + virtual void drawArrays(GLenum mode, GLsizei count, GLsizei instances); + virtual void drawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices, gl::Buffer *elementArrayBuffer, const TranslatedIndexData &indexInfo, GLsizei instances); + + virtual void clear(const gl::ClearParameters &clearParams, gl::Framebuffer *frameBuffer); + + virtual void markAllStateDirty(); + + // lost device + void notifyDeviceLost(); + virtual bool isDeviceLost(); + virtual bool testDeviceLost(bool notify); + virtual bool testDeviceResettable(); + + // Renderer capabilities + virtual DWORD getAdapterVendor() const; + virtual std::string getRendererDescription() const; + virtual GUID getAdapterIdentifier() const; + + virtual bool getBGRATextureSupport() const; + virtual bool getDXT1TextureSupport(); + virtual bool getDXT3TextureSupport(); + virtual bool getDXT5TextureSupport(); + virtual bool getEventQuerySupport(); + virtual bool getFloat32TextureSupport(bool *filtering, bool *renderable); + virtual bool getFloat16TextureSupport(bool *filtering, bool *renderable); + virtual bool getLuminanceTextureSupport(); + virtual bool getLuminanceAlphaTextureSupport(); + virtual unsigned int getMaxVertexTextureImageUnits() const; + virtual unsigned int getMaxCombinedTextureImageUnits() const; + virtual unsigned int getReservedVertexUniformVectors() const; + virtual unsigned int getReservedFragmentUniformVectors() const; + virtual unsigned int getMaxVertexUniformVectors() const; + virtual unsigned int getMaxFragmentUniformVectors() const; + virtual unsigned int getMaxVaryingVectors() const; + virtual bool getNonPower2TextureSupport() const; + virtual bool getDepthTextureSupport() const; + virtual bool getOcclusionQuerySupport() const; + virtual bool getInstancingSupport() const; + virtual bool getTextureFilterAnisotropySupport() const; + virtual float getTextureMaxAnisotropy() const; + virtual bool getShareHandleSupport() const; + virtual bool getDerivativeInstructionSupport() const; + virtual bool getPostSubBufferSupport() const; + + virtual int getMajorShaderModel() const; + virtual float getMaxPointSize() const; + virtual int getMaxViewportDimension() const; + virtual int getMaxTextureWidth() const; + virtual int getMaxTextureHeight() const; + virtual bool get32BitIndexSupport() const; + virtual int getMinSwapInterval() const; + virtual int getMaxSwapInterval() const; + + virtual GLsizei getMaxSupportedSamples() const; + int getNearestSupportedSamples(DXGI_FORMAT format, unsigned int requested) const; + + virtual unsigned int getMaxRenderTargets() const; + + // Pixel operations + virtual bool copyToRenderTarget(TextureStorageInterface2D *dest, TextureStorageInterface2D *source); + virtual bool copyToRenderTarget(TextureStorageInterfaceCube *dest, TextureStorageInterfaceCube *source); + + virtual bool copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterface2D *storage, GLint level); + virtual bool copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterfaceCube *storage, GLenum target, GLint level); + + bool copyTexture(ID3D11ShaderResourceView *source, const gl::Rectangle &sourceArea, unsigned int sourceWidth, unsigned int sourceHeight, + ID3D11RenderTargetView *dest, const gl::Rectangle &destArea, unsigned int destWidth, unsigned int destHeight, GLenum destFormat); + + virtual bool blitRect(gl::Framebuffer *readTarget, const gl::Rectangle &readRect, gl::Framebuffer *drawTarget, const gl::Rectangle &drawRect, + bool blitRenderTarget, bool blitDepthStencil); + virtual void readPixels(gl::Framebuffer *framebuffer, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, + GLsizei outputPitch, bool packReverseRowOrder, GLint packAlignment, void* pixels); + + // RenderTarget creation + virtual RenderTarget *createRenderTarget(SwapChain *swapChain, bool depth); + virtual RenderTarget *createRenderTarget(int width, int height, GLenum format, GLsizei samples, bool depth); + + // Shader operations + virtual ShaderExecutable *loadExecutable(const void *function, size_t length, rx::ShaderType type); + virtual ShaderExecutable *compileToExecutable(gl::InfoLog &infoLog, const char *shaderHLSL, rx::ShaderType type); + + // Image operations + virtual Image *createImage(); + virtual void generateMipmap(Image *dest, Image *source); + virtual TextureStorage *createTextureStorage2D(SwapChain *swapChain); + virtual TextureStorage *createTextureStorage2D(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height); + virtual TextureStorage *createTextureStorageCube(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size); + + // Buffer creation + virtual VertexBuffer *createVertexBuffer(); + virtual IndexBuffer *createIndexBuffer(); + virtual BufferStorage *createBufferStorage(); + + // Query and Fence creation + virtual QueryImpl *createQuery(GLenum type); + virtual FenceImpl *createFence(); + + // D3D11-renderer specific methods + ID3D11Device *getDevice() { return mDevice; } + ID3D11DeviceContext *getDeviceContext() { return mDeviceContext; }; + IDXGIFactory *getDxgiFactory() { return mDxgiFactory; }; + + bool getRenderTargetResource(gl::Renderbuffer *colorbuffer, unsigned int *subresourceIndex, ID3D11Texture2D **resource); + void unapplyRenderTargets(); + void setOneTimeRenderTarget(ID3D11RenderTargetView *renderTargetView); + + private: + DISALLOW_COPY_AND_ASSIGN(Renderer11); + + void drawLineLoop(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer); + void drawTriangleFan(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer, int instances); + + void readTextureData(ID3D11Texture2D *texture, unsigned int subResource, const gl::Rectangle &area, + GLenum format, GLenum type, GLsizei outputPitch, bool packReverseRowOrder, + GLint packAlignment, void *pixels); + + void maskedClear(const gl::ClearParameters &clearParams); + rx::Range getViewportBounds() const; + + bool blitRenderbufferRect(const gl::Rectangle &readRect, const gl::Rectangle &drawRect, RenderTarget *readRenderTarget, + RenderTarget *drawRenderTarget, bool wholeBufferCopy); + ID3D11Texture2D *resolveMultisampledTexture(ID3D11Texture2D *source, unsigned int subresource); + + HMODULE mD3d11Module; + HMODULE mDxgiModule; + HDC mDc; + + bool mDeviceLost; + + void initializeDevice(); + void releaseDeviceResources(); + int getMinorShaderModel() const; + void release(); + + RenderStateCache mStateCache; + + // Support flags + bool mFloat16TextureSupport; + bool mFloat16FilterSupport; + bool mFloat16RenderSupport; + + bool mFloat32TextureSupport; + bool mFloat32FilterSupport; + bool mFloat32RenderSupport; + + bool mDXT1TextureSupport; + bool mDXT3TextureSupport; + bool mDXT5TextureSupport; + + bool mDepthTextureSupport; + + // Multisample format support + struct MultisampleSupportInfo + { + unsigned int qualityLevels[D3D11_MAX_MULTISAMPLE_SAMPLE_COUNT]; + }; + + typedef std::unordered_map<DXGI_FORMAT, MultisampleSupportInfo> MultisampleSupportMap; + MultisampleSupportMap mMultisampleSupportMap; + + unsigned int mMaxSupportedSamples; + + // current render target states + unsigned int mAppliedRenderTargetSerials[gl::IMPLEMENTATION_MAX_DRAW_BUFFERS]; + unsigned int mAppliedDepthbufferSerial; + unsigned int mAppliedStencilbufferSerial; + bool mDepthStencilInitialized; + bool mRenderTargetDescInitialized; + rx::RenderTarget::Desc mRenderTargetDesc; + unsigned int mCurDepthSize; + unsigned int mCurStencilSize; + + // Currently applied sampler states + bool mForceSetVertexSamplerStates[gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS]; + gl::SamplerState mCurVertexSamplerStates[gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS]; + + bool mForceSetPixelSamplerStates[gl::MAX_TEXTURE_IMAGE_UNITS]; + gl::SamplerState mCurPixelSamplerStates[gl::MAX_TEXTURE_IMAGE_UNITS]; + + // Currently applied textures + unsigned int mCurVertexTextureSerials[gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS]; + unsigned int mCurPixelTextureSerials[gl::MAX_TEXTURE_IMAGE_UNITS]; + + // Currently applied blend state + bool mForceSetBlendState; + gl::BlendState mCurBlendState; + gl::Color mCurBlendColor; + unsigned int mCurSampleMask; + + // Currently applied rasterizer state + bool mForceSetRasterState; + gl::RasterizerState mCurRasterState; + + // Currently applied depth stencil state + bool mForceSetDepthStencilState; + gl::DepthStencilState mCurDepthStencilState; + int mCurStencilRef; + int mCurStencilBackRef; + + // Currently applied scissor rectangle + bool mForceSetScissor; + bool mScissorEnabled; + gl::Rectangle mCurScissor; + + // Currently applied viewport + bool mForceSetViewport; + gl::Rectangle mCurViewport; + float mCurNear; + float mCurFar; + + unsigned int mAppliedIBSerial; + unsigned int mAppliedStorageIBSerial; + unsigned int mAppliedIBOffset; + + unsigned int mAppliedProgramBinarySerial; + bool mIsGeometryShaderActive; + + dx_VertexConstants mVertexConstants; + dx_VertexConstants mAppliedVertexConstants; + ID3D11Buffer *mDriverConstantBufferVS; + + dx_PixelConstants mPixelConstants; + dx_PixelConstants mAppliedPixelConstants; + ID3D11Buffer *mDriverConstantBufferPS; + + // Vertex, index and input layouts + VertexDataManager *mVertexDataManager; + IndexDataManager *mIndexDataManager; + InputLayoutCache mInputLayoutCache; + + StreamingIndexBufferInterface *mLineLoopIB; + StreamingIndexBufferInterface *mTriangleFanIB; + + // Texture copy resources + bool mCopyResourcesInitialized; + ID3D11Buffer *mCopyVB; + ID3D11SamplerState *mCopySampler; + ID3D11InputLayout *mCopyIL; + ID3D11VertexShader *mCopyVS; + ID3D11PixelShader *mCopyRGBAPS; + ID3D11PixelShader *mCopyRGBPS; + ID3D11PixelShader *mCopyLumPS; + ID3D11PixelShader *mCopyLumAlphaPS; + + // Masked clear resources + bool mClearResourcesInitialized; + ID3D11Buffer *mClearVB; + ID3D11InputLayout *mClearIL; + ID3D11VertexShader *mClearVS; + ID3D11PixelShader *mClearPS; + ID3D11RasterizerState *mClearScissorRS; + ID3D11RasterizerState *mClearNoScissorRS; + + // Sync query + ID3D11Query *mSyncQuery; + + ID3D11Device *mDevice; + D3D_FEATURE_LEVEL mFeatureLevel; + ID3D11DeviceContext *mDeviceContext; + IDXGIAdapter *mDxgiAdapter; + DXGI_ADAPTER_DESC mAdapterDescription; + char mDescription[128]; + IDXGIFactory *mDxgiFactory; + + // Cached device caps + bool mBGRATextureSupport; +}; + +} +#endif // LIBGLESV2_RENDERER_RENDERER11_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Renderer9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer9.cpp new file mode 100644 index 0000000000..8acbce4be7 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer9.cpp @@ -0,0 +1,3211 @@ +#include "precompiled.h" +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Renderer9.cpp: Implements a back-end specific class for the D3D9 renderer. + +#include "libGLESv2/main.h" +#include "libGLESv2/Buffer.h" +#include "libGLESv2/Texture.h" +#include "libGLESv2/Framebuffer.h" +#include "libGLESv2/Renderbuffer.h" +#include "libGLESv2/ProgramBinary.h" +#include "libGLESv2/renderer/IndexDataManager.h" +#include "libGLESv2/renderer/Renderer9.h" +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/renderer/ShaderExecutable9.h" +#include "libGLESv2/renderer/SwapChain9.h" +#include "libGLESv2/renderer/TextureStorage9.h" +#include "libGLESv2/renderer/Image9.h" +#include "libGLESv2/renderer/Blit.h" +#include "libGLESv2/renderer/RenderTarget9.h" +#include "libGLESv2/renderer/VertexBuffer9.h" +#include "libGLESv2/renderer/IndexBuffer9.h" +#include "libGLESv2/renderer/BufferStorage9.h" +#include "libGLESv2/renderer/Query9.h" +#include "libGLESv2/renderer/Fence9.h" + +#include "libEGL/Display.h" + +// Can also be enabled by defining FORCE_REF_RAST in the project's predefined macros +#define REF_RAST 0 + +// The "Debug This Pixel..." feature in PIX often fails when using the +// D3D9Ex interfaces. In order to get debug pixel to work on a Vista/Win 7 +// machine, define "ANGLE_ENABLE_D3D9EX=0" in your project file. +#if !defined(ANGLE_ENABLE_D3D9EX) +// Enables use of the IDirect3D9Ex interface, when available +#define ANGLE_ENABLE_D3D9EX 1 +#endif // !defined(ANGLE_ENABLE_D3D9EX) + +#if !defined(ANGLE_COMPILE_OPTIMIZATION_LEVEL) +#define ANGLE_COMPILE_OPTIMIZATION_LEVEL D3DCOMPILE_OPTIMIZATION_LEVEL3 +#endif + +namespace rx +{ +static const D3DFORMAT RenderTargetFormats[] = + { + D3DFMT_A1R5G5B5, + // D3DFMT_A2R10G10B10, // The color_ramp conformance test uses ReadPixels with UNSIGNED_BYTE causing it to think that rendering skipped a colour value. + D3DFMT_A8R8G8B8, + D3DFMT_R5G6B5, + // D3DFMT_X1R5G5B5, // Has no compatible OpenGL ES renderbuffer format + D3DFMT_X8R8G8B8 + }; + +static const D3DFORMAT DepthStencilFormats[] = + { + D3DFMT_UNKNOWN, + // D3DFMT_D16_LOCKABLE, + D3DFMT_D32, + // D3DFMT_D15S1, + D3DFMT_D24S8, + D3DFMT_D24X8, + // D3DFMT_D24X4S4, + D3DFMT_D16, + // D3DFMT_D32F_LOCKABLE, + // D3DFMT_D24FS8 + }; + +enum +{ + MAX_VERTEX_CONSTANT_VECTORS_D3D9 = 256, + MAX_PIXEL_CONSTANT_VECTORS_SM2 = 32, + MAX_PIXEL_CONSTANT_VECTORS_SM3 = 224, + MAX_VARYING_VECTORS_SM2 = 8, + MAX_VARYING_VECTORS_SM3 = 10, + + MAX_TEXTURE_IMAGE_UNITS_VTF_SM3 = 4 +}; + +Renderer9::Renderer9(egl::Display *display, HDC hDc, bool softwareDevice) : Renderer(display), mDc(hDc), mSoftwareDevice(softwareDevice) +{ + mD3d9Module = NULL; + + mD3d9 = NULL; + mD3d9Ex = NULL; + mDevice = NULL; + mDeviceEx = NULL; + mDeviceWindow = NULL; + mBlit = NULL; + + mAdapter = D3DADAPTER_DEFAULT; + + #if REF_RAST == 1 || defined(FORCE_REF_RAST) + mDeviceType = D3DDEVTYPE_REF; + #else + mDeviceType = D3DDEVTYPE_HAL; + #endif + + mDeviceLost = false; + + mMaxSupportedSamples = 0; + + mMaskedClearSavedState = NULL; + + mVertexDataManager = NULL; + mIndexDataManager = NULL; + mLineLoopIB = NULL; + + mMaxNullColorbufferLRU = 0; + for (int i = 0; i < NUM_NULL_COLORBUFFER_CACHE_ENTRIES; i++) + { + mNullColorbufferCache[i].lruCount = 0; + mNullColorbufferCache[i].width = 0; + mNullColorbufferCache[i].height = 0; + mNullColorbufferCache[i].buffer = NULL; + } +} + +Renderer9::~Renderer9() +{ + releaseDeviceResources(); + + if (mDevice) + { + // If the device is lost, reset it first to prevent leaving the driver in an unstable state + if (testDeviceLost(false)) + { + resetDevice(); + } + + mDevice->Release(); + mDevice = NULL; + } + + if (mDeviceEx) + { + mDeviceEx->Release(); + mDeviceEx = NULL; + } + + if (mD3d9) + { + mD3d9->Release(); + mD3d9 = NULL; + } + + if (mDeviceWindow) + { + DestroyWindow(mDeviceWindow); + mDeviceWindow = NULL; + } + + if (mD3d9Ex) + { + mD3d9Ex->Release(); + mD3d9Ex = NULL; + } + + if (mD3d9Module) + { + mD3d9Module = NULL; + } + + while (!mMultiSampleSupport.empty()) + { + delete [] mMultiSampleSupport.begin()->second; + mMultiSampleSupport.erase(mMultiSampleSupport.begin()); + } +} + +Renderer9 *Renderer9::makeRenderer9(Renderer *renderer) +{ + ASSERT(HAS_DYNAMIC_TYPE(rx::Renderer9*, renderer)); + return static_cast<rx::Renderer9*>(renderer); +} + +EGLint Renderer9::initialize() +{ + if (!initializeCompiler()) + { + return EGL_NOT_INITIALIZED; + } + + if (mSoftwareDevice) + { + mD3d9Module = GetModuleHandle(TEXT("swiftshader_d3d9.dll")); + } + else + { + mD3d9Module = GetModuleHandle(TEXT("d3d9.dll")); + } + + if (mD3d9Module == NULL) + { + ERR("No D3D9 module found - aborting!\n"); + return EGL_NOT_INITIALIZED; + } + + typedef HRESULT (WINAPI *Direct3DCreate9ExFunc)(UINT, IDirect3D9Ex**); + Direct3DCreate9ExFunc Direct3DCreate9ExPtr = reinterpret_cast<Direct3DCreate9ExFunc>(GetProcAddress(mD3d9Module, "Direct3DCreate9Ex")); + + // Use Direct3D9Ex if available. Among other things, this version is less + // inclined to report a lost context, for example when the user switches + // desktop. Direct3D9Ex is available in Windows Vista and later if suitable drivers are available. + if (ANGLE_ENABLE_D3D9EX && Direct3DCreate9ExPtr && SUCCEEDED(Direct3DCreate9ExPtr(D3D_SDK_VERSION, &mD3d9Ex))) + { + ASSERT(mD3d9Ex); + mD3d9Ex->QueryInterface(IID_IDirect3D9, reinterpret_cast<void**>(&mD3d9)); + ASSERT(mD3d9); + } + else + { + mD3d9 = Direct3DCreate9(D3D_SDK_VERSION); + } + + if (!mD3d9) + { + ERR("Could not create D3D9 device - aborting!\n"); + return EGL_NOT_INITIALIZED; + } + + if (mDc != NULL) + { + // UNIMPLEMENTED(); // FIXME: Determine which adapter index the device context corresponds to + } + + HRESULT result; + + // Give up on getting device caps after about one second. + for (int i = 0; i < 10; ++i) + { + result = mD3d9->GetDeviceCaps(mAdapter, mDeviceType, &mDeviceCaps); + if (SUCCEEDED(result)) + { + break; + } + else if (result == D3DERR_NOTAVAILABLE) + { + Sleep(100); // Give the driver some time to initialize/recover + } + else if (FAILED(result)) // D3DERR_OUTOFVIDEOMEMORY, E_OUTOFMEMORY, D3DERR_INVALIDDEVICE, or another error we can't recover from + { + ERR("failed to get device caps (0x%x)\n", result); + return EGL_NOT_INITIALIZED; + } + } + + if (mDeviceCaps.PixelShaderVersion < D3DPS_VERSION(2, 0)) + { + ERR("Renderer does not support PS 2.0. aborting!\n"); + return EGL_NOT_INITIALIZED; + } + + // When DirectX9 is running with an older DirectX8 driver, a StretchRect from a regular texture to a render target texture is not supported. + // This is required by Texture2D::convertToRenderTarget. + if ((mDeviceCaps.DevCaps2 & D3DDEVCAPS2_CAN_STRETCHRECT_FROM_TEXTURES) == 0) + { + ERR("Renderer does not support stretctrect from textures!\n"); + return EGL_NOT_INITIALIZED; + } + + mD3d9->GetAdapterIdentifier(mAdapter, 0, &mAdapterIdentifier); + + // ATI cards on XP have problems with non-power-of-two textures. + mSupportsNonPower2Textures = !(mDeviceCaps.TextureCaps & D3DPTEXTURECAPS_POW2) && + !(mDeviceCaps.TextureCaps & D3DPTEXTURECAPS_CUBEMAP_POW2) && + !(mDeviceCaps.TextureCaps & D3DPTEXTURECAPS_NONPOW2CONDITIONAL) && + !(getComparableOSVersion() < versionWindowsVista && mAdapterIdentifier.VendorId == VENDOR_ID_AMD); + + // Must support a minimum of 2:1 anisotropy for max anisotropy to be considered supported, per the spec + mSupportsTextureFilterAnisotropy = ((mDeviceCaps.RasterCaps & D3DPRASTERCAPS_ANISOTROPY) && (mDeviceCaps.MaxAnisotropy >= 2)); + + mMinSwapInterval = 4; + mMaxSwapInterval = 0; + + if (mDeviceCaps.PresentationIntervals & D3DPRESENT_INTERVAL_IMMEDIATE) + { + mMinSwapInterval = std::min(mMinSwapInterval, 0); + mMaxSwapInterval = std::max(mMaxSwapInterval, 0); + } + if (mDeviceCaps.PresentationIntervals & D3DPRESENT_INTERVAL_ONE) + { + mMinSwapInterval = std::min(mMinSwapInterval, 1); + mMaxSwapInterval = std::max(mMaxSwapInterval, 1); + } + if (mDeviceCaps.PresentationIntervals & D3DPRESENT_INTERVAL_TWO) + { + mMinSwapInterval = std::min(mMinSwapInterval, 2); + mMaxSwapInterval = std::max(mMaxSwapInterval, 2); + } + if (mDeviceCaps.PresentationIntervals & D3DPRESENT_INTERVAL_THREE) + { + mMinSwapInterval = std::min(mMinSwapInterval, 3); + mMaxSwapInterval = std::max(mMaxSwapInterval, 3); + } + if (mDeviceCaps.PresentationIntervals & D3DPRESENT_INTERVAL_FOUR) + { + mMinSwapInterval = std::min(mMinSwapInterval, 4); + mMaxSwapInterval = std::max(mMaxSwapInterval, 4); + } + + int max = 0; + for (unsigned int i = 0; i < ArraySize(RenderTargetFormats); ++i) + { + bool *multisampleArray = new bool[D3DMULTISAMPLE_16_SAMPLES + 1]; + getMultiSampleSupport(RenderTargetFormats[i], multisampleArray); + mMultiSampleSupport[RenderTargetFormats[i]] = multisampleArray; + + for (int j = D3DMULTISAMPLE_16_SAMPLES; j >= 0; --j) + { + if (multisampleArray[j] && j != D3DMULTISAMPLE_NONMASKABLE && j > max) + { + max = j; + } + } + } + + for (unsigned int i = 0; i < ArraySize(DepthStencilFormats); ++i) + { + if (DepthStencilFormats[i] == D3DFMT_UNKNOWN) + continue; + + bool *multisampleArray = new bool[D3DMULTISAMPLE_16_SAMPLES + 1]; + getMultiSampleSupport(DepthStencilFormats[i], multisampleArray); + mMultiSampleSupport[DepthStencilFormats[i]] = multisampleArray; + + for (int j = D3DMULTISAMPLE_16_SAMPLES; j >= 0; --j) + { + if (multisampleArray[j] && j != D3DMULTISAMPLE_NONMASKABLE && j > max) + { + max = j; + } + } + } + + mMaxSupportedSamples = max; + + static const TCHAR windowName[] = TEXT("AngleHiddenWindow"); + static const TCHAR className[] = TEXT("STATIC"); + + mDeviceWindow = CreateWindowEx(WS_EX_NOACTIVATE, className, windowName, WS_DISABLED | WS_POPUP, 0, 0, 1, 1, HWND_MESSAGE, NULL, GetModuleHandle(NULL), NULL); + + D3DPRESENT_PARAMETERS presentParameters = getDefaultPresentParameters(); + DWORD behaviorFlags = D3DCREATE_FPU_PRESERVE | D3DCREATE_NOWINDOWCHANGES; + + result = mD3d9->CreateDevice(mAdapter, mDeviceType, mDeviceWindow, behaviorFlags | D3DCREATE_HARDWARE_VERTEXPROCESSING | D3DCREATE_PUREDEVICE, &presentParameters, &mDevice); + if (result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY || result == D3DERR_DEVICELOST) + { + return EGL_BAD_ALLOC; + } + + if (FAILED(result)) + { + result = mD3d9->CreateDevice(mAdapter, mDeviceType, mDeviceWindow, behaviorFlags | D3DCREATE_SOFTWARE_VERTEXPROCESSING, &presentParameters, &mDevice); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY || result == D3DERR_NOTAVAILABLE || result == D3DERR_DEVICELOST); + return EGL_BAD_ALLOC; + } + } + + if (mD3d9Ex) + { + result = mDevice->QueryInterface(IID_IDirect3DDevice9Ex, (void**) &mDeviceEx); + ASSERT(SUCCEEDED(result)); + } + + mVertexShaderCache.initialize(mDevice); + mPixelShaderCache.initialize(mDevice); + + // Check occlusion query support + IDirect3DQuery9 *occlusionQuery = NULL; + if (SUCCEEDED(mDevice->CreateQuery(D3DQUERYTYPE_OCCLUSION, &occlusionQuery)) && occlusionQuery) + { + occlusionQuery->Release(); + mOcclusionQuerySupport = true; + } + else + { + mOcclusionQuerySupport = false; + } + + // Check event query support + IDirect3DQuery9 *eventQuery = NULL; + if (SUCCEEDED(mDevice->CreateQuery(D3DQUERYTYPE_EVENT, &eventQuery)) && eventQuery) + { + eventQuery->Release(); + mEventQuerySupport = true; + } + else + { + mEventQuerySupport = false; + } + + D3DDISPLAYMODE currentDisplayMode; + mD3d9->GetAdapterDisplayMode(mAdapter, ¤tDisplayMode); + + // Check vertex texture support + // Only Direct3D 10 ready devices support all the necessary vertex texture formats. + // We test this using D3D9 by checking support for the R16F format. + mVertexTextureSupport = mDeviceCaps.PixelShaderVersion >= D3DPS_VERSION(3, 0) && + SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, + D3DUSAGE_QUERY_VERTEXTEXTURE, D3DRTYPE_TEXTURE, D3DFMT_R16F)); + + // Check depth texture support + // we use INTZ for depth textures in Direct3D9 + // we also want NULL texture support to ensure the we can make depth-only FBOs + // see http://aras-p.info/texts/D3D9GPUHacks.html + mDepthTextureSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, + D3DUSAGE_DEPTHSTENCIL, D3DRTYPE_TEXTURE, D3DFMT_INTZ)) && + SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, + D3DUSAGE_RENDERTARGET, D3DRTYPE_SURFACE, D3DFMT_NULL)); + + // Check 32 bit floating point texture support + mFloat32FilterSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_QUERY_FILTER, + D3DRTYPE_TEXTURE, D3DFMT_A32B32G32R32F)) && + SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_QUERY_FILTER, + D3DRTYPE_CUBETEXTURE, D3DFMT_A32B32G32R32F)); + + mFloat32RenderSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_RENDERTARGET, + D3DRTYPE_TEXTURE, D3DFMT_A32B32G32R32F)) && + SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_RENDERTARGET, + D3DRTYPE_CUBETEXTURE, D3DFMT_A32B32G32R32F)); + + if (!mFloat32FilterSupport && !mFloat32RenderSupport) + { + mFloat32TextureSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, + D3DRTYPE_TEXTURE, D3DFMT_A32B32G32R32F)) && + SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, + D3DRTYPE_CUBETEXTURE, D3DFMT_A32B32G32R32F)); + } + else + { + mFloat32TextureSupport = true; + } + + // Check 16 bit floating point texture support + mFloat16FilterSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_QUERY_FILTER, + D3DRTYPE_TEXTURE, D3DFMT_A16B16G16R16F)) && + SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_QUERY_FILTER, + D3DRTYPE_CUBETEXTURE, D3DFMT_A16B16G16R16F)); + + mFloat16RenderSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_RENDERTARGET, + D3DRTYPE_TEXTURE, D3DFMT_A16B16G16R16F)) && + SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_RENDERTARGET, + D3DRTYPE_CUBETEXTURE, D3DFMT_A16B16G16R16F)); + + if (!mFloat16FilterSupport && !mFloat16RenderSupport) + { + mFloat16TextureSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, + D3DRTYPE_TEXTURE, D3DFMT_A16B16G16R16F)) && + SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, + D3DRTYPE_CUBETEXTURE, D3DFMT_A16B16G16R16F)); + } + else + { + mFloat16TextureSupport = true; + } + + // Check DXT texture support + mDXT1TextureSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, D3DRTYPE_TEXTURE, D3DFMT_DXT1)); + mDXT3TextureSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, D3DRTYPE_TEXTURE, D3DFMT_DXT3)); + mDXT5TextureSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, D3DRTYPE_TEXTURE, D3DFMT_DXT5)); + + // Check luminance[alpha] texture support + mLuminanceTextureSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, D3DRTYPE_TEXTURE, D3DFMT_L8)); + mLuminanceAlphaTextureSupport = SUCCEEDED(mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, 0, D3DRTYPE_TEXTURE, D3DFMT_A8L8)); + + initializeDevice(); + + return EGL_SUCCESS; +} + +// do any one-time device initialization +// NOTE: this is also needed after a device lost/reset +// to reset the scene status and ensure the default states are reset. +void Renderer9::initializeDevice() +{ + // Permanent non-default states + mDevice->SetRenderState(D3DRS_POINTSPRITEENABLE, TRUE); + mDevice->SetRenderState(D3DRS_LASTPIXEL, FALSE); + + if (mDeviceCaps.PixelShaderVersion >= D3DPS_VERSION(3, 0)) + { + mDevice->SetRenderState(D3DRS_POINTSIZE_MAX, (DWORD&)mDeviceCaps.MaxPointSize); + } + else + { + mDevice->SetRenderState(D3DRS_POINTSIZE_MAX, 0x3F800000); // 1.0f + } + + markAllStateDirty(); + + mSceneStarted = false; + + ASSERT(!mBlit && !mVertexDataManager && !mIndexDataManager); + mBlit = new Blit(this); + mVertexDataManager = new rx::VertexDataManager(this); + mIndexDataManager = new rx::IndexDataManager(this); +} + +D3DPRESENT_PARAMETERS Renderer9::getDefaultPresentParameters() +{ + D3DPRESENT_PARAMETERS presentParameters = {0}; + + // The default swap chain is never actually used. Surface will create a new swap chain with the proper parameters. + presentParameters.AutoDepthStencilFormat = D3DFMT_UNKNOWN; + presentParameters.BackBufferCount = 1; + presentParameters.BackBufferFormat = D3DFMT_UNKNOWN; + presentParameters.BackBufferWidth = 1; + presentParameters.BackBufferHeight = 1; + presentParameters.EnableAutoDepthStencil = FALSE; + presentParameters.Flags = 0; + presentParameters.hDeviceWindow = mDeviceWindow; + presentParameters.MultiSampleQuality = 0; + presentParameters.MultiSampleType = D3DMULTISAMPLE_NONE; + presentParameters.PresentationInterval = D3DPRESENT_INTERVAL_DEFAULT; + presentParameters.SwapEffect = D3DSWAPEFFECT_DISCARD; + presentParameters.Windowed = TRUE; + + return presentParameters; +} + +int Renderer9::generateConfigs(ConfigDesc **configDescList) +{ + D3DDISPLAYMODE currentDisplayMode; + mD3d9->GetAdapterDisplayMode(mAdapter, ¤tDisplayMode); + + unsigned int numRenderFormats = ArraySize(RenderTargetFormats); + unsigned int numDepthFormats = ArraySize(DepthStencilFormats); + (*configDescList) = new ConfigDesc[numRenderFormats * numDepthFormats]; + int numConfigs = 0; + + for (unsigned int formatIndex = 0; formatIndex < numRenderFormats; formatIndex++) + { + D3DFORMAT renderTargetFormat = RenderTargetFormats[formatIndex]; + + HRESULT result = mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_RENDERTARGET, D3DRTYPE_SURFACE, renderTargetFormat); + + if (SUCCEEDED(result)) + { + for (unsigned int depthStencilIndex = 0; depthStencilIndex < numDepthFormats; depthStencilIndex++) + { + D3DFORMAT depthStencilFormat = DepthStencilFormats[depthStencilIndex]; + HRESULT result = D3D_OK; + + if(depthStencilFormat != D3DFMT_UNKNOWN) + { + result = mD3d9->CheckDeviceFormat(mAdapter, mDeviceType, currentDisplayMode.Format, D3DUSAGE_DEPTHSTENCIL, D3DRTYPE_SURFACE, depthStencilFormat); + } + + if (SUCCEEDED(result)) + { + if(depthStencilFormat != D3DFMT_UNKNOWN) + { + result = mD3d9->CheckDepthStencilMatch(mAdapter, mDeviceType, currentDisplayMode.Format, renderTargetFormat, depthStencilFormat); + } + + if (SUCCEEDED(result)) + { + ConfigDesc newConfig; + newConfig.renderTargetFormat = d3d9_gl::ConvertBackBufferFormat(renderTargetFormat); + newConfig.depthStencilFormat = d3d9_gl::ConvertDepthStencilFormat(depthStencilFormat); + newConfig.multiSample = 0; // FIXME: enumerate multi-sampling + newConfig.fastConfig = (currentDisplayMode.Format == renderTargetFormat); + + (*configDescList)[numConfigs++] = newConfig; + } + } + } + } + } + + return numConfigs; +} + +void Renderer9::deleteConfigs(ConfigDesc *configDescList) +{ + delete [] (configDescList); +} + +void Renderer9::startScene() +{ + if (!mSceneStarted) + { + long result = mDevice->BeginScene(); + if (SUCCEEDED(result)) { + // This is defensive checking against the device being + // lost at unexpected times. + mSceneStarted = true; + } + } +} + +void Renderer9::endScene() +{ + if (mSceneStarted) + { + // EndScene can fail if the device was lost, for example due + // to a TDR during a draw call. + mDevice->EndScene(); + mSceneStarted = false; + } +} + +void Renderer9::sync(bool block) +{ + HRESULT result; + + IDirect3DQuery9* query = allocateEventQuery(); + if (!query) + { + return; + } + + result = query->Issue(D3DISSUE_END); + ASSERT(SUCCEEDED(result)); + + do + { + result = query->GetData(NULL, 0, D3DGETDATA_FLUSH); + + if(block && result == S_FALSE) + { + // Keep polling, but allow other threads to do something useful first + Sleep(0); + // explicitly check for device loss + // some drivers seem to return S_FALSE even if the device is lost + // instead of D3DERR_DEVICELOST like they should + if (testDeviceLost(false)) + { + result = D3DERR_DEVICELOST; + } + } + } + while(block && result == S_FALSE); + + freeEventQuery(query); + + if (d3d9::isDeviceLostError(result)) + { + notifyDeviceLost(); + } +} + +SwapChain *Renderer9::createSwapChain(HWND window, HANDLE shareHandle, GLenum backBufferFormat, GLenum depthBufferFormat) +{ + return new rx::SwapChain9(this, window, shareHandle, backBufferFormat, depthBufferFormat); +} + +IDirect3DQuery9* Renderer9::allocateEventQuery() +{ + IDirect3DQuery9 *query = NULL; + + if (mEventQueryPool.empty()) + { + HRESULT result = mDevice->CreateQuery(D3DQUERYTYPE_EVENT, &query); + ASSERT(SUCCEEDED(result)); + } + else + { + query = mEventQueryPool.back(); + mEventQueryPool.pop_back(); + } + + return query; +} + +void Renderer9::freeEventQuery(IDirect3DQuery9* query) +{ + if (mEventQueryPool.size() > 1000) + { + query->Release(); + } + else + { + mEventQueryPool.push_back(query); + } +} + +IDirect3DVertexShader9 *Renderer9::createVertexShader(const DWORD *function, size_t length) +{ + return mVertexShaderCache.create(function, length); +} + +IDirect3DPixelShader9 *Renderer9::createPixelShader(const DWORD *function, size_t length) +{ + return mPixelShaderCache.create(function, length); +} + +HRESULT Renderer9::createVertexBuffer(UINT Length, DWORD Usage, IDirect3DVertexBuffer9 **ppVertexBuffer) +{ + D3DPOOL Pool = getBufferPool(Usage); + return mDevice->CreateVertexBuffer(Length, Usage, 0, Pool, ppVertexBuffer, NULL); +} + +VertexBuffer *Renderer9::createVertexBuffer() +{ + return new VertexBuffer9(this); +} + +HRESULT Renderer9::createIndexBuffer(UINT Length, DWORD Usage, D3DFORMAT Format, IDirect3DIndexBuffer9 **ppIndexBuffer) +{ + D3DPOOL Pool = getBufferPool(Usage); + return mDevice->CreateIndexBuffer(Length, Usage, Format, Pool, ppIndexBuffer, NULL); +} + +IndexBuffer *Renderer9::createIndexBuffer() +{ + return new IndexBuffer9(this); +} + +BufferStorage *Renderer9::createBufferStorage() +{ + return new BufferStorage9(); +} + +QueryImpl *Renderer9::createQuery(GLenum type) +{ + return new Query9(this, type); +} + +FenceImpl *Renderer9::createFence() +{ + return new Fence9(this); +} + +void Renderer9::setSamplerState(gl::SamplerType type, int index, const gl::SamplerState &samplerState) +{ + bool *forceSetSamplers = (type == gl::SAMPLER_PIXEL) ? mForceSetPixelSamplerStates : mForceSetVertexSamplerStates; + gl::SamplerState *appliedSamplers = (type == gl::SAMPLER_PIXEL) ? mCurPixelSamplerStates: mCurVertexSamplerStates; + + if (forceSetSamplers[index] || memcmp(&samplerState, &appliedSamplers[index], sizeof(gl::SamplerState)) != 0) + { + int d3dSamplerOffset = (type == gl::SAMPLER_PIXEL) ? 0 : D3DVERTEXTEXTURESAMPLER0; + int d3dSampler = index + d3dSamplerOffset; + + mDevice->SetSamplerState(d3dSampler, D3DSAMP_ADDRESSU, gl_d3d9::ConvertTextureWrap(samplerState.wrapS)); + mDevice->SetSamplerState(d3dSampler, D3DSAMP_ADDRESSV, gl_d3d9::ConvertTextureWrap(samplerState.wrapT)); + + mDevice->SetSamplerState(d3dSampler, D3DSAMP_MAGFILTER, gl_d3d9::ConvertMagFilter(samplerState.magFilter, samplerState.maxAnisotropy)); + D3DTEXTUREFILTERTYPE d3dMinFilter, d3dMipFilter; + gl_d3d9::ConvertMinFilter(samplerState.minFilter, &d3dMinFilter, &d3dMipFilter, samplerState.maxAnisotropy); + mDevice->SetSamplerState(d3dSampler, D3DSAMP_MINFILTER, d3dMinFilter); + mDevice->SetSamplerState(d3dSampler, D3DSAMP_MIPFILTER, d3dMipFilter); + mDevice->SetSamplerState(d3dSampler, D3DSAMP_MAXMIPLEVEL, samplerState.lodOffset); + if (mSupportsTextureFilterAnisotropy) + { + mDevice->SetSamplerState(d3dSampler, D3DSAMP_MAXANISOTROPY, (DWORD)samplerState.maxAnisotropy); + } + } + + forceSetSamplers[index] = false; + appliedSamplers[index] = samplerState; +} + +void Renderer9::setTexture(gl::SamplerType type, int index, gl::Texture *texture) +{ + int d3dSamplerOffset = (type == gl::SAMPLER_PIXEL) ? 0 : D3DVERTEXTEXTURESAMPLER0; + int d3dSampler = index + d3dSamplerOffset; + IDirect3DBaseTexture9 *d3dTexture = NULL; + unsigned int serial = 0; + bool forceSetTexture = false; + + unsigned int *appliedSerials = (type == gl::SAMPLER_PIXEL) ? mCurPixelTextureSerials : mCurVertexTextureSerials; + + if (texture) + { + TextureStorageInterface *texStorage = texture->getNativeTexture(); + if (texStorage) + { + TextureStorage9 *storage9 = TextureStorage9::makeTextureStorage9(texStorage->getStorageInstance()); + d3dTexture = storage9->getBaseTexture(); + } + // If we get NULL back from getBaseTexture here, something went wrong + // in the texture class and we're unexpectedly missing the d3d texture + ASSERT(d3dTexture != NULL); + + serial = texture->getTextureSerial(); + forceSetTexture = texture->hasDirtyImages(); + } + + if (forceSetTexture || appliedSerials[index] != serial) + { + mDevice->SetTexture(d3dSampler, d3dTexture); + } + + appliedSerials[index] = serial; +} + +void Renderer9::setRasterizerState(const gl::RasterizerState &rasterState) +{ + bool rasterStateChanged = mForceSetRasterState || memcmp(&rasterState, &mCurRasterState, sizeof(gl::RasterizerState)) != 0; + + if (rasterStateChanged) + { + // Set the cull mode + if (rasterState.cullFace) + { + mDevice->SetRenderState(D3DRS_CULLMODE, gl_d3d9::ConvertCullMode(rasterState.cullMode, rasterState.frontFace)); + } + else + { + mDevice->SetRenderState(D3DRS_CULLMODE, D3DCULL_NONE); + } + + if (rasterState.polygonOffsetFill) + { + if (mCurDepthSize > 0) + { + mDevice->SetRenderState(D3DRS_SLOPESCALEDEPTHBIAS, *(DWORD*)&rasterState.polygonOffsetFactor); + + float depthBias = ldexp(rasterState.polygonOffsetUnits, -static_cast<int>(mCurDepthSize)); + mDevice->SetRenderState(D3DRS_DEPTHBIAS, *(DWORD*)&depthBias); + } + } + else + { + mDevice->SetRenderState(D3DRS_SLOPESCALEDEPTHBIAS, 0); + mDevice->SetRenderState(D3DRS_DEPTHBIAS, 0); + } + + mCurRasterState = rasterState; + } + + mForceSetRasterState = false; +} + +void Renderer9::setBlendState(const gl::BlendState &blendState, const gl::Color &blendColor, unsigned int sampleMask) +{ + bool blendStateChanged = mForceSetBlendState || memcmp(&blendState, &mCurBlendState, sizeof(gl::BlendState)) != 0; + bool blendColorChanged = mForceSetBlendState || memcmp(&blendColor, &mCurBlendColor, sizeof(gl::Color)) != 0; + bool sampleMaskChanged = mForceSetBlendState || sampleMask != mCurSampleMask; + + if (blendStateChanged || blendColorChanged) + { + if (blendState.blend) + { + mDevice->SetRenderState(D3DRS_ALPHABLENDENABLE, TRUE); + + if (blendState.sourceBlendRGB != GL_CONSTANT_ALPHA && blendState.sourceBlendRGB != GL_ONE_MINUS_CONSTANT_ALPHA && + blendState.destBlendRGB != GL_CONSTANT_ALPHA && blendState.destBlendRGB != GL_ONE_MINUS_CONSTANT_ALPHA) + { + mDevice->SetRenderState(D3DRS_BLENDFACTOR, gl_d3d9::ConvertColor(blendColor)); + } + else + { + mDevice->SetRenderState(D3DRS_BLENDFACTOR, D3DCOLOR_RGBA(gl::unorm<8>(blendColor.alpha), + gl::unorm<8>(blendColor.alpha), + gl::unorm<8>(blendColor.alpha), + gl::unorm<8>(blendColor.alpha))); + } + + mDevice->SetRenderState(D3DRS_SRCBLEND, gl_d3d9::ConvertBlendFunc(blendState.sourceBlendRGB)); + mDevice->SetRenderState(D3DRS_DESTBLEND, gl_d3d9::ConvertBlendFunc(blendState.destBlendRGB)); + mDevice->SetRenderState(D3DRS_BLENDOP, gl_d3d9::ConvertBlendOp(blendState.blendEquationRGB)); + + if (blendState.sourceBlendRGB != blendState.sourceBlendAlpha || + blendState.destBlendRGB != blendState.destBlendAlpha || + blendState.blendEquationRGB != blendState.blendEquationAlpha) + { + mDevice->SetRenderState(D3DRS_SEPARATEALPHABLENDENABLE, TRUE); + + mDevice->SetRenderState(D3DRS_SRCBLENDALPHA, gl_d3d9::ConvertBlendFunc(blendState.sourceBlendAlpha)); + mDevice->SetRenderState(D3DRS_DESTBLENDALPHA, gl_d3d9::ConvertBlendFunc(blendState.destBlendAlpha)); + mDevice->SetRenderState(D3DRS_BLENDOPALPHA, gl_d3d9::ConvertBlendOp(blendState.blendEquationAlpha)); + } + else + { + mDevice->SetRenderState(D3DRS_SEPARATEALPHABLENDENABLE, FALSE); + } + } + else + { + mDevice->SetRenderState(D3DRS_ALPHABLENDENABLE, FALSE); + } + + if (blendState.sampleAlphaToCoverage) + { + FIXME("Sample alpha to coverage is unimplemented."); + } + + // Set the color mask + bool zeroColorMaskAllowed = getAdapterVendor() != VENDOR_ID_AMD; + // Apparently some ATI cards have a bug where a draw with a zero color + // write mask can cause later draws to have incorrect results. Instead, + // set a nonzero color write mask but modify the blend state so that no + // drawing is done. + // http://code.google.com/p/angleproject/issues/detail?id=169 + + DWORD colorMask = gl_d3d9::ConvertColorMask(blendState.colorMaskRed, blendState.colorMaskGreen, + blendState.colorMaskBlue, blendState.colorMaskAlpha); + if (colorMask == 0 && !zeroColorMaskAllowed) + { + // Enable green channel, but set blending so nothing will be drawn. + mDevice->SetRenderState(D3DRS_COLORWRITEENABLE, D3DCOLORWRITEENABLE_GREEN); + mDevice->SetRenderState(D3DRS_ALPHABLENDENABLE, TRUE); + + mDevice->SetRenderState(D3DRS_SRCBLEND, D3DBLEND_ZERO); + mDevice->SetRenderState(D3DRS_DESTBLEND, D3DBLEND_ONE); + mDevice->SetRenderState(D3DRS_BLENDOP, D3DBLENDOP_ADD); + } + else + { + mDevice->SetRenderState(D3DRS_COLORWRITEENABLE, colorMask); + } + + mDevice->SetRenderState(D3DRS_DITHERENABLE, blendState.dither ? TRUE : FALSE); + + mCurBlendState = blendState; + mCurBlendColor = blendColor; + } + + if (sampleMaskChanged) + { + // Set the multisample mask + mDevice->SetRenderState(D3DRS_MULTISAMPLEANTIALIAS, TRUE); + mDevice->SetRenderState(D3DRS_MULTISAMPLEMASK, static_cast<DWORD>(sampleMask)); + + mCurSampleMask = sampleMask; + } + + mForceSetBlendState = false; +} + +void Renderer9::setDepthStencilState(const gl::DepthStencilState &depthStencilState, int stencilRef, + int stencilBackRef, bool frontFaceCCW) +{ + bool depthStencilStateChanged = mForceSetDepthStencilState || + memcmp(&depthStencilState, &mCurDepthStencilState, sizeof(gl::DepthStencilState)) != 0; + bool stencilRefChanged = mForceSetDepthStencilState || stencilRef != mCurStencilRef || + stencilBackRef != mCurStencilBackRef; + bool frontFaceCCWChanged = mForceSetDepthStencilState || frontFaceCCW != mCurFrontFaceCCW; + + if (depthStencilStateChanged) + { + if (depthStencilState.depthTest) + { + mDevice->SetRenderState(D3DRS_ZENABLE, D3DZB_TRUE); + mDevice->SetRenderState(D3DRS_ZFUNC, gl_d3d9::ConvertComparison(depthStencilState.depthFunc)); + } + else + { + mDevice->SetRenderState(D3DRS_ZENABLE, D3DZB_FALSE); + } + + mCurDepthStencilState = depthStencilState; + } + + if (depthStencilStateChanged || stencilRefChanged || frontFaceCCWChanged) + { + if (depthStencilState.stencilTest && mCurStencilSize > 0) + { + mDevice->SetRenderState(D3DRS_STENCILENABLE, TRUE); + mDevice->SetRenderState(D3DRS_TWOSIDEDSTENCILMODE, TRUE); + + // FIXME: Unsupported by D3D9 + const D3DRENDERSTATETYPE D3DRS_CCW_STENCILREF = D3DRS_STENCILREF; + const D3DRENDERSTATETYPE D3DRS_CCW_STENCILMASK = D3DRS_STENCILMASK; + const D3DRENDERSTATETYPE D3DRS_CCW_STENCILWRITEMASK = D3DRS_STENCILWRITEMASK; + if (depthStencilState.stencilWritemask != depthStencilState.stencilBackWritemask || + stencilRef != stencilBackRef || + depthStencilState.stencilMask != depthStencilState.stencilBackMask) + { + ERR("Separate front/back stencil writemasks, reference values, or stencil mask values are invalid under WebGL."); + return gl::error(GL_INVALID_OPERATION); + } + + // get the maximum size of the stencil ref + unsigned int maxStencil = (1 << mCurStencilSize) - 1; + + mDevice->SetRenderState(frontFaceCCW ? D3DRS_STENCILWRITEMASK : D3DRS_CCW_STENCILWRITEMASK, + depthStencilState.stencilWritemask); + mDevice->SetRenderState(frontFaceCCW ? D3DRS_STENCILFUNC : D3DRS_CCW_STENCILFUNC, + gl_d3d9::ConvertComparison(depthStencilState.stencilFunc)); + + mDevice->SetRenderState(frontFaceCCW ? D3DRS_STENCILREF : D3DRS_CCW_STENCILREF, + (stencilRef < (int)maxStencil) ? stencilRef : maxStencil); + mDevice->SetRenderState(frontFaceCCW ? D3DRS_STENCILMASK : D3DRS_CCW_STENCILMASK, + depthStencilState.stencilMask); + + mDevice->SetRenderState(frontFaceCCW ? D3DRS_STENCILFAIL : D3DRS_CCW_STENCILFAIL, + gl_d3d9::ConvertStencilOp(depthStencilState.stencilFail)); + mDevice->SetRenderState(frontFaceCCW ? D3DRS_STENCILZFAIL : D3DRS_CCW_STENCILZFAIL, + gl_d3d9::ConvertStencilOp(depthStencilState.stencilPassDepthFail)); + mDevice->SetRenderState(frontFaceCCW ? D3DRS_STENCILPASS : D3DRS_CCW_STENCILPASS, + gl_d3d9::ConvertStencilOp(depthStencilState.stencilPassDepthPass)); + + mDevice->SetRenderState(!frontFaceCCW ? D3DRS_STENCILWRITEMASK : D3DRS_CCW_STENCILWRITEMASK, + depthStencilState.stencilBackWritemask); + mDevice->SetRenderState(!frontFaceCCW ? D3DRS_STENCILFUNC : D3DRS_CCW_STENCILFUNC, + gl_d3d9::ConvertComparison(depthStencilState.stencilBackFunc)); + + mDevice->SetRenderState(!frontFaceCCW ? D3DRS_STENCILREF : D3DRS_CCW_STENCILREF, + (stencilBackRef < (int)maxStencil) ? stencilBackRef : maxStencil); + mDevice->SetRenderState(!frontFaceCCW ? D3DRS_STENCILMASK : D3DRS_CCW_STENCILMASK, + depthStencilState.stencilBackMask); + + mDevice->SetRenderState(!frontFaceCCW ? D3DRS_STENCILFAIL : D3DRS_CCW_STENCILFAIL, + gl_d3d9::ConvertStencilOp(depthStencilState.stencilBackFail)); + mDevice->SetRenderState(!frontFaceCCW ? D3DRS_STENCILZFAIL : D3DRS_CCW_STENCILZFAIL, + gl_d3d9::ConvertStencilOp(depthStencilState.stencilBackPassDepthFail)); + mDevice->SetRenderState(!frontFaceCCW ? D3DRS_STENCILPASS : D3DRS_CCW_STENCILPASS, + gl_d3d9::ConvertStencilOp(depthStencilState.stencilBackPassDepthPass)); + } + else + { + mDevice->SetRenderState(D3DRS_STENCILENABLE, FALSE); + } + + mDevice->SetRenderState(D3DRS_ZWRITEENABLE, depthStencilState.depthMask ? TRUE : FALSE); + + mCurStencilRef = stencilRef; + mCurStencilBackRef = stencilBackRef; + mCurFrontFaceCCW = frontFaceCCW; + } + + mForceSetDepthStencilState = false; +} + +void Renderer9::setScissorRectangle(const gl::Rectangle &scissor, bool enabled) +{ + bool scissorChanged = mForceSetScissor || + memcmp(&scissor, &mCurScissor, sizeof(gl::Rectangle)) != 0 || + enabled != mScissorEnabled; + + if (scissorChanged) + { + if (enabled) + { + RECT rect; + rect.left = gl::clamp(scissor.x, 0, static_cast<int>(mRenderTargetDesc.width)); + rect.top = gl::clamp(scissor.y, 0, static_cast<int>(mRenderTargetDesc.height)); + rect.right = gl::clamp(scissor.x + scissor.width, 0, static_cast<int>(mRenderTargetDesc.width)); + rect.bottom = gl::clamp(scissor.y + scissor.height, 0, static_cast<int>(mRenderTargetDesc.height)); + mDevice->SetScissorRect(&rect); + } + + mDevice->SetRenderState(D3DRS_SCISSORTESTENABLE, enabled ? TRUE : FALSE); + + mScissorEnabled = enabled; + mCurScissor = scissor; + } + + mForceSetScissor = false; +} + +bool Renderer9::setViewport(const gl::Rectangle &viewport, float zNear, float zFar, GLenum drawMode, GLenum frontFace, + bool ignoreViewport) +{ + gl::Rectangle actualViewport = viewport; + float actualZNear = gl::clamp01(zNear); + float actualZFar = gl::clamp01(zFar); + if (ignoreViewport) + { + actualViewport.x = 0; + actualViewport.y = 0; + actualViewport.width = mRenderTargetDesc.width; + actualViewport.height = mRenderTargetDesc.height; + actualZNear = 0.0f; + actualZFar = 1.0f; + } + + D3DVIEWPORT9 dxViewport; + dxViewport.X = gl::clamp(actualViewport.x, 0, static_cast<int>(mRenderTargetDesc.width)); + dxViewport.Y = gl::clamp(actualViewport.y, 0, static_cast<int>(mRenderTargetDesc.height)); + dxViewport.Width = gl::clamp(actualViewport.width, 0, static_cast<int>(mRenderTargetDesc.width) - static_cast<int>(dxViewport.X)); + dxViewport.Height = gl::clamp(actualViewport.height, 0, static_cast<int>(mRenderTargetDesc.height) - static_cast<int>(dxViewport.Y)); + dxViewport.MinZ = actualZNear; + dxViewport.MaxZ = actualZFar; + + if (dxViewport.Width <= 0 || dxViewport.Height <= 0) + { + return false; // Nothing to render + } + + bool viewportChanged = mForceSetViewport || memcmp(&actualViewport, &mCurViewport, sizeof(gl::Rectangle)) != 0 || + actualZNear != mCurNear || actualZFar != mCurFar; + if (viewportChanged) + { + mDevice->SetViewport(&dxViewport); + + mCurViewport = actualViewport; + mCurNear = actualZNear; + mCurFar = actualZFar; + + dx_VertexConstants vc = {0}; + dx_PixelConstants pc = {0}; + + vc.viewAdjust[0] = (float)((actualViewport.width - (int)dxViewport.Width) + 2 * (actualViewport.x - (int)dxViewport.X) - 1) / dxViewport.Width; + vc.viewAdjust[1] = (float)((actualViewport.height - (int)dxViewport.Height) + 2 * (actualViewport.y - (int)dxViewport.Y) - 1) / dxViewport.Height; + vc.viewAdjust[2] = (float)actualViewport.width / dxViewport.Width; + vc.viewAdjust[3] = (float)actualViewport.height / dxViewport.Height; + + pc.viewCoords[0] = actualViewport.width * 0.5f; + pc.viewCoords[1] = actualViewport.height * 0.5f; + pc.viewCoords[2] = actualViewport.x + (actualViewport.width * 0.5f); + pc.viewCoords[3] = actualViewport.y + (actualViewport.height * 0.5f); + + pc.depthFront[0] = (actualZFar - actualZNear) * 0.5f; + pc.depthFront[1] = (actualZNear + actualZFar) * 0.5f; + pc.depthFront[2] = !gl::IsTriangleMode(drawMode) ? 0.0f : (frontFace == GL_CCW ? 1.0f : -1.0f);; + + vc.depthRange[0] = actualZNear; + vc.depthRange[1] = actualZFar; + vc.depthRange[2] = actualZFar - actualZNear; + + pc.depthRange[0] = actualZNear; + pc.depthRange[1] = actualZFar; + pc.depthRange[2] = actualZFar - actualZNear; + + if (memcmp(&vc, &mVertexConstants, sizeof(dx_VertexConstants)) != 0) + { + mVertexConstants = vc; + mDxUniformsDirty = true; + } + + if (memcmp(&pc, &mPixelConstants, sizeof(dx_PixelConstants)) != 0) + { + mPixelConstants = pc; + mDxUniformsDirty = true; + } + } + + mForceSetViewport = false; + return true; +} + +bool Renderer9::applyPrimitiveType(GLenum mode, GLsizei count) +{ + switch (mode) + { + case GL_POINTS: + mPrimitiveType = D3DPT_POINTLIST; + mPrimitiveCount = count; + break; + case GL_LINES: + mPrimitiveType = D3DPT_LINELIST; + mPrimitiveCount = count / 2; + break; + case GL_LINE_LOOP: + mPrimitiveType = D3DPT_LINESTRIP; + mPrimitiveCount = count - 1; // D3D doesn't support line loops, so we draw the last line separately + break; + case GL_LINE_STRIP: + mPrimitiveType = D3DPT_LINESTRIP; + mPrimitiveCount = count - 1; + break; + case GL_TRIANGLES: + mPrimitiveType = D3DPT_TRIANGLELIST; + mPrimitiveCount = count / 3; + break; + case GL_TRIANGLE_STRIP: + mPrimitiveType = D3DPT_TRIANGLESTRIP; + mPrimitiveCount = count - 2; + break; + case GL_TRIANGLE_FAN: + mPrimitiveType = D3DPT_TRIANGLEFAN; + mPrimitiveCount = count - 2; + break; + default: + return gl::error(GL_INVALID_ENUM, false); + } + + return mPrimitiveCount > 0; +} + + +gl::Renderbuffer *Renderer9::getNullColorbuffer(gl::Renderbuffer *depthbuffer) +{ + if (!depthbuffer) + { + ERR("Unexpected null depthbuffer for depth-only FBO."); + return NULL; + } + + GLsizei width = depthbuffer->getWidth(); + GLsizei height = depthbuffer->getHeight(); + + // search cached nullcolorbuffers + for (int i = 0; i < NUM_NULL_COLORBUFFER_CACHE_ENTRIES; i++) + { + if (mNullColorbufferCache[i].buffer != NULL && + mNullColorbufferCache[i].width == width && + mNullColorbufferCache[i].height == height) + { + mNullColorbufferCache[i].lruCount = ++mMaxNullColorbufferLRU; + return mNullColorbufferCache[i].buffer; + } + } + + gl::Renderbuffer *nullbuffer = new gl::Renderbuffer(this, 0, new gl::Colorbuffer(this, width, height, GL_NONE, 0)); + + // add nullbuffer to the cache + NullColorbufferCacheEntry *oldest = &mNullColorbufferCache[0]; + for (int i = 1; i < NUM_NULL_COLORBUFFER_CACHE_ENTRIES; i++) + { + if (mNullColorbufferCache[i].lruCount < oldest->lruCount) + { + oldest = &mNullColorbufferCache[i]; + } + } + + delete oldest->buffer; + oldest->buffer = nullbuffer; + oldest->lruCount = ++mMaxNullColorbufferLRU; + oldest->width = width; + oldest->height = height; + + return nullbuffer; +} + +bool Renderer9::applyRenderTarget(gl::Framebuffer *framebuffer) +{ + // if there is no color attachment we must synthesize a NULL colorattachment + // to keep the D3D runtime happy. This should only be possible if depth texturing. + gl::Renderbuffer *renderbufferObject = NULL; + if (framebuffer->getColorbufferType(0) != GL_NONE) + { + renderbufferObject = framebuffer->getColorbuffer(0); + } + else + { + renderbufferObject = getNullColorbuffer(framebuffer->getDepthbuffer()); + } + if (!renderbufferObject) + { + ERR("unable to locate renderbuffer for FBO."); + return false; + } + + bool renderTargetChanged = false; + unsigned int renderTargetSerial = renderbufferObject->getSerial(); + if (renderTargetSerial != mAppliedRenderTargetSerial) + { + // Apply the render target on the device + IDirect3DSurface9 *renderTargetSurface = NULL; + + RenderTarget *renderTarget = renderbufferObject->getRenderTarget(); + if (renderTarget) + { + renderTargetSurface = RenderTarget9::makeRenderTarget9(renderTarget)->getSurface(); + } + + if (!renderTargetSurface) + { + ERR("render target pointer unexpectedly null."); + return false; // Context must be lost + } + + mDevice->SetRenderTarget(0, renderTargetSurface); + renderTargetSurface->Release(); + + mAppliedRenderTargetSerial = renderTargetSerial; + renderTargetChanged = true; + } + + gl::Renderbuffer *depthStencil = NULL; + unsigned int depthbufferSerial = 0; + unsigned int stencilbufferSerial = 0; + if (framebuffer->getDepthbufferType() != GL_NONE) + { + depthStencil = framebuffer->getDepthbuffer(); + if (!depthStencil) + { + ERR("Depth stencil pointer unexpectedly null."); + return false; + } + + depthbufferSerial = depthStencil->getSerial(); + } + else if (framebuffer->getStencilbufferType() != GL_NONE) + { + depthStencil = framebuffer->getStencilbuffer(); + if (!depthStencil) + { + ERR("Depth stencil pointer unexpectedly null."); + return false; + } + + stencilbufferSerial = depthStencil->getSerial(); + } + + if (depthbufferSerial != mAppliedDepthbufferSerial || + stencilbufferSerial != mAppliedStencilbufferSerial || + !mDepthStencilInitialized) + { + unsigned int depthSize = 0; + unsigned int stencilSize = 0; + + // Apply the depth stencil on the device + if (depthStencil) + { + IDirect3DSurface9 *depthStencilSurface = NULL; + RenderTarget *depthStencilRenderTarget = depthStencil->getDepthStencil(); + + if (depthStencilRenderTarget) + { + depthStencilSurface = RenderTarget9::makeRenderTarget9(depthStencilRenderTarget)->getSurface(); + } + + if (!depthStencilSurface) + { + ERR("depth stencil pointer unexpectedly null."); + return false; // Context must be lost + } + + mDevice->SetDepthStencilSurface(depthStencilSurface); + depthStencilSurface->Release(); + + depthSize = depthStencil->getDepthSize(); + stencilSize = depthStencil->getStencilSize(); + } + else + { + mDevice->SetDepthStencilSurface(NULL); + } + + if (!mDepthStencilInitialized || depthSize != mCurDepthSize) + { + mCurDepthSize = depthSize; + mForceSetRasterState = true; + } + + if (!mDepthStencilInitialized || stencilSize != mCurStencilSize) + { + mCurStencilSize = stencilSize; + mForceSetDepthStencilState = true; + } + + mAppliedDepthbufferSerial = depthbufferSerial; + mAppliedStencilbufferSerial = stencilbufferSerial; + mDepthStencilInitialized = true; + } + + if (renderTargetChanged || !mRenderTargetDescInitialized) + { + mForceSetScissor = true; + mForceSetViewport = true; + + mRenderTargetDesc.width = renderbufferObject->getWidth(); + mRenderTargetDesc.height = renderbufferObject->getHeight(); + mRenderTargetDesc.format = renderbufferObject->getActualFormat(); + mRenderTargetDescInitialized = true; + } + + return true; +} + +GLenum Renderer9::applyVertexBuffer(gl::ProgramBinary *programBinary, gl::VertexAttribute vertexAttributes[], GLint first, GLsizei count, GLsizei instances) +{ + TranslatedAttribute attributes[gl::MAX_VERTEX_ATTRIBS]; + GLenum err = mVertexDataManager->prepareVertexData(vertexAttributes, programBinary, first, count, attributes, instances); + if (err != GL_NO_ERROR) + { + return err; + } + + return mVertexDeclarationCache.applyDeclaration(mDevice, attributes, programBinary, instances, &mRepeatDraw); +} + +// Applies the indices and element array bindings to the Direct3D 9 device +GLenum Renderer9::applyIndexBuffer(const GLvoid *indices, gl::Buffer *elementArrayBuffer, GLsizei count, GLenum mode, GLenum type, TranslatedIndexData *indexInfo) +{ + GLenum err = mIndexDataManager->prepareIndexData(type, count, elementArrayBuffer, indices, indexInfo); + + if (err == GL_NO_ERROR) + { + // Directly binding the storage buffer is not supported for d3d9 + ASSERT(indexInfo->storage == NULL); + + if (indexInfo->serial != mAppliedIBSerial) + { + IndexBuffer9* indexBuffer = IndexBuffer9::makeIndexBuffer9(indexInfo->indexBuffer); + + mDevice->SetIndices(indexBuffer->getBuffer()); + mAppliedIBSerial = indexInfo->serial; + } + } + + return err; +} + +void Renderer9::drawArrays(GLenum mode, GLsizei count, GLsizei instances) +{ + startScene(); + + if (mode == GL_LINE_LOOP) + { + drawLineLoop(count, GL_NONE, NULL, 0, NULL); + } + else if (instances > 0) + { + StaticIndexBufferInterface *countingIB = mIndexDataManager->getCountingIndices(count); + if (countingIB) + { + if (mAppliedIBSerial != countingIB->getSerial()) + { + IndexBuffer9 *indexBuffer = IndexBuffer9::makeIndexBuffer9(countingIB->getIndexBuffer()); + + mDevice->SetIndices(indexBuffer->getBuffer()); + mAppliedIBSerial = countingIB->getSerial(); + } + + for (int i = 0; i < mRepeatDraw; i++) + { + mDevice->DrawIndexedPrimitive(mPrimitiveType, 0, 0, count, 0, mPrimitiveCount); + } + } + else + { + ERR("Could not create a counting index buffer for glDrawArraysInstanced."); + return gl::error(GL_OUT_OF_MEMORY); + } + } + else // Regular case + { + mDevice->DrawPrimitive(mPrimitiveType, 0, mPrimitiveCount); + } +} + +void Renderer9::drawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices, gl::Buffer *elementArrayBuffer, const TranslatedIndexData &indexInfo, GLsizei /*instances*/) +{ + startScene(); + + if (mode == GL_POINTS) + { + drawIndexedPoints(count, type, indices, elementArrayBuffer); + } + else if (mode == GL_LINE_LOOP) + { + drawLineLoop(count, type, indices, indexInfo.minIndex, elementArrayBuffer); + } + else + { + for (int i = 0; i < mRepeatDraw; i++) + { + GLsizei vertexCount = indexInfo.maxIndex - indexInfo.minIndex + 1; + mDevice->DrawIndexedPrimitive(mPrimitiveType, -(INT)indexInfo.minIndex, indexInfo.minIndex, vertexCount, indexInfo.startIndex, mPrimitiveCount); + } + } +} + +void Renderer9::drawLineLoop(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer) +{ + // Get the raw indices for an indexed draw + if (type != GL_NONE && elementArrayBuffer) + { + gl::Buffer *indexBuffer = elementArrayBuffer; + BufferStorage *storage = indexBuffer->getStorage(); + intptr_t offset = reinterpret_cast<intptr_t>(indices); + indices = static_cast<const GLubyte*>(storage->getData()) + offset; + } + + UINT startIndex = 0; + + if (get32BitIndexSupport()) + { + if (!mLineLoopIB) + { + mLineLoopIB = new StreamingIndexBufferInterface(this); + if (!mLineLoopIB->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_INT)) + { + delete mLineLoopIB; + mLineLoopIB = NULL; + + ERR("Could not create a 32-bit looping index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + } + + const int spaceNeeded = (count + 1) * sizeof(unsigned int); + if (!mLineLoopIB->reserveBufferSpace(spaceNeeded, GL_UNSIGNED_INT)) + { + ERR("Could not reserve enough space in looping index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + + void* mappedMemory = NULL; + int offset = mLineLoopIB->mapBuffer(spaceNeeded, &mappedMemory); + if (offset == -1 || mappedMemory == NULL) + { + ERR("Could not map index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + + startIndex = static_cast<UINT>(offset) / 4; + unsigned int *data = reinterpret_cast<unsigned int*>(mappedMemory); + + switch (type) + { + case GL_NONE: // Non-indexed draw + for (int i = 0; i < count; i++) + { + data[i] = i; + } + data[count] = 0; + break; + case GL_UNSIGNED_BYTE: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLubyte*>(indices)[i]; + } + data[count] = static_cast<const GLubyte*>(indices)[0]; + break; + case GL_UNSIGNED_SHORT: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLushort*>(indices)[i]; + } + data[count] = static_cast<const GLushort*>(indices)[0]; + break; + case GL_UNSIGNED_INT: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLuint*>(indices)[i]; + } + data[count] = static_cast<const GLuint*>(indices)[0]; + break; + default: UNREACHABLE(); + } + + if (!mLineLoopIB->unmapBuffer()) + { + ERR("Could not unmap index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + } + else + { + if (!mLineLoopIB) + { + mLineLoopIB = new StreamingIndexBufferInterface(this); + if (!mLineLoopIB->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_SHORT)) + { + delete mLineLoopIB; + mLineLoopIB = NULL; + + ERR("Could not create a 16-bit looping index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + } + + const int spaceNeeded = (count + 1) * sizeof(unsigned short); + if (!mLineLoopIB->reserveBufferSpace(spaceNeeded, GL_UNSIGNED_SHORT)) + { + ERR("Could not reserve enough space in looping index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + + void* mappedMemory = NULL; + int offset = mLineLoopIB->mapBuffer(spaceNeeded, &mappedMemory); + if (offset == -1 || mappedMemory == NULL) + { + ERR("Could not map index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + + startIndex = static_cast<UINT>(offset) / 2; + unsigned short *data = reinterpret_cast<unsigned short*>(mappedMemory); + + switch (type) + { + case GL_NONE: // Non-indexed draw + for (int i = 0; i < count; i++) + { + data[i] = i; + } + data[count] = 0; + break; + case GL_UNSIGNED_BYTE: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLubyte*>(indices)[i]; + } + data[count] = static_cast<const GLubyte*>(indices)[0]; + break; + case GL_UNSIGNED_SHORT: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLushort*>(indices)[i]; + } + data[count] = static_cast<const GLushort*>(indices)[0]; + break; + case GL_UNSIGNED_INT: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLuint*>(indices)[i]; + } + data[count] = static_cast<const GLuint*>(indices)[0]; + break; + default: UNREACHABLE(); + } + + if (!mLineLoopIB->unmapBuffer()) + { + ERR("Could not unmap index buffer for GL_LINE_LOOP."); + return gl::error(GL_OUT_OF_MEMORY); + } + } + + if (mAppliedIBSerial != mLineLoopIB->getSerial()) + { + IndexBuffer9 *indexBuffer = IndexBuffer9::makeIndexBuffer9(mLineLoopIB->getIndexBuffer()); + + mDevice->SetIndices(indexBuffer->getBuffer()); + mAppliedIBSerial = mLineLoopIB->getSerial(); + } + + mDevice->DrawIndexedPrimitive(D3DPT_LINESTRIP, -minIndex, minIndex, count, startIndex, count); +} + +template <typename T> +static void drawPoints(IDirect3DDevice9* device, GLsizei count, const GLvoid *indices) +{ + for (int i = 0; i < count; i++) + { + unsigned int indexValue = static_cast<unsigned int>(static_cast<const T*>(indices)[i]); + device->DrawPrimitive(D3DPT_POINTLIST, indexValue, 1); + } +} + +void Renderer9::drawIndexedPoints(GLsizei count, GLenum type, const GLvoid *indices, gl::Buffer *elementArrayBuffer) +{ + // Drawing index point lists is unsupported in d3d9, fall back to a regular DrawPrimitive call + // for each individual point. This call is not expected to happen often. + + if (elementArrayBuffer) + { + BufferStorage *storage = elementArrayBuffer->getStorage(); + intptr_t offset = reinterpret_cast<intptr_t>(indices); + indices = static_cast<const GLubyte*>(storage->getData()) + offset; + } + + switch (type) + { + case GL_UNSIGNED_BYTE: drawPoints<GLubyte>(mDevice, count, indices); break; + case GL_UNSIGNED_SHORT: drawPoints<GLushort>(mDevice, count, indices); break; + case GL_UNSIGNED_INT: drawPoints<GLuint>(mDevice, count, indices); break; + default: UNREACHABLE(); + } +} + +void Renderer9::applyShaders(gl::ProgramBinary *programBinary) +{ + unsigned int programBinarySerial = programBinary->getSerial(); + if (programBinarySerial != mAppliedProgramBinarySerial) + { + ShaderExecutable *vertexExe = programBinary->getVertexExecutable(); + ShaderExecutable *pixelExe = programBinary->getPixelExecutable(); + + IDirect3DVertexShader9 *vertexShader = NULL; + if (vertexExe) vertexShader = ShaderExecutable9::makeShaderExecutable9(vertexExe)->getVertexShader(); + + IDirect3DPixelShader9 *pixelShader = NULL; + if (pixelExe) pixelShader = ShaderExecutable9::makeShaderExecutable9(pixelExe)->getPixelShader(); + + mDevice->SetPixelShader(pixelShader); + mDevice->SetVertexShader(vertexShader); + programBinary->dirtyAllUniforms(); + mDxUniformsDirty = true; + + mAppliedProgramBinarySerial = programBinarySerial; + } +} + +void Renderer9::applyUniforms(gl::ProgramBinary *programBinary, gl::UniformArray *uniformArray) +{ + for (std::vector<gl::Uniform*>::const_iterator ub = uniformArray->begin(), ue = uniformArray->end(); ub != ue; ++ub) + { + gl::Uniform *targetUniform = *ub; + + if (targetUniform->dirty) + { + GLfloat *f = (GLfloat*)targetUniform->data; + GLint *i = (GLint*)targetUniform->data; + + switch (targetUniform->type) + { + case GL_SAMPLER_2D: + case GL_SAMPLER_CUBE: + break; + case GL_BOOL: + case GL_BOOL_VEC2: + case GL_BOOL_VEC3: + case GL_BOOL_VEC4: + applyUniformnbv(targetUniform, i); + break; + case GL_FLOAT: + case GL_FLOAT_VEC2: + case GL_FLOAT_VEC3: + case GL_FLOAT_VEC4: + case GL_FLOAT_MAT2: + case GL_FLOAT_MAT3: + case GL_FLOAT_MAT4: + applyUniformnfv(targetUniform, f); + break; + case GL_INT: + case GL_INT_VEC2: + case GL_INT_VEC3: + case GL_INT_VEC4: + applyUniformniv(targetUniform, i); + break; + default: + UNREACHABLE(); + } + + targetUniform->dirty = false; + } + } + + // Driver uniforms + if (mDxUniformsDirty) + { + mDevice->SetVertexShaderConstantF(0, (float*)&mVertexConstants, sizeof(dx_VertexConstants) / sizeof(float[4])); + mDevice->SetPixelShaderConstantF(0, (float*)&mPixelConstants, sizeof(dx_PixelConstants) / sizeof(float[4])); + mDxUniformsDirty = false; + } +} + +void Renderer9::applyUniformnfv(gl::Uniform *targetUniform, const GLfloat *v) +{ + if (targetUniform->psRegisterIndex >= 0) + { + mDevice->SetPixelShaderConstantF(targetUniform->psRegisterIndex, v, targetUniform->registerCount); + } + + if (targetUniform->vsRegisterIndex >= 0) + { + mDevice->SetVertexShaderConstantF(targetUniform->vsRegisterIndex, v, targetUniform->registerCount); + } +} + +void Renderer9::applyUniformniv(gl::Uniform *targetUniform, const GLint *v) +{ + ASSERT(targetUniform->registerCount <= MAX_VERTEX_CONSTANT_VECTORS_D3D9); + GLfloat vector[MAX_VERTEX_CONSTANT_VECTORS_D3D9][4]; + + for (unsigned int i = 0; i < targetUniform->registerCount; i++) + { + vector[i][0] = (GLfloat)v[4 * i + 0]; + vector[i][1] = (GLfloat)v[4 * i + 1]; + vector[i][2] = (GLfloat)v[4 * i + 2]; + vector[i][3] = (GLfloat)v[4 * i + 3]; + } + + applyUniformnfv(targetUniform, (GLfloat*)vector); +} + +void Renderer9::applyUniformnbv(gl::Uniform *targetUniform, const GLint *v) +{ + ASSERT(targetUniform->registerCount <= MAX_VERTEX_CONSTANT_VECTORS_D3D9); + GLfloat vector[MAX_VERTEX_CONSTANT_VECTORS_D3D9][4]; + + for (unsigned int i = 0; i < targetUniform->registerCount; i++) + { + vector[i][0] = (v[4 * i + 0] == GL_FALSE) ? 0.0f : 1.0f; + vector[i][1] = (v[4 * i + 1] == GL_FALSE) ? 0.0f : 1.0f; + vector[i][2] = (v[4 * i + 2] == GL_FALSE) ? 0.0f : 1.0f; + vector[i][3] = (v[4 * i + 3] == GL_FALSE) ? 0.0f : 1.0f; + } + + applyUniformnfv(targetUniform, (GLfloat*)vector); +} + +void Renderer9::clear(const gl::ClearParameters &clearParams, gl::Framebuffer *frameBuffer) +{ + D3DCOLOR color = D3DCOLOR_ARGB(gl::unorm<8>(clearParams.colorClearValue.alpha), + gl::unorm<8>(clearParams.colorClearValue.red), + gl::unorm<8>(clearParams.colorClearValue.green), + gl::unorm<8>(clearParams.colorClearValue.blue)); + float depth = gl::clamp01(clearParams.depthClearValue); + int stencil = clearParams.stencilClearValue & 0x000000FF; + + unsigned int stencilUnmasked = 0x0; + if ((clearParams.mask & GL_STENCIL_BUFFER_BIT) && frameBuffer->hasStencil()) + { + unsigned int stencilSize = gl::GetStencilSize(frameBuffer->getStencilbuffer()->getActualFormat()); + stencilUnmasked = (0x1 << stencilSize) - 1; + } + + bool alphaUnmasked = (gl::GetAlphaSize(mRenderTargetDesc.format) == 0) || clearParams.colorMaskAlpha; + + const bool needMaskedStencilClear = (clearParams.mask & GL_STENCIL_BUFFER_BIT) && + (clearParams.stencilWriteMask & stencilUnmasked) != stencilUnmasked; + const bool needMaskedColorClear = (clearParams.mask & GL_COLOR_BUFFER_BIT) && + !(clearParams.colorMaskRed && clearParams.colorMaskGreen && + clearParams.colorMaskBlue && alphaUnmasked); + + if (needMaskedColorClear || needMaskedStencilClear) + { + // State which is altered in all paths from this point to the clear call is saved. + // State which is altered in only some paths will be flagged dirty in the case that + // that path is taken. + HRESULT hr; + if (mMaskedClearSavedState == NULL) + { + hr = mDevice->BeginStateBlock(); + ASSERT(SUCCEEDED(hr) || hr == D3DERR_OUTOFVIDEOMEMORY || hr == E_OUTOFMEMORY); + + mDevice->SetRenderState(D3DRS_ZWRITEENABLE, FALSE); + mDevice->SetRenderState(D3DRS_ZFUNC, D3DCMP_ALWAYS); + mDevice->SetRenderState(D3DRS_ZENABLE, FALSE); + mDevice->SetRenderState(D3DRS_CULLMODE, D3DCULL_NONE); + mDevice->SetRenderState(D3DRS_FILLMODE, D3DFILL_SOLID); + mDevice->SetRenderState(D3DRS_ALPHATESTENABLE, FALSE); + mDevice->SetRenderState(D3DRS_ALPHABLENDENABLE, FALSE); + mDevice->SetRenderState(D3DRS_CLIPPLANEENABLE, 0); + mDevice->SetRenderState(D3DRS_COLORWRITEENABLE, 0); + mDevice->SetRenderState(D3DRS_STENCILENABLE, FALSE); + mDevice->SetPixelShader(NULL); + mDevice->SetVertexShader(NULL); + mDevice->SetFVF(D3DFVF_XYZRHW | D3DFVF_DIFFUSE); + mDevice->SetStreamSource(0, NULL, 0, 0); + mDevice->SetRenderState(D3DRS_SEPARATEALPHABLENDENABLE, TRUE); + mDevice->SetTextureStageState(0, D3DTSS_COLOROP, D3DTOP_SELECTARG1); + mDevice->SetTextureStageState(0, D3DTSS_COLORARG1, D3DTA_TFACTOR); + mDevice->SetTextureStageState(0, D3DTSS_ALPHAOP, D3DTOP_SELECTARG1); + mDevice->SetTextureStageState(0, D3DTSS_ALPHAARG1, D3DTA_TFACTOR); + mDevice->SetRenderState(D3DRS_TEXTUREFACTOR, color); + mDevice->SetRenderState(D3DRS_MULTISAMPLEMASK, 0xFFFFFFFF); + + for(int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + mDevice->SetStreamSourceFreq(i, 1); + } + + hr = mDevice->EndStateBlock(&mMaskedClearSavedState); + ASSERT(SUCCEEDED(hr) || hr == D3DERR_OUTOFVIDEOMEMORY || hr == E_OUTOFMEMORY); + } + + ASSERT(mMaskedClearSavedState != NULL); + + if (mMaskedClearSavedState != NULL) + { + hr = mMaskedClearSavedState->Capture(); + ASSERT(SUCCEEDED(hr)); + } + + mDevice->SetRenderState(D3DRS_ZWRITEENABLE, FALSE); + mDevice->SetRenderState(D3DRS_ZFUNC, D3DCMP_ALWAYS); + mDevice->SetRenderState(D3DRS_ZENABLE, FALSE); + mDevice->SetRenderState(D3DRS_CULLMODE, D3DCULL_NONE); + mDevice->SetRenderState(D3DRS_FILLMODE, D3DFILL_SOLID); + mDevice->SetRenderState(D3DRS_ALPHATESTENABLE, FALSE); + mDevice->SetRenderState(D3DRS_ALPHABLENDENABLE, FALSE); + mDevice->SetRenderState(D3DRS_CLIPPLANEENABLE, 0); + + if (clearParams.mask & GL_COLOR_BUFFER_BIT) + { + mDevice->SetRenderState(D3DRS_COLORWRITEENABLE, + gl_d3d9::ConvertColorMask(clearParams.colorMaskRed, + clearParams.colorMaskGreen, + clearParams.colorMaskBlue, + clearParams.colorMaskAlpha)); + } + else + { + mDevice->SetRenderState(D3DRS_COLORWRITEENABLE, 0); + } + + if (stencilUnmasked != 0x0 && (clearParams.mask & GL_STENCIL_BUFFER_BIT)) + { + mDevice->SetRenderState(D3DRS_STENCILENABLE, TRUE); + mDevice->SetRenderState(D3DRS_TWOSIDEDSTENCILMODE, FALSE); + mDevice->SetRenderState(D3DRS_STENCILFUNC, D3DCMP_ALWAYS); + mDevice->SetRenderState(D3DRS_STENCILREF, stencil); + mDevice->SetRenderState(D3DRS_STENCILWRITEMASK, clearParams.stencilWriteMask); + mDevice->SetRenderState(D3DRS_STENCILFAIL, D3DSTENCILOP_REPLACE); + mDevice->SetRenderState(D3DRS_STENCILZFAIL, D3DSTENCILOP_REPLACE); + mDevice->SetRenderState(D3DRS_STENCILPASS, D3DSTENCILOP_REPLACE); + } + else + { + mDevice->SetRenderState(D3DRS_STENCILENABLE, FALSE); + } + + mDevice->SetPixelShader(NULL); + mDevice->SetVertexShader(NULL); + mDevice->SetFVF(D3DFVF_XYZRHW); + mDevice->SetRenderState(D3DRS_SEPARATEALPHABLENDENABLE, TRUE); + mDevice->SetTextureStageState(0, D3DTSS_COLOROP, D3DTOP_SELECTARG1); + mDevice->SetTextureStageState(0, D3DTSS_COLORARG1, D3DTA_TFACTOR); + mDevice->SetTextureStageState(0, D3DTSS_ALPHAOP, D3DTOP_SELECTARG1); + mDevice->SetTextureStageState(0, D3DTSS_ALPHAARG1, D3DTA_TFACTOR); + mDevice->SetRenderState(D3DRS_TEXTUREFACTOR, color); + mDevice->SetRenderState(D3DRS_MULTISAMPLEMASK, 0xFFFFFFFF); + + for(int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + mDevice->SetStreamSourceFreq(i, 1); + } + + float quad[4][4]; // A quadrilateral covering the target, aligned to match the edges + quad[0][0] = -0.5f; + quad[0][1] = mRenderTargetDesc.height - 0.5f; + quad[0][2] = 0.0f; + quad[0][3] = 1.0f; + + quad[1][0] = mRenderTargetDesc.width - 0.5f; + quad[1][1] = mRenderTargetDesc.height - 0.5f; + quad[1][2] = 0.0f; + quad[1][3] = 1.0f; + + quad[2][0] = -0.5f; + quad[2][1] = -0.5f; + quad[2][2] = 0.0f; + quad[2][3] = 1.0f; + + quad[3][0] = mRenderTargetDesc.width - 0.5f; + quad[3][1] = -0.5f; + quad[3][2] = 0.0f; + quad[3][3] = 1.0f; + + startScene(); + mDevice->DrawPrimitiveUP(D3DPT_TRIANGLESTRIP, 2, quad, sizeof(float[4])); + + if (clearParams.mask & GL_DEPTH_BUFFER_BIT) + { + mDevice->SetRenderState(D3DRS_ZENABLE, TRUE); + mDevice->SetRenderState(D3DRS_ZWRITEENABLE, TRUE); + mDevice->Clear(0, NULL, D3DCLEAR_ZBUFFER, color, depth, stencil); + } + + if (mMaskedClearSavedState != NULL) + { + mMaskedClearSavedState->Apply(); + } + } + else if (clearParams.mask) + { + DWORD dxClearFlags = 0; + if (clearParams.mask & GL_COLOR_BUFFER_BIT) + { + dxClearFlags |= D3DCLEAR_TARGET; + } + if (clearParams.mask & GL_DEPTH_BUFFER_BIT) + { + dxClearFlags |= D3DCLEAR_ZBUFFER; + } + if (clearParams.mask & GL_STENCIL_BUFFER_BIT) + { + dxClearFlags |= D3DCLEAR_STENCIL; + } + + mDevice->Clear(0, NULL, dxClearFlags, color, depth, stencil); + } +} + +void Renderer9::markAllStateDirty() +{ + mAppliedRenderTargetSerial = 0; + mAppliedDepthbufferSerial = 0; + mAppliedStencilbufferSerial = 0; + mDepthStencilInitialized = false; + mRenderTargetDescInitialized = false; + + mForceSetDepthStencilState = true; + mForceSetRasterState = true; + mForceSetScissor = true; + mForceSetViewport = true; + mForceSetBlendState = true; + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS; i++) + { + mForceSetVertexSamplerStates[i] = true; + mCurVertexTextureSerials[i] = 0; + } + for (unsigned int i = 0; i < gl::MAX_TEXTURE_IMAGE_UNITS; i++) + { + mForceSetPixelSamplerStates[i] = true; + mCurPixelTextureSerials[i] = 0; + } + + mAppliedIBSerial = 0; + mAppliedProgramBinarySerial = 0; + mDxUniformsDirty = true; + + mVertexDeclarationCache.markStateDirty(); +} + +void Renderer9::releaseDeviceResources() +{ + while (!mEventQueryPool.empty()) + { + mEventQueryPool.back()->Release(); + mEventQueryPool.pop_back(); + } + + if (mMaskedClearSavedState) + { + mMaskedClearSavedState->Release(); + mMaskedClearSavedState = NULL; + } + + mVertexShaderCache.clear(); + mPixelShaderCache.clear(); + + delete mBlit; + mBlit = NULL; + + delete mVertexDataManager; + mVertexDataManager = NULL; + + delete mIndexDataManager; + mIndexDataManager = NULL; + + delete mLineLoopIB; + mLineLoopIB = NULL; + + for (int i = 0; i < NUM_NULL_COLORBUFFER_CACHE_ENTRIES; i++) + { + delete mNullColorbufferCache[i].buffer; + mNullColorbufferCache[i].buffer = NULL; + } + +} + + +void Renderer9::notifyDeviceLost() +{ + mDeviceLost = true; + mDisplay->notifyDeviceLost(); +} + +bool Renderer9::isDeviceLost() +{ + return mDeviceLost; +} + +// set notify to true to broadcast a message to all contexts of the device loss +bool Renderer9::testDeviceLost(bool notify) +{ + HRESULT status = S_OK; + + if (mDeviceEx) + { + status = mDeviceEx->CheckDeviceState(NULL); + + if (status == S_PRESENT_MODE_CHANGED) + { + // Reset the device so that D3D stops reporting S_PRESENT_MODE_CHANGED. Otherwise it will report + // it continuously, potentially masking a lost device. D3D resources are not lost on a mode change with WDDM. + D3DPRESENT_PARAMETERS presentParameters = getDefaultPresentParameters(); + mDeviceEx->Reset(&presentParameters); + + // Existing swap chains sometimes crash on the next present after a reset. + mDisplay->recreateSwapChains(); + + // Reset will not always cause the device loss to be reported so issue a dummy present. + mDeviceEx->Present(NULL, NULL, NULL, NULL); + + // Retest the device status to see if the mode change really indicated a lost device. + status = mDeviceEx->CheckDeviceState(NULL); + } + } + else if (mDevice) + { + status = mDevice->TestCooperativeLevel(); + } + else + { + // No device yet, so no reset required + } + + bool isLost = FAILED(status) || d3d9::isDeviceLostError(status); + + if (isLost) + { + // ensure we note the device loss -- + // we'll probably get this done again by notifyDeviceLost + // but best to remember it! + // Note that we don't want to clear the device loss status here + // -- this needs to be done by resetDevice + mDeviceLost = true; + if (notify) + { + notifyDeviceLost(); + } + } + + return isLost; +} + +bool Renderer9::testDeviceResettable() +{ + HRESULT status = D3D_OK; + + if (mDeviceEx) + { + status = mDeviceEx->CheckDeviceState(NULL); + } + else if (mDevice) + { + status = mDevice->TestCooperativeLevel(); + } + + // On D3D9Ex, DEVICELOST represents a hung device that needs to be restarted + // DEVICEREMOVED indicates the device has been stopped and must be recreated + switch (status) + { + case D3DERR_DEVICENOTRESET: + case D3DERR_DEVICEHUNG: + return true; + case D3DERR_DEVICELOST: + return (mDeviceEx != NULL); + case D3DERR_DEVICEREMOVED: + UNIMPLEMENTED(); + return false; + default: + return false; + } +} + +bool Renderer9::resetDevice() +{ + releaseDeviceResources(); + + D3DPRESENT_PARAMETERS presentParameters = getDefaultPresentParameters(); + + HRESULT result = D3D_OK; + bool lost = testDeviceLost(false); + int attempts = 3; + + while (lost && attempts > 0) + { + if (mDeviceEx) + { + Sleep(500); // Give the graphics driver some CPU time + result = mDeviceEx->ResetEx(&presentParameters, NULL); + } + else + { + result = mDevice->TestCooperativeLevel(); + while (result == D3DERR_DEVICELOST) + { + Sleep(100); // Give the graphics driver some CPU time + result = mDevice->TestCooperativeLevel(); + } + + if (result == D3DERR_DEVICENOTRESET) + { + result = mDevice->Reset(&presentParameters); + } + } + + lost = testDeviceLost(false); + attempts --; + } + + if (FAILED(result)) + { + ERR("Reset/ResetEx failed multiple times: 0x%08X", result); + return false; + } + + // reset device defaults + initializeDevice(); + mDeviceLost = false; + + return true; +} + +DWORD Renderer9::getAdapterVendor() const +{ + return mAdapterIdentifier.VendorId; +} + +std::string Renderer9::getRendererDescription() const +{ + std::ostringstream rendererString; + + rendererString << mAdapterIdentifier.Description; + if (getShareHandleSupport()) + { + rendererString << " Direct3D9Ex"; + } + else + { + rendererString << " Direct3D9"; + } + + rendererString << " vs_" << D3DSHADER_VERSION_MAJOR(mDeviceCaps.VertexShaderVersion) << "_" << D3DSHADER_VERSION_MINOR(mDeviceCaps.VertexShaderVersion); + rendererString << " ps_" << D3DSHADER_VERSION_MAJOR(mDeviceCaps.PixelShaderVersion) << "_" << D3DSHADER_VERSION_MINOR(mDeviceCaps.PixelShaderVersion); + + return rendererString.str(); +} + +GUID Renderer9::getAdapterIdentifier() const +{ + return mAdapterIdentifier.DeviceIdentifier; +} + +void Renderer9::getMultiSampleSupport(D3DFORMAT format, bool *multiSampleArray) +{ + for (int multiSampleIndex = 0; multiSampleIndex <= D3DMULTISAMPLE_16_SAMPLES; multiSampleIndex++) + { + HRESULT result = mD3d9->CheckDeviceMultiSampleType(mAdapter, mDeviceType, format, + TRUE, (D3DMULTISAMPLE_TYPE)multiSampleIndex, NULL); + + multiSampleArray[multiSampleIndex] = SUCCEEDED(result); + } +} + +bool Renderer9::getBGRATextureSupport() const +{ + // DirectX 9 always supports BGRA + return true; +} + +bool Renderer9::getDXT1TextureSupport() +{ + return mDXT1TextureSupport; +} + +bool Renderer9::getDXT3TextureSupport() +{ + return mDXT3TextureSupport; +} + +bool Renderer9::getDXT5TextureSupport() +{ + return mDXT5TextureSupport; +} + +bool Renderer9::getDepthTextureSupport() const +{ + return mDepthTextureSupport; +} + +bool Renderer9::getFloat32TextureSupport(bool *filtering, bool *renderable) +{ + *filtering = mFloat32FilterSupport; + *renderable = mFloat32RenderSupport; + return mFloat32TextureSupport; +} + +bool Renderer9::getFloat16TextureSupport(bool *filtering, bool *renderable) +{ + *filtering = mFloat16FilterSupport; + *renderable = mFloat16RenderSupport; + return mFloat16TextureSupport; +} + +bool Renderer9::getLuminanceTextureSupport() +{ + return mLuminanceTextureSupport; +} + +bool Renderer9::getLuminanceAlphaTextureSupport() +{ + return mLuminanceAlphaTextureSupport; +} + +bool Renderer9::getTextureFilterAnisotropySupport() const +{ + return mSupportsTextureFilterAnisotropy; +} + +float Renderer9::getTextureMaxAnisotropy() const +{ + if (mSupportsTextureFilterAnisotropy) + { + return static_cast<float>(mDeviceCaps.MaxAnisotropy); + } + return 1.0f; +} + +bool Renderer9::getEventQuerySupport() +{ + return mEventQuerySupport; +} + +unsigned int Renderer9::getMaxVertexTextureImageUnits() const +{ + META_ASSERT(MAX_TEXTURE_IMAGE_UNITS_VTF_SM3 <= gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS); + return mVertexTextureSupport ? MAX_TEXTURE_IMAGE_UNITS_VTF_SM3 : 0; +} + +unsigned int Renderer9::getMaxCombinedTextureImageUnits() const +{ + return gl::MAX_TEXTURE_IMAGE_UNITS + getMaxVertexTextureImageUnits(); +} + +unsigned int Renderer9::getReservedVertexUniformVectors() const +{ + return 2; // dx_ViewAdjust and dx_DepthRange. +} + +unsigned int Renderer9::getReservedFragmentUniformVectors() const +{ + return 3; // dx_ViewCoords, dx_DepthFront and dx_DepthRange. +} + +unsigned int Renderer9::getMaxVertexUniformVectors() const +{ + return MAX_VERTEX_CONSTANT_VECTORS_D3D9 - getReservedVertexUniformVectors(); +} + +unsigned int Renderer9::getMaxFragmentUniformVectors() const +{ + const int maxPixelConstantVectors = (getMajorShaderModel() >= 3) ? MAX_PIXEL_CONSTANT_VECTORS_SM3 : MAX_PIXEL_CONSTANT_VECTORS_SM2; + + return maxPixelConstantVectors - getReservedFragmentUniformVectors(); +} + +unsigned int Renderer9::getMaxVaryingVectors() const +{ + return (getMajorShaderModel() >= 3) ? MAX_VARYING_VECTORS_SM3 : MAX_VARYING_VECTORS_SM2; +} + +bool Renderer9::getNonPower2TextureSupport() const +{ + return mSupportsNonPower2Textures; +} + +bool Renderer9::getOcclusionQuerySupport() const +{ + return mOcclusionQuerySupport; +} + +bool Renderer9::getInstancingSupport() const +{ + return mDeviceCaps.PixelShaderVersion >= D3DPS_VERSION(3, 0); +} + +bool Renderer9::getShareHandleSupport() const +{ + // PIX doesn't seem to support using share handles, so disable them. + return (mD3d9Ex != NULL) && !gl::perfActive(); +} + +bool Renderer9::getDerivativeInstructionSupport() const +{ + return (mDeviceCaps.PS20Caps.Caps & D3DPS20CAPS_GRADIENTINSTRUCTIONS) != 0; +} + +bool Renderer9::getPostSubBufferSupport() const +{ + return true; +} + +int Renderer9::getMajorShaderModel() const +{ + return D3DSHADER_VERSION_MAJOR(mDeviceCaps.PixelShaderVersion); +} + +float Renderer9::getMaxPointSize() const +{ + // Point size clamped at 1.0f for SM2 + return getMajorShaderModel() == 3 ? mDeviceCaps.MaxPointSize : 1.0f; +} + +int Renderer9::getMaxViewportDimension() const +{ + int maxTextureDimension = std::min(std::min(getMaxTextureWidth(), getMaxTextureHeight()), + (int)gl::IMPLEMENTATION_MAX_TEXTURE_SIZE); + return maxTextureDimension; +} + +int Renderer9::getMaxTextureWidth() const +{ + return (int)mDeviceCaps.MaxTextureWidth; +} + +int Renderer9::getMaxTextureHeight() const +{ + return (int)mDeviceCaps.MaxTextureHeight; +} + +bool Renderer9::get32BitIndexSupport() const +{ + return mDeviceCaps.MaxVertexIndex >= (1 << 16); +} + +DWORD Renderer9::getCapsDeclTypes() const +{ + return mDeviceCaps.DeclTypes; +} + +int Renderer9::getMinSwapInterval() const +{ + return mMinSwapInterval; +} + +int Renderer9::getMaxSwapInterval() const +{ + return mMaxSwapInterval; +} + +int Renderer9::getMaxSupportedSamples() const +{ + return mMaxSupportedSamples; +} + +int Renderer9::getNearestSupportedSamples(D3DFORMAT format, int requested) const +{ + if (requested == 0) + { + return requested; + } + + std::map<D3DFORMAT, bool *>::const_iterator itr = mMultiSampleSupport.find(format); + if (itr == mMultiSampleSupport.end()) + { + if (format == D3DFMT_UNKNOWN) + return 0; + return -1; + } + + for (int i = requested; i <= D3DMULTISAMPLE_16_SAMPLES; ++i) + { + if (itr->second[i] && i != D3DMULTISAMPLE_NONMASKABLE) + { + return i; + } + } + + return -1; +} + +unsigned int Renderer9::getMaxRenderTargets() const +{ + // we do not support MRT in d3d9 + return 1; +} + +D3DFORMAT Renderer9::ConvertTextureInternalFormat(GLint internalformat) +{ + switch (internalformat) + { + case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT32_OES: + case GL_DEPTH24_STENCIL8_OES: + return D3DFMT_INTZ; + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return D3DFMT_DXT1; + case GL_COMPRESSED_RGBA_S3TC_DXT3_ANGLE: + return D3DFMT_DXT3; + case GL_COMPRESSED_RGBA_S3TC_DXT5_ANGLE: + return D3DFMT_DXT5; + case GL_RGBA32F_EXT: + case GL_RGB32F_EXT: + case GL_ALPHA32F_EXT: + case GL_LUMINANCE32F_EXT: + case GL_LUMINANCE_ALPHA32F_EXT: + return D3DFMT_A32B32G32R32F; + case GL_RGBA16F_EXT: + case GL_RGB16F_EXT: + case GL_ALPHA16F_EXT: + case GL_LUMINANCE16F_EXT: + case GL_LUMINANCE_ALPHA16F_EXT: + return D3DFMT_A16B16G16R16F; + case GL_LUMINANCE8_EXT: + if (getLuminanceTextureSupport()) + { + return D3DFMT_L8; + } + break; + case GL_LUMINANCE8_ALPHA8_EXT: + if (getLuminanceAlphaTextureSupport()) + { + return D3DFMT_A8L8; + } + break; + case GL_RGB8_OES: + case GL_RGB565: + return D3DFMT_X8R8G8B8; + } + + return D3DFMT_A8R8G8B8; +} + +bool Renderer9::copyToRenderTarget(TextureStorageInterface2D *dest, TextureStorageInterface2D *source) +{ + bool result = false; + + if (source && dest) + { + TextureStorage9_2D *source9 = TextureStorage9_2D::makeTextureStorage9_2D(source->getStorageInstance()); + TextureStorage9_2D *dest9 = TextureStorage9_2D::makeTextureStorage9_2D(dest->getStorageInstance()); + + int levels = source9->levelCount(); + for (int i = 0; i < levels; ++i) + { + IDirect3DSurface9 *srcSurf = source9->getSurfaceLevel(i, false); + IDirect3DSurface9 *dstSurf = dest9->getSurfaceLevel(i, false); + + result = copyToRenderTarget(dstSurf, srcSurf, source9->isManaged()); + + if (srcSurf) srcSurf->Release(); + if (dstSurf) dstSurf->Release(); + + if (!result) + return false; + } + } + + return result; +} + +bool Renderer9::copyToRenderTarget(TextureStorageInterfaceCube *dest, TextureStorageInterfaceCube *source) +{ + bool result = false; + + if (source && dest) + { + TextureStorage9_Cube *source9 = TextureStorage9_Cube::makeTextureStorage9_Cube(source->getStorageInstance()); + TextureStorage9_Cube *dest9 = TextureStorage9_Cube::makeTextureStorage9_Cube(dest->getStorageInstance()); + int levels = source9->levelCount(); + for (int f = 0; f < 6; f++) + { + for (int i = 0; i < levels; i++) + { + IDirect3DSurface9 *srcSurf = source9->getCubeMapSurface(GL_TEXTURE_CUBE_MAP_POSITIVE_X + f, i, false); + IDirect3DSurface9 *dstSurf = dest9->getCubeMapSurface(GL_TEXTURE_CUBE_MAP_POSITIVE_X + f, i, true); + + result = copyToRenderTarget(dstSurf, srcSurf, source9->isManaged()); + + if (srcSurf) srcSurf->Release(); + if (dstSurf) dstSurf->Release(); + + if (!result) + return false; + } + } + } + + return result; +} + +D3DPOOL Renderer9::getBufferPool(DWORD usage) const +{ + if (mD3d9Ex != NULL) + { + return D3DPOOL_DEFAULT; + } + else + { + if (!(usage & D3DUSAGE_DYNAMIC)) + { + return D3DPOOL_MANAGED; + } + } + + return D3DPOOL_DEFAULT; +} + +bool Renderer9::copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterface2D *storage, GLint level) +{ + RECT rect; + rect.left = sourceRect.x; + rect.top = sourceRect.y; + rect.right = sourceRect.x + sourceRect.width; + rect.bottom = sourceRect.y + sourceRect.height; + + return mBlit->copy(framebuffer, rect, destFormat, xoffset, yoffset, storage, level); +} + +bool Renderer9::copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterfaceCube *storage, GLenum target, GLint level) +{ + RECT rect; + rect.left = sourceRect.x; + rect.top = sourceRect.y; + rect.right = sourceRect.x + sourceRect.width; + rect.bottom = sourceRect.y + sourceRect.height; + + return mBlit->copy(framebuffer, rect, destFormat, xoffset, yoffset, storage, target, level); +} + +bool Renderer9::blitRect(gl::Framebuffer *readFramebuffer, const gl::Rectangle &readRect, gl::Framebuffer *drawFramebuffer, const gl::Rectangle &drawRect, + bool blitRenderTarget, bool blitDepthStencil) +{ + endScene(); + + if (blitRenderTarget) + { + gl::Renderbuffer *readBuffer = readFramebuffer->getColorbuffer(0); + gl::Renderbuffer *drawBuffer = drawFramebuffer->getColorbuffer(0); + RenderTarget9 *readRenderTarget = NULL; + RenderTarget9 *drawRenderTarget = NULL; + IDirect3DSurface9* readSurface = NULL; + IDirect3DSurface9* drawSurface = NULL; + + if (readBuffer) + { + readRenderTarget = RenderTarget9::makeRenderTarget9(readBuffer->getRenderTarget()); + } + if (drawBuffer) + { + drawRenderTarget = RenderTarget9::makeRenderTarget9(drawBuffer->getRenderTarget()); + } + + if (readRenderTarget) + { + readSurface = readRenderTarget->getSurface(); + } + if (drawRenderTarget) + { + drawSurface = drawRenderTarget->getSurface(); + } + + if (!readSurface || !drawSurface) + { + ERR("Failed to retrieve the render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + RECT srcRect; + srcRect.left = readRect.x; + srcRect.right = readRect.x + readRect.width; + srcRect.top = readRect.y; + srcRect.bottom = readRect.y + readRect.height; + + RECT dstRect; + dstRect.left = drawRect.x; + dstRect.right = drawRect.x + drawRect.width; + dstRect.top = drawRect.y; + dstRect.bottom = drawRect.y + drawRect.height; + + HRESULT result = mDevice->StretchRect(readSurface, &srcRect, drawSurface, &dstRect, D3DTEXF_NONE); + + readSurface->Release(); + drawSurface->Release(); + + if (FAILED(result)) + { + ERR("BlitFramebufferANGLE failed: StretchRect returned %x.", result); + return false; + } + } + + if (blitDepthStencil) + { + gl::Renderbuffer *readBuffer = readFramebuffer->getDepthOrStencilbuffer(); + gl::Renderbuffer *drawBuffer = drawFramebuffer->getDepthOrStencilbuffer(); + RenderTarget9 *readDepthStencil = NULL; + RenderTarget9 *drawDepthStencil = NULL; + IDirect3DSurface9* readSurface = NULL; + IDirect3DSurface9* drawSurface = NULL; + + if (readBuffer) + { + readDepthStencil = RenderTarget9::makeRenderTarget9(readBuffer->getDepthStencil()); + } + if (drawBuffer) + { + drawDepthStencil = RenderTarget9::makeRenderTarget9(drawBuffer->getDepthStencil()); + } + + if (readDepthStencil) + { + readSurface = readDepthStencil->getSurface(); + } + if (drawDepthStencil) + { + drawSurface = drawDepthStencil->getSurface(); + } + + if (!readSurface || !drawSurface) + { + ERR("Failed to retrieve the render target."); + return gl::error(GL_OUT_OF_MEMORY, false); + } + + HRESULT result = mDevice->StretchRect(readSurface, NULL, drawSurface, NULL, D3DTEXF_NONE); + + readSurface->Release(); + drawSurface->Release(); + + if (FAILED(result)) + { + ERR("BlitFramebufferANGLE failed: StretchRect returned %x.", result); + return false; + } + } + + return true; +} + +void Renderer9::readPixels(gl::Framebuffer *framebuffer, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, + GLsizei outputPitch, bool packReverseRowOrder, GLint packAlignment, void* pixels) +{ + RenderTarget9 *renderTarget = NULL; + IDirect3DSurface9 *surface = NULL; + gl::Renderbuffer *colorbuffer = framebuffer->getColorbuffer(0); + + if (colorbuffer) + { + renderTarget = RenderTarget9::makeRenderTarget9(colorbuffer->getRenderTarget()); + } + + if (renderTarget) + { + surface = renderTarget->getSurface(); + } + + if (!surface) + { + // context must be lost + return; + } + + D3DSURFACE_DESC desc; + surface->GetDesc(&desc); + + if (desc.MultiSampleType != D3DMULTISAMPLE_NONE) + { + UNIMPLEMENTED(); // FIXME: Requires resolve using StretchRect into non-multisampled render target + surface->Release(); + return gl::error(GL_OUT_OF_MEMORY); + } + + HRESULT result; + IDirect3DSurface9 *systemSurface = NULL; + bool directToPixels = !packReverseRowOrder && packAlignment <= 4 && getShareHandleSupport() && + x == 0 && y == 0 && UINT(width) == desc.Width && UINT(height) == desc.Height && + desc.Format == D3DFMT_A8R8G8B8 && format == GL_BGRA_EXT && type == GL_UNSIGNED_BYTE; + if (directToPixels) + { + // Use the pixels ptr as a shared handle to write directly into client's memory + result = mDevice->CreateOffscreenPlainSurface(desc.Width, desc.Height, desc.Format, + D3DPOOL_SYSTEMMEM, &systemSurface, &pixels); + if (FAILED(result)) + { + // Try again without the shared handle + directToPixels = false; + } + } + + if (!directToPixels) + { + result = mDevice->CreateOffscreenPlainSurface(desc.Width, desc.Height, desc.Format, + D3DPOOL_SYSTEMMEM, &systemSurface, NULL); + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + surface->Release(); + return gl::error(GL_OUT_OF_MEMORY); + } + } + + result = mDevice->GetRenderTargetData(surface, systemSurface); + surface->Release(); + surface = NULL; + + if (FAILED(result)) + { + systemSurface->Release(); + + // It turns out that D3D will sometimes produce more error + // codes than those documented. + if (d3d9::isDeviceLostError(result)) + { + notifyDeviceLost(); + return gl::error(GL_OUT_OF_MEMORY); + } + else + { + UNREACHABLE(); + return; + } + + } + + if (directToPixels) + { + systemSurface->Release(); + return; + } + + RECT rect; + rect.left = gl::clamp(x, 0L, static_cast<LONG>(desc.Width)); + rect.top = gl::clamp(y, 0L, static_cast<LONG>(desc.Height)); + rect.right = gl::clamp(x + width, 0L, static_cast<LONG>(desc.Width)); + rect.bottom = gl::clamp(y + height, 0L, static_cast<LONG>(desc.Height)); + + D3DLOCKED_RECT lock; + result = systemSurface->LockRect(&lock, &rect, D3DLOCK_READONLY); + + if (FAILED(result)) + { + UNREACHABLE(); + systemSurface->Release(); + + return; // No sensible error to generate + } + + unsigned char *dest = (unsigned char*)pixels; + unsigned short *dest16 = (unsigned short*)pixels; + + unsigned char *source; + int inputPitch; + if (packReverseRowOrder) + { + source = ((unsigned char*)lock.pBits) + lock.Pitch * (rect.bottom - rect.top - 1); + inputPitch = -lock.Pitch; + } + else + { + source = (unsigned char*)lock.pBits; + inputPitch = lock.Pitch; + } + + unsigned int fastPixelSize = 0; + + if (desc.Format == D3DFMT_A8R8G8B8 && + format == GL_BGRA_EXT && + type == GL_UNSIGNED_BYTE) + { + fastPixelSize = 4; + } + else if ((desc.Format == D3DFMT_A4R4G4B4 && + format == GL_BGRA_EXT && + type == GL_UNSIGNED_SHORT_4_4_4_4_REV_EXT) || + (desc.Format == D3DFMT_A1R5G5B5 && + format == GL_BGRA_EXT && + type == GL_UNSIGNED_SHORT_1_5_5_5_REV_EXT)) + { + fastPixelSize = 2; + } + else if (desc.Format == D3DFMT_A16B16G16R16F && + format == GL_RGBA && + type == GL_HALF_FLOAT_OES) + { + fastPixelSize = 8; + } + else if (desc.Format == D3DFMT_A32B32G32R32F && + format == GL_RGBA && + type == GL_FLOAT) + { + fastPixelSize = 16; + } + + for (int j = 0; j < rect.bottom - rect.top; j++) + { + if (fastPixelSize != 0) + { + // Fast path for formats which require no translation: + // D3DFMT_A8R8G8B8 to BGRA/UNSIGNED_BYTE + // D3DFMT_A4R4G4B4 to BGRA/UNSIGNED_SHORT_4_4_4_4_REV_EXT + // D3DFMT_A1R5G5B5 to BGRA/UNSIGNED_SHORT_1_5_5_5_REV_EXT + // D3DFMT_A16B16G16R16F to RGBA/HALF_FLOAT_OES + // D3DFMT_A32B32G32R32F to RGBA/FLOAT + // + // Note that buffers with no alpha go through the slow path below. + memcpy(dest + j * outputPitch, + source + j * inputPitch, + (rect.right - rect.left) * fastPixelSize); + continue; + } + else if (desc.Format == D3DFMT_A8R8G8B8 && + format == GL_RGBA && + type == GL_UNSIGNED_BYTE) + { + // Fast path for swapping red with blue + for (int i = 0; i < rect.right - rect.left; i++) + { + unsigned int argb = *(unsigned int*)(source + 4 * i + j * inputPitch); + *(unsigned int*)(dest + 4 * i + j * outputPitch) = + (argb & 0xFF00FF00) | // Keep alpha and green + (argb & 0x00FF0000) >> 16 | // Move red to blue + (argb & 0x000000FF) << 16; // Move blue to red + } + continue; + } + + for (int i = 0; i < rect.right - rect.left; i++) + { + float r; + float g; + float b; + float a; + + switch (desc.Format) + { + case D3DFMT_R5G6B5: + { + unsigned short rgb = *(unsigned short*)(source + 2 * i + j * inputPitch); + + a = 1.0f; + b = (rgb & 0x001F) * (1.0f / 0x001F); + g = (rgb & 0x07E0) * (1.0f / 0x07E0); + r = (rgb & 0xF800) * (1.0f / 0xF800); + } + break; + case D3DFMT_A1R5G5B5: + { + unsigned short argb = *(unsigned short*)(source + 2 * i + j * inputPitch); + + a = (argb & 0x8000) ? 1.0f : 0.0f; + b = (argb & 0x001F) * (1.0f / 0x001F); + g = (argb & 0x03E0) * (1.0f / 0x03E0); + r = (argb & 0x7C00) * (1.0f / 0x7C00); + } + break; + case D3DFMT_A8R8G8B8: + { + unsigned int argb = *(unsigned int*)(source + 4 * i + j * inputPitch); + + a = (argb & 0xFF000000) * (1.0f / 0xFF000000); + b = (argb & 0x000000FF) * (1.0f / 0x000000FF); + g = (argb & 0x0000FF00) * (1.0f / 0x0000FF00); + r = (argb & 0x00FF0000) * (1.0f / 0x00FF0000); + } + break; + case D3DFMT_X8R8G8B8: + { + unsigned int xrgb = *(unsigned int*)(source + 4 * i + j * inputPitch); + + a = 1.0f; + b = (xrgb & 0x000000FF) * (1.0f / 0x000000FF); + g = (xrgb & 0x0000FF00) * (1.0f / 0x0000FF00); + r = (xrgb & 0x00FF0000) * (1.0f / 0x00FF0000); + } + break; + case D3DFMT_A2R10G10B10: + { + unsigned int argb = *(unsigned int*)(source + 4 * i + j * inputPitch); + + a = (argb & 0xC0000000) * (1.0f / 0xC0000000); + b = (argb & 0x000003FF) * (1.0f / 0x000003FF); + g = (argb & 0x000FFC00) * (1.0f / 0x000FFC00); + r = (argb & 0x3FF00000) * (1.0f / 0x3FF00000); + } + break; + case D3DFMT_A32B32G32R32F: + { + // float formats in D3D are stored rgba, rather than the other way round + r = *((float*)(source + 16 * i + j * inputPitch) + 0); + g = *((float*)(source + 16 * i + j * inputPitch) + 1); + b = *((float*)(source + 16 * i + j * inputPitch) + 2); + a = *((float*)(source + 16 * i + j * inputPitch) + 3); + } + break; + case D3DFMT_A16B16G16R16F: + { + // float formats in D3D are stored rgba, rather than the other way round + r = gl::float16ToFloat32(*((unsigned short*)(source + 8 * i + j * inputPitch) + 0)); + g = gl::float16ToFloat32(*((unsigned short*)(source + 8 * i + j * inputPitch) + 1)); + b = gl::float16ToFloat32(*((unsigned short*)(source + 8 * i + j * inputPitch) + 2)); + a = gl::float16ToFloat32(*((unsigned short*)(source + 8 * i + j * inputPitch) + 3)); + } + break; + default: + UNIMPLEMENTED(); // FIXME + UNREACHABLE(); + return; + } + + switch (format) + { + case GL_RGBA: + switch (type) + { + case GL_UNSIGNED_BYTE: + dest[4 * i + j * outputPitch + 0] = (unsigned char)(255 * r + 0.5f); + dest[4 * i + j * outputPitch + 1] = (unsigned char)(255 * g + 0.5f); + dest[4 * i + j * outputPitch + 2] = (unsigned char)(255 * b + 0.5f); + dest[4 * i + j * outputPitch + 3] = (unsigned char)(255 * a + 0.5f); + break; + default: UNREACHABLE(); + } + break; + case GL_BGRA_EXT: + switch (type) + { + case GL_UNSIGNED_BYTE: + dest[4 * i + j * outputPitch + 0] = (unsigned char)(255 * b + 0.5f); + dest[4 * i + j * outputPitch + 1] = (unsigned char)(255 * g + 0.5f); + dest[4 * i + j * outputPitch + 2] = (unsigned char)(255 * r + 0.5f); + dest[4 * i + j * outputPitch + 3] = (unsigned char)(255 * a + 0.5f); + break; + case GL_UNSIGNED_SHORT_4_4_4_4_REV_EXT: + // According to the desktop GL spec in the "Transfer of Pixel Rectangles" section + // this type is packed as follows: + // 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 + // -------------------------------------------------------------------------------- + // | 4th | 3rd | 2nd | 1st component | + // -------------------------------------------------------------------------------- + // in the case of BGRA_EXT, B is the first component, G the second, and so forth. + dest16[i + j * outputPitch / sizeof(unsigned short)] = + ((unsigned short)(15 * a + 0.5f) << 12)| + ((unsigned short)(15 * r + 0.5f) << 8) | + ((unsigned short)(15 * g + 0.5f) << 4) | + ((unsigned short)(15 * b + 0.5f) << 0); + break; + case GL_UNSIGNED_SHORT_1_5_5_5_REV_EXT: + // According to the desktop GL spec in the "Transfer of Pixel Rectangles" section + // this type is packed as follows: + // 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 + // -------------------------------------------------------------------------------- + // | 4th | 3rd | 2nd | 1st component | + // -------------------------------------------------------------------------------- + // in the case of BGRA_EXT, B is the first component, G the second, and so forth. + dest16[i + j * outputPitch / sizeof(unsigned short)] = + ((unsigned short)( a + 0.5f) << 15) | + ((unsigned short)(31 * r + 0.5f) << 10) | + ((unsigned short)(31 * g + 0.5f) << 5) | + ((unsigned short)(31 * b + 0.5f) << 0); + break; + default: UNREACHABLE(); + } + break; + case GL_RGB: + switch (type) + { + case GL_UNSIGNED_SHORT_5_6_5: + dest16[i + j * outputPitch / sizeof(unsigned short)] = + ((unsigned short)(31 * b + 0.5f) << 0) | + ((unsigned short)(63 * g + 0.5f) << 5) | + ((unsigned short)(31 * r + 0.5f) << 11); + break; + case GL_UNSIGNED_BYTE: + dest[3 * i + j * outputPitch + 0] = (unsigned char)(255 * r + 0.5f); + dest[3 * i + j * outputPitch + 1] = (unsigned char)(255 * g + 0.5f); + dest[3 * i + j * outputPitch + 2] = (unsigned char)(255 * b + 0.5f); + break; + default: UNREACHABLE(); + } + break; + default: UNREACHABLE(); + } + } + } + + systemSurface->UnlockRect(); + + systemSurface->Release(); +} + +RenderTarget *Renderer9::createRenderTarget(SwapChain *swapChain, bool depth) +{ + SwapChain9 *swapChain9 = SwapChain9::makeSwapChain9(swapChain); + IDirect3DSurface9 *surface = NULL; + if (depth) + { + surface = swapChain9->getDepthStencil(); + } + else + { + surface = swapChain9->getRenderTarget(); + } + + RenderTarget9 *renderTarget = new RenderTarget9(this, surface); + + return renderTarget; +} + +RenderTarget *Renderer9::createRenderTarget(int width, int height, GLenum format, GLsizei samples, bool depth) +{ + RenderTarget9 *renderTarget = new RenderTarget9(this, width, height, format, samples); + return renderTarget; +} + +ShaderExecutable *Renderer9::loadExecutable(const void *function, size_t length, rx::ShaderType type) +{ + ShaderExecutable9 *executable = NULL; + + switch (type) + { + case rx::SHADER_VERTEX: + { + IDirect3DVertexShader9 *vshader = createVertexShader((DWORD*)function, length); + if (vshader) + { + executable = new ShaderExecutable9(function, length, vshader); + } + } + break; + case rx::SHADER_PIXEL: + { + IDirect3DPixelShader9 *pshader = createPixelShader((DWORD*)function, length); + if (pshader) + { + executable = new ShaderExecutable9(function, length, pshader); + } + } + break; + default: + UNREACHABLE(); + break; + } + + return executable; +} + +ShaderExecutable *Renderer9::compileToExecutable(gl::InfoLog &infoLog, const char *shaderHLSL, rx::ShaderType type) +{ + const char *profile = NULL; + + switch (type) + { + case rx::SHADER_VERTEX: + profile = getMajorShaderModel() >= 3 ? "vs_3_0" : "vs_2_0"; + break; + case rx::SHADER_PIXEL: + profile = getMajorShaderModel() >= 3 ? "ps_3_0" : "ps_2_0"; + break; + default: + UNREACHABLE(); + return NULL; + } + + ID3DBlob *binary = (ID3DBlob*)compileToBinary(infoLog, shaderHLSL, profile, ANGLE_COMPILE_OPTIMIZATION_LEVEL, true); + if (!binary) + return NULL; + + ShaderExecutable *executable = loadExecutable(binary->GetBufferPointer(), binary->GetBufferSize(), type); + binary->Release(); + + return executable; +} + +bool Renderer9::boxFilter(IDirect3DSurface9 *source, IDirect3DSurface9 *dest) +{ + return mBlit->boxFilter(source, dest); +} + +D3DPOOL Renderer9::getTexturePool(DWORD usage) const +{ + if (mD3d9Ex != NULL) + { + return D3DPOOL_DEFAULT; + } + else + { + if (!(usage & (D3DUSAGE_DEPTHSTENCIL | D3DUSAGE_RENDERTARGET))) + { + return D3DPOOL_MANAGED; + } + } + + return D3DPOOL_DEFAULT; +} + +bool Renderer9::copyToRenderTarget(IDirect3DSurface9 *dest, IDirect3DSurface9 *source, bool fromManaged) +{ + if (source && dest) + { + HRESULT result = D3DERR_OUTOFVIDEOMEMORY; + + if (fromManaged) + { + D3DSURFACE_DESC desc; + source->GetDesc(&desc); + + IDirect3DSurface9 *surf = 0; + result = mDevice->CreateOffscreenPlainSurface(desc.Width, desc.Height, desc.Format, D3DPOOL_SYSTEMMEM, &surf, NULL); + + if (SUCCEEDED(result)) + { + Image9::copyLockableSurfaces(surf, source); + result = mDevice->UpdateSurface(surf, NULL, dest, NULL); + surf->Release(); + } + } + else + { + endScene(); + result = mDevice->StretchRect(source, NULL, dest, NULL, D3DTEXF_NONE); + } + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + return false; + } + } + + return true; +} + +Image *Renderer9::createImage() +{ + return new Image9(); +} + +void Renderer9::generateMipmap(Image *dest, Image *src) +{ + Image9 *src9 = Image9::makeImage9(src); + Image9 *dst9 = Image9::makeImage9(dest); + Image9::generateMipmap(dst9, src9); +} + +TextureStorage *Renderer9::createTextureStorage2D(SwapChain *swapChain) +{ + SwapChain9 *swapChain9 = SwapChain9::makeSwapChain9(swapChain); + return new TextureStorage9_2D(this, swapChain9); +} + +TextureStorage *Renderer9::createTextureStorage2D(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height) +{ + return new TextureStorage9_2D(this, levels, internalformat, usage, forceRenderable, width, height); +} + +TextureStorage *Renderer9::createTextureStorageCube(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size) +{ + return new TextureStorage9_Cube(this, levels, internalformat, usage, forceRenderable, size); +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/Renderer9.h b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer9.h new file mode 100644 index 0000000000..527a5010ae --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/Renderer9.h @@ -0,0 +1,347 @@ +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Renderer9.h: Defines a back-end specific class for the D3D9 renderer. + +#ifndef LIBGLESV2_RENDERER_RENDERER9_H_ +#define LIBGLESV2_RENDERER_RENDERER9_H_ + +#include "common/angleutils.h" +#include "libGLESv2/mathutil.h" +#include "libGLESv2/renderer/ShaderCache.h" +#include "libGLESv2/renderer/VertexDeclarationCache.h" +#include "libGLESv2/renderer/Renderer.h" +#include "libGLESv2/renderer/RenderTarget.h" + +namespace gl +{ +class Renderbuffer; +} + +namespace rx +{ +class VertexDataManager; +class IndexDataManager; +class StreamingIndexBufferInterface; +struct TranslatedAttribute; + +class Renderer9 : public Renderer +{ + public: + Renderer9(egl::Display *display, HDC hDc, bool softwareDevice); + virtual ~Renderer9(); + + static Renderer9 *makeRenderer9(Renderer *renderer); + + virtual EGLint initialize(); + virtual bool resetDevice(); + + virtual int generateConfigs(ConfigDesc **configDescList); + virtual void deleteConfigs(ConfigDesc *configDescList); + + void startScene(); + void endScene(); + + virtual void sync(bool block); + + virtual SwapChain *createSwapChain(HWND window, HANDLE shareHandle, GLenum backBufferFormat, GLenum depthBufferFormat); + + IDirect3DQuery9* allocateEventQuery(); + void freeEventQuery(IDirect3DQuery9* query); + + // resource creation + IDirect3DVertexShader9 *createVertexShader(const DWORD *function, size_t length); + IDirect3DPixelShader9 *createPixelShader(const DWORD *function, size_t length); + HRESULT createVertexBuffer(UINT Length, DWORD Usage, IDirect3DVertexBuffer9 **ppVertexBuffer); + HRESULT createIndexBuffer(UINT Length, DWORD Usage, D3DFORMAT Format, IDirect3DIndexBuffer9 **ppIndexBuffer); +#if 0 + void *createTexture2D(); + void *createTextureCube(); + void *createQuery(); + void *createIndexBuffer(); + void *createVertexbuffer(); + + // state setup + void applyShaders(); + void applyConstants(); +#endif + virtual void setSamplerState(gl::SamplerType type, int index, const gl::SamplerState &sampler); + virtual void setTexture(gl::SamplerType type, int index, gl::Texture *texture); + + virtual void setRasterizerState(const gl::RasterizerState &rasterState); + virtual void setBlendState(const gl::BlendState &blendState, const gl::Color &blendColor, + unsigned int sampleMask); + virtual void setDepthStencilState(const gl::DepthStencilState &depthStencilState, int stencilRef, + int stencilBackRef, bool frontFaceCCW); + + virtual void setScissorRectangle(const gl::Rectangle &scissor, bool enabled); + virtual bool setViewport(const gl::Rectangle &viewport, float zNear, float zFar, GLenum drawMode, GLenum frontFace, + bool ignoreViewport); + + virtual bool applyRenderTarget(gl::Framebuffer *frameBuffer); + virtual void applyShaders(gl::ProgramBinary *programBinary); + virtual void applyUniforms(gl::ProgramBinary *programBinary, gl::UniformArray *uniformArray); + virtual bool applyPrimitiveType(GLenum primitiveType, GLsizei elementCount); + virtual GLenum applyVertexBuffer(gl::ProgramBinary *programBinary, gl::VertexAttribute vertexAttributes[], GLint first, GLsizei count, GLsizei instances); + virtual GLenum applyIndexBuffer(const GLvoid *indices, gl::Buffer *elementArrayBuffer, GLsizei count, GLenum mode, GLenum type, TranslatedIndexData *indexInfo); + + virtual void drawArrays(GLenum mode, GLsizei count, GLsizei instances); + virtual void drawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices, gl::Buffer *elementArrayBuffer, const TranslatedIndexData &indexInfo, GLsizei instances); + + virtual void clear(const gl::ClearParameters &clearParams, gl::Framebuffer *frameBuffer); + + virtual void markAllStateDirty(); + + // lost device + void notifyDeviceLost(); + virtual bool isDeviceLost(); + virtual bool testDeviceLost(bool notify); + virtual bool testDeviceResettable(); + + // Renderer capabilities + IDirect3DDevice9 *getDevice() { return mDevice; } + virtual DWORD getAdapterVendor() const; + virtual std::string getRendererDescription() const; + virtual GUID getAdapterIdentifier() const; + + virtual bool getBGRATextureSupport() const; + virtual bool getDXT1TextureSupport(); + virtual bool getDXT3TextureSupport(); + virtual bool getDXT5TextureSupport(); + virtual bool getEventQuerySupport(); + virtual bool getFloat32TextureSupport(bool *filtering, bool *renderable); + virtual bool getFloat16TextureSupport(bool *filtering, bool *renderable); + virtual bool getLuminanceTextureSupport(); + virtual bool getLuminanceAlphaTextureSupport(); + virtual unsigned int getMaxVertexTextureImageUnits() const; + virtual unsigned int getMaxCombinedTextureImageUnits() const; + virtual unsigned int getReservedVertexUniformVectors() const; + virtual unsigned int getReservedFragmentUniformVectors() const; + virtual unsigned int getMaxVertexUniformVectors() const; + virtual unsigned int getMaxFragmentUniformVectors() const; + virtual unsigned int getMaxVaryingVectors() const; + virtual bool getNonPower2TextureSupport() const; + virtual bool getDepthTextureSupport() const; + virtual bool getOcclusionQuerySupport() const; + virtual bool getInstancingSupport() const; + virtual bool getTextureFilterAnisotropySupport() const; + virtual float getTextureMaxAnisotropy() const; + virtual bool getShareHandleSupport() const; + virtual bool getDerivativeInstructionSupport() const; + virtual bool getPostSubBufferSupport() const; + + virtual int getMajorShaderModel() const; + virtual float getMaxPointSize() const; + virtual int getMaxViewportDimension() const; + virtual int getMaxTextureWidth() const; + virtual int getMaxTextureHeight() const; + virtual bool get32BitIndexSupport() const; + DWORD getCapsDeclTypes() const; + virtual int getMinSwapInterval() const; + virtual int getMaxSwapInterval() const; + + virtual GLsizei getMaxSupportedSamples() const; + int getNearestSupportedSamples(D3DFORMAT format, int requested) const; + + virtual unsigned int getMaxRenderTargets() const; + + D3DFORMAT ConvertTextureInternalFormat(GLint internalformat); + + // Pixel operations + virtual bool copyToRenderTarget(TextureStorageInterface2D *dest, TextureStorageInterface2D *source); + virtual bool copyToRenderTarget(TextureStorageInterfaceCube *dest, TextureStorageInterfaceCube *source); + + virtual bool copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterface2D *storage, GLint level); + virtual bool copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + GLint xoffset, GLint yoffset, TextureStorageInterfaceCube *storage, GLenum target, GLint level); + + virtual bool blitRect(gl::Framebuffer *readTarget, const gl::Rectangle &readRect, gl::Framebuffer *drawTarget, const gl::Rectangle &drawRect, + bool blitRenderTarget, bool blitDepthStencil); + virtual void readPixels(gl::Framebuffer *framebuffer, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, + GLsizei outputPitch, bool packReverseRowOrder, GLint packAlignment, void* pixels); + + // RenderTarget creation + virtual RenderTarget *createRenderTarget(SwapChain *swapChain, bool depth); + virtual RenderTarget *createRenderTarget(int width, int height, GLenum format, GLsizei samples, bool depth); + + // Shader operations + virtual ShaderExecutable *loadExecutable(const void *function, size_t length, rx::ShaderType type); + virtual ShaderExecutable *compileToExecutable(gl::InfoLog &infoLog, const char *shaderHLSL, rx::ShaderType type); + + // Image operations + virtual Image *createImage(); + virtual void generateMipmap(Image *dest, Image *source); + virtual TextureStorage *createTextureStorage2D(SwapChain *swapChain); + virtual TextureStorage *createTextureStorage2D(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height); + virtual TextureStorage *createTextureStorageCube(int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size); + + // Buffer creation + virtual VertexBuffer *createVertexBuffer(); + virtual IndexBuffer *createIndexBuffer(); + virtual BufferStorage *createBufferStorage(); + + // Query and Fence creation + virtual QueryImpl *createQuery(GLenum type); + virtual FenceImpl *createFence(); + + // D3D9-renderer specific methods + bool boxFilter(IDirect3DSurface9 *source, IDirect3DSurface9 *dest); + + D3DPOOL getTexturePool(DWORD usage) const; + + private: + DISALLOW_COPY_AND_ASSIGN(Renderer9); + + void applyUniformnfv(gl::Uniform *targetUniform, const GLfloat *v); + void applyUniformniv(gl::Uniform *targetUniform, const GLint *v); + void applyUniformnbv(gl::Uniform *targetUniform, const GLint *v); + + void drawLineLoop(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer); + void drawIndexedPoints(GLsizei count, GLenum type, const GLvoid *indices, gl::Buffer *elementArrayBuffer); + + void getMultiSampleSupport(D3DFORMAT format, bool *multiSampleArray); + bool copyToRenderTarget(IDirect3DSurface9 *dest, IDirect3DSurface9 *source, bool fromManaged); + gl::Renderbuffer *getNullColorbuffer(gl::Renderbuffer *depthbuffer); + + D3DPOOL getBufferPool(DWORD usage) const; + + HMODULE mD3d9Module; + HDC mDc; + + void initializeDevice(); + D3DPRESENT_PARAMETERS getDefaultPresentParameters(); + void releaseDeviceResources(); + + UINT mAdapter; + D3DDEVTYPE mDeviceType; + bool mSoftwareDevice; // FIXME: Deprecate + IDirect3D9 *mD3d9; // Always valid after successful initialization. + IDirect3D9Ex *mD3d9Ex; // Might be null if D3D9Ex is not supported. + IDirect3DDevice9 *mDevice; + IDirect3DDevice9Ex *mDeviceEx; // Might be null if D3D9Ex is not supported. + + Blit *mBlit; + + HWND mDeviceWindow; + + bool mDeviceLost; + D3DCAPS9 mDeviceCaps; + D3DADAPTER_IDENTIFIER9 mAdapterIdentifier; + + D3DPRIMITIVETYPE mPrimitiveType; + int mPrimitiveCount; + GLsizei mRepeatDraw; + + bool mSceneStarted; + bool mSupportsNonPower2Textures; + bool mSupportsTextureFilterAnisotropy; + int mMinSwapInterval; + int mMaxSwapInterval; + + bool mOcclusionQuerySupport; + bool mEventQuerySupport; + bool mVertexTextureSupport; + + bool mDepthTextureSupport; + + bool mFloat32TextureSupport; + bool mFloat32FilterSupport; + bool mFloat32RenderSupport; + + bool mFloat16TextureSupport; + bool mFloat16FilterSupport; + bool mFloat16RenderSupport; + + bool mDXT1TextureSupport; + bool mDXT3TextureSupport; + bool mDXT5TextureSupport; + + bool mLuminanceTextureSupport; + bool mLuminanceAlphaTextureSupport; + + std::map<D3DFORMAT, bool *> mMultiSampleSupport; + GLsizei mMaxSupportedSamples; + + // current render target states + unsigned int mAppliedRenderTargetSerial; + unsigned int mAppliedDepthbufferSerial; + unsigned int mAppliedStencilbufferSerial; + bool mDepthStencilInitialized; + bool mRenderTargetDescInitialized; + rx::RenderTarget::Desc mRenderTargetDesc; + unsigned int mCurStencilSize; + unsigned int mCurDepthSize; + + IDirect3DStateBlock9 *mMaskedClearSavedState; + + // previously set render states + bool mForceSetDepthStencilState; + gl::DepthStencilState mCurDepthStencilState; + int mCurStencilRef; + int mCurStencilBackRef; + bool mCurFrontFaceCCW; + + bool mForceSetRasterState; + gl::RasterizerState mCurRasterState; + + bool mForceSetScissor; + gl::Rectangle mCurScissor; + bool mScissorEnabled; + + bool mForceSetViewport; + gl::Rectangle mCurViewport; + float mCurNear; + float mCurFar; + + bool mForceSetBlendState; + gl::BlendState mCurBlendState; + gl::Color mCurBlendColor; + GLuint mCurSampleMask; + + // Currently applied sampler states + bool mForceSetVertexSamplerStates[gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS]; + gl::SamplerState mCurVertexSamplerStates[gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS]; + + bool mForceSetPixelSamplerStates[gl::MAX_TEXTURE_IMAGE_UNITS]; + gl::SamplerState mCurPixelSamplerStates[gl::MAX_TEXTURE_IMAGE_UNITS]; + + // Currently applied textures + unsigned int mCurVertexTextureSerials[gl::IMPLEMENTATION_MAX_VERTEX_TEXTURE_IMAGE_UNITS]; + unsigned int mCurPixelTextureSerials[gl::MAX_TEXTURE_IMAGE_UNITS]; + + unsigned int mAppliedIBSerial; + unsigned int mAppliedProgramBinarySerial; + + rx::dx_VertexConstants mVertexConstants; + rx::dx_PixelConstants mPixelConstants; + bool mDxUniformsDirty; + + // A pool of event queries that are currently unused. + std::vector<IDirect3DQuery9*> mEventQueryPool; + VertexShaderCache mVertexShaderCache; + PixelShaderCache mPixelShaderCache; + + VertexDataManager *mVertexDataManager; + VertexDeclarationCache mVertexDeclarationCache; + + IndexDataManager *mIndexDataManager; + StreamingIndexBufferInterface *mLineLoopIB; + + enum { NUM_NULL_COLORBUFFER_CACHE_ENTRIES = 12 }; + struct NullColorbufferCacheEntry + { + UINT lruCount; + int width; + int height; + gl::Renderbuffer *buffer; + } mNullColorbufferCache[NUM_NULL_COLORBUFFER_CACHE_ENTRIES]; + UINT mMaxNullColorbufferLRU; + +}; + +} +#endif // LIBGLESV2_RENDERER_RENDERER9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/ShaderCache.h b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderCache.h new file mode 100644 index 0000000000..4391ac271a --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderCache.h @@ -0,0 +1,110 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ShaderCache: Defines rx::ShaderCache, a cache of Direct3D shader objects +// keyed by their byte code. + +#ifndef LIBGLESV2_RENDERER_SHADER_CACHE_H_ +#define LIBGLESV2_RENDERER_SHADER_CACHE_H_ + +#include "common/debug.h" + +namespace rx +{ +template <typename ShaderObject> +class ShaderCache +{ + public: + ShaderCache() : mDevice(NULL) + { + } + + ~ShaderCache() + { + // Call clear while the device is still valid. + ASSERT(mMap.empty()); + } + + void initialize(IDirect3DDevice9* device) + { + mDevice = device; + } + + ShaderObject *create(const DWORD *function, size_t length) + { + std::string key(reinterpret_cast<const char*>(function), length); + typename Map::iterator it = mMap.find(key); + if (it != mMap.end()) + { + it->second->AddRef(); + return it->second; + } + + ShaderObject *shader; + HRESULT result = createShader(function, &shader); + if (FAILED(result)) + { + return NULL; + } + + // Random eviction policy. + if (mMap.size() >= kMaxMapSize) + { + mMap.begin()->second->Release(); + mMap.erase(mMap.begin()); + } + + shader->AddRef(); + mMap[key] = shader; + + return shader; + } + + void clear() + { + for (typename Map::iterator it = mMap.begin(); it != mMap.end(); ++it) + { + it->second->Release(); + } + + mMap.clear(); + } + + private: + DISALLOW_COPY_AND_ASSIGN(ShaderCache); + + const static size_t kMaxMapSize = 100; + + HRESULT createShader(const DWORD *function, IDirect3DVertexShader9 **shader) + { + return mDevice->CreateVertexShader(function, shader); + } + + HRESULT createShader(const DWORD *function, IDirect3DPixelShader9 **shader) + { + return mDevice->CreatePixelShader(function, shader); + } + +#ifndef HASH_MAP +# ifdef _MSC_VER +# define HASH_MAP stdext::hash_map +# else +# define HASH_MAP std::unordered_map +# endif +#endif + + typedef HASH_MAP<std::string, ShaderObject*> Map; + Map mMap; + + IDirect3DDevice9 *mDevice; +}; + +typedef ShaderCache<IDirect3DVertexShader9> VertexShaderCache; +typedef ShaderCache<IDirect3DPixelShader9> PixelShaderCache; + +} + +#endif // LIBGLESV2_RENDERER_SHADER_CACHE_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable.h b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable.h new file mode 100644 index 0000000000..128d123fbe --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable.h @@ -0,0 +1,51 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ShaderExecutable.h: Defines a renderer-agnostic class to contain shader +// executable implementation details. + +#ifndef LIBGLESV2_RENDERER_SHADEREXECUTABLE_H_ +#define LIBGLESV2_RENDERER_SHADEREXECUTABLE_H_ + +#include "common/angleutils.h" + +namespace rx +{ + +class ShaderExecutable +{ + public: + ShaderExecutable(const void *function, size_t length) : mLength(length) + { + mFunction = new char[length]; + memcpy(mFunction, function, length); + } + + virtual ~ShaderExecutable() + { + delete mFunction; + } + + void *getFunction() const + { + return mFunction; + } + + size_t getLength() const + { + return mLength; + } + + private: + DISALLOW_COPY_AND_ASSIGN(ShaderExecutable); + + void *mFunction; + const size_t mLength; +}; + +} + +#endif // LIBGLESV2_RENDERER_SHADEREXECUTABLE9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable11.cpp new file mode 100644 index 0000000000..e1eb560334 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable11.cpp @@ -0,0 +1,109 @@ +#include "precompiled.h" +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ShaderExecutable11.cpp: Implements a D3D11-specific class to contain shader +// executable implementation details. + +#include "libGLESv2/renderer/ShaderExecutable11.h" + +#include "common/debug.h" + +namespace rx +{ + +ShaderExecutable11::ShaderExecutable11(const void *function, size_t length, ID3D11PixelShader *executable) + : ShaderExecutable(function, length) +{ + mPixelExecutable = executable; + mVertexExecutable = NULL; + mGeometryExecutable = NULL; + + mConstantBuffer = NULL; +} + +ShaderExecutable11::ShaderExecutable11(const void *function, size_t length, ID3D11VertexShader *executable) + : ShaderExecutable(function, length) +{ + mVertexExecutable = executable; + mPixelExecutable = NULL; + mGeometryExecutable = NULL; + + mConstantBuffer = NULL; +} + +ShaderExecutable11::ShaderExecutable11(const void *function, size_t length, ID3D11GeometryShader *executable) + : ShaderExecutable(function, length) +{ + mGeometryExecutable = executable; + mVertexExecutable = NULL; + mPixelExecutable = NULL; + + mConstantBuffer = NULL; +} + +ShaderExecutable11::~ShaderExecutable11() +{ + if (mVertexExecutable) + { + mVertexExecutable->Release(); + } + if (mPixelExecutable) + { + mPixelExecutable->Release(); + } + if (mGeometryExecutable) + { + mGeometryExecutable->Release(); + } + + if (mConstantBuffer) + { + mConstantBuffer->Release(); + } +} + +ShaderExecutable11 *ShaderExecutable11::makeShaderExecutable11(ShaderExecutable *executable) +{ + ASSERT(HAS_DYNAMIC_TYPE(ShaderExecutable11*, executable)); + return static_cast<ShaderExecutable11*>(executable); +} + +ID3D11VertexShader *ShaderExecutable11::getVertexShader() const +{ + return mVertexExecutable; +} + +ID3D11PixelShader *ShaderExecutable11::getPixelShader() const +{ + return mPixelExecutable; +} + +ID3D11GeometryShader *ShaderExecutable11::getGeometryShader() const +{ + return mGeometryExecutable; +} + +ID3D11Buffer *ShaderExecutable11::getConstantBuffer(ID3D11Device *device, unsigned int registerCount) +{ + if (!mConstantBuffer && registerCount > 0) + { + D3D11_BUFFER_DESC constantBufferDescription = {0}; + constantBufferDescription.ByteWidth = registerCount * sizeof(float[4]); + constantBufferDescription.Usage = D3D11_USAGE_DYNAMIC; + constantBufferDescription.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + constantBufferDescription.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + constantBufferDescription.MiscFlags = 0; + constantBufferDescription.StructureByteStride = 0; + + HRESULT result = device->CreateBuffer(&constantBufferDescription, NULL, &mConstantBuffer); + ASSERT(SUCCEEDED(result)); + } + + return mConstantBuffer; +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable11.h b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable11.h new file mode 100644 index 0000000000..c6ec1cf7d2 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable11.h @@ -0,0 +1,47 @@ +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ShaderExecutable11.h: Defines a D3D11-specific class to contain shader +// executable implementation details. + +#ifndef LIBGLESV2_RENDERER_SHADEREXECUTABLE11_H_ +#define LIBGLESV2_RENDERER_SHADEREXECUTABLE11_H_ + +#include "libGLESv2/renderer/ShaderExecutable.h" + +namespace rx +{ + +class ShaderExecutable11 : public ShaderExecutable +{ + public: + ShaderExecutable11(const void *function, size_t length, ID3D11PixelShader *executable); + ShaderExecutable11(const void *function, size_t length, ID3D11VertexShader *executable); + ShaderExecutable11(const void *function, size_t length, ID3D11GeometryShader *executable); + + virtual ~ShaderExecutable11(); + + static ShaderExecutable11 *makeShaderExecutable11(ShaderExecutable *executable); + + ID3D11PixelShader *getPixelShader() const; + ID3D11VertexShader *getVertexShader() const; + ID3D11GeometryShader *getGeometryShader() const; + + ID3D11Buffer *getConstantBuffer(ID3D11Device *device, unsigned int registerCount); + + private: + DISALLOW_COPY_AND_ASSIGN(ShaderExecutable11); + + ID3D11PixelShader *mPixelExecutable; + ID3D11VertexShader *mVertexExecutable; + ID3D11GeometryShader *mGeometryExecutable; + + ID3D11Buffer *mConstantBuffer; +}; + +} + +#endif // LIBGLESV2_RENDERER_SHADEREXECUTABLE11_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable9.cpp new file mode 100644 index 0000000000..98868a3fbf --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable9.cpp @@ -0,0 +1,60 @@ +#include "precompiled.h" +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ShaderExecutable9.cpp: Implements a D3D9-specific class to contain shader +// executable implementation details. + +#include "libGLESv2/renderer/ShaderExecutable9.h" + +#include "common/debug.h" + +namespace rx +{ + +ShaderExecutable9::ShaderExecutable9(const void *function, size_t length, IDirect3DPixelShader9 *executable) + : ShaderExecutable(function, length) +{ + mPixelExecutable = executable; + mVertexExecutable = NULL; +} + +ShaderExecutable9::ShaderExecutable9(const void *function, size_t length, IDirect3DVertexShader9 *executable) + : ShaderExecutable(function, length) +{ + mVertexExecutable = executable; + mPixelExecutable = NULL; +} + +ShaderExecutable9::~ShaderExecutable9() +{ + if (mVertexExecutable) + { + mVertexExecutable->Release(); + } + if (mPixelExecutable) + { + mPixelExecutable->Release(); + } +} + +ShaderExecutable9 *ShaderExecutable9::makeShaderExecutable9(ShaderExecutable *executable) +{ + ASSERT(HAS_DYNAMIC_TYPE(ShaderExecutable9*, executable)); + return static_cast<ShaderExecutable9*>(executable); +} + +IDirect3DVertexShader9 *ShaderExecutable9::getVertexShader() const +{ + return mVertexExecutable; +} + +IDirect3DPixelShader9 *ShaderExecutable9::getPixelShader() const +{ + return mPixelExecutable; +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable9.h b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable9.h new file mode 100644 index 0000000000..fa1e6c2844 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/ShaderExecutable9.h @@ -0,0 +1,39 @@ +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ShaderExecutable9.h: Defines a D3D9-specific class to contain shader +// executable implementation details. + +#ifndef LIBGLESV2_RENDERER_SHADEREXECUTABLE9_H_ +#define LIBGLESV2_RENDERER_SHADEREXECUTABLE9_H_ + +#include "libGLESv2/renderer/ShaderExecutable.h" + +namespace rx +{ + +class ShaderExecutable9 : public ShaderExecutable +{ + public: + ShaderExecutable9(const void *function, size_t length, IDirect3DPixelShader9 *executable); + ShaderExecutable9(const void *function, size_t length, IDirect3DVertexShader9 *executable); + virtual ~ShaderExecutable9(); + + static ShaderExecutable9 *makeShaderExecutable9(ShaderExecutable *executable); + + IDirect3DPixelShader9 *getPixelShader() const; + IDirect3DVertexShader9 *getVertexShader() const; + + private: + DISALLOW_COPY_AND_ASSIGN(ShaderExecutable9); + + IDirect3DPixelShader9 *mPixelExecutable; + IDirect3DVertexShader9 *mVertexExecutable; +}; + +} + +#endif // LIBGLESV2_RENDERER_SHADEREXECUTABLE9_H_
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain.h b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain.h new file mode 100644 index 0000000000..14c0515fc8 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain.h @@ -0,0 +1,44 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChain.h: Defines a back-end specific class that hides the details of the +// implementation-specific swapchain. + +#ifndef LIBGLESV2_RENDERER_SWAPCHAIN_H_ +#define LIBGLESV2_RENDERER_SWAPCHAIN_H_ + +#include "common/angleutils.h" + +namespace rx +{ + +class SwapChain +{ + public: + SwapChain(HWND window, HANDLE shareHandle, GLenum backBufferFormat, GLenum depthBufferFormat) + : mWindow(window), mShareHandle(shareHandle), mBackBufferFormat(backBufferFormat), mDepthBufferFormat(depthBufferFormat) + { + } + + virtual ~SwapChain() {}; + + virtual EGLint resize(EGLint backbufferWidth, EGLint backbufferSize) = 0; + virtual EGLint reset(EGLint backbufferWidth, EGLint backbufferHeight, EGLint swapInterval) = 0; + virtual EGLint swapRect(EGLint x, EGLint y, EGLint width, EGLint height) = 0; + virtual void recreate() = 0; + + virtual HANDLE getShareHandle() {return mShareHandle;}; + + protected: + const HWND mWindow; // Window that the surface is created for. + const GLenum mBackBufferFormat; + const GLenum mDepthBufferFormat; + + HANDLE mShareHandle; +}; + +} +#endif // LIBGLESV2_RENDERER_SWAPCHAIN_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain11.cpp new file mode 100644 index 0000000000..87422be727 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain11.cpp @@ -0,0 +1,767 @@ +#include "precompiled.h" +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChain11.cpp: Implements a back-end specific class for the D3D11 swap chain. + +#include "libGLESv2/renderer/SwapChain11.h" + +#include "libGLESv2/renderer/renderer11_utils.h" +#include "libGLESv2/renderer/Renderer11.h" +#include "libGLESv2/renderer/shaders/compiled/passthrough11vs.h" +#include "libGLESv2/renderer/shaders/compiled/passthroughrgba11ps.h" + +namespace rx +{ + +SwapChain11::SwapChain11(Renderer11 *renderer, HWND window, HANDLE shareHandle, + GLenum backBufferFormat, GLenum depthBufferFormat) + : mRenderer(renderer), SwapChain(window, shareHandle, backBufferFormat, depthBufferFormat) +{ + mSwapChain = NULL; + mBackBufferTexture = NULL; + mBackBufferRTView = NULL; + mOffscreenTexture = NULL; + mOffscreenRTView = NULL; + mOffscreenSRView = NULL; + mDepthStencilTexture = NULL; + mDepthStencilDSView = NULL; + mQuadVB = NULL; + mPassThroughSampler = NULL; + mPassThroughIL = NULL; + mPassThroughVS = NULL; + mPassThroughPS = NULL; + mWidth = -1; + mHeight = -1; + mSwapInterval = 0; + mAppCreatedShareHandle = mShareHandle != NULL; + mPassThroughResourcesInit = false; +} + +SwapChain11::~SwapChain11() +{ + release(); +} + +void SwapChain11::release() +{ + if (mSwapChain) + { + mSwapChain->Release(); + mSwapChain = NULL; + } + + if (mBackBufferTexture) + { + mBackBufferTexture->Release(); + mBackBufferTexture = NULL; + } + + if (mBackBufferRTView) + { + mBackBufferRTView->Release(); + mBackBufferRTView = NULL; + } + + if (mOffscreenTexture) + { + mOffscreenTexture->Release(); + mOffscreenTexture = NULL; + } + + if (mOffscreenRTView) + { + mOffscreenRTView->Release(); + mOffscreenRTView = NULL; + } + + if (mOffscreenSRView) + { + mOffscreenSRView->Release(); + mOffscreenSRView = NULL; + } + + if (mDepthStencilTexture) + { + mDepthStencilTexture->Release(); + mDepthStencilTexture = NULL; + } + + if (mDepthStencilDSView) + { + mDepthStencilDSView->Release(); + mDepthStencilDSView = NULL; + } + + if (mQuadVB) + { + mQuadVB->Release(); + mQuadVB = NULL; + } + + if (mPassThroughSampler) + { + mPassThroughSampler->Release(); + mPassThroughSampler = NULL; + } + + if (mPassThroughIL) + { + mPassThroughIL->Release(); + mPassThroughIL = NULL; + } + + if (mPassThroughVS) + { + mPassThroughVS->Release(); + mPassThroughVS = NULL; + } + + if (mPassThroughPS) + { + mPassThroughPS->Release(); + mPassThroughPS = NULL; + } + + if (!mAppCreatedShareHandle) + { + mShareHandle = NULL; + } +} + +void SwapChain11::releaseOffscreenTexture() +{ + if (mOffscreenTexture) + { + mOffscreenTexture->Release(); + mOffscreenTexture = NULL; + } + + if (mOffscreenRTView) + { + mOffscreenRTView->Release(); + mOffscreenRTView = NULL; + } + + if (mOffscreenSRView) + { + mOffscreenSRView->Release(); + mOffscreenSRView = NULL; + } + + if (mDepthStencilTexture) + { + mDepthStencilTexture->Release(); + mDepthStencilTexture = NULL; + } + + if (mDepthStencilDSView) + { + mDepthStencilDSView->Release(); + mDepthStencilDSView = NULL; + } +} + +EGLint SwapChain11::resetOffscreenTexture(int backbufferWidth, int backbufferHeight) +{ + ID3D11Device *device = mRenderer->getDevice(); + + ASSERT(device != NULL); + + // D3D11 does not allow zero size textures + ASSERT(backbufferWidth >= 1); + ASSERT(backbufferHeight >= 1); + + // Preserve the render target content + ID3D11Texture2D *previousOffscreenTexture = mOffscreenTexture; + if (previousOffscreenTexture) + { + previousOffscreenTexture->AddRef(); + } + const int previousWidth = mWidth; + const int previousHeight = mHeight; + + releaseOffscreenTexture(); + + // If the app passed in a share handle, open the resource + // See EGL_ANGLE_d3d_share_handle_client_buffer + if (mAppCreatedShareHandle) + { + ID3D11Resource *tempResource11; + HRESULT result = device->OpenSharedResource(mShareHandle, __uuidof(ID3D11Resource), (void**)&tempResource11); + + if (FAILED(result)) + { + ERR("Failed to open the swap chain pbuffer share handle: %08lX", result); + release(); + return EGL_BAD_PARAMETER; + } + + result = tempResource11->QueryInterface(__uuidof(ID3D11Texture2D), (void**)&mOffscreenTexture); + tempResource11->Release(); + + if (FAILED(result)) + { + ERR("Failed to query texture2d interface in pbuffer share handle: %08lX", result); + release(); + return EGL_BAD_PARAMETER; + } + + // Validate offscreen texture parameters + D3D11_TEXTURE2D_DESC offscreenTextureDesc = {0}; + mOffscreenTexture->GetDesc(&offscreenTextureDesc); + + if (offscreenTextureDesc.Width != (UINT)backbufferWidth + || offscreenTextureDesc.Height != (UINT)backbufferHeight + || offscreenTextureDesc.Format != gl_d3d11::ConvertRenderbufferFormat(mBackBufferFormat) + || offscreenTextureDesc.MipLevels != 1 + || offscreenTextureDesc.ArraySize != 1) + { + ERR("Invalid texture parameters in the shared offscreen texture pbuffer"); + release(); + return EGL_BAD_PARAMETER; + } + } + else + { + const bool useSharedResource = !mWindow && mRenderer->getShareHandleSupport(); + + D3D11_TEXTURE2D_DESC offscreenTextureDesc = {0}; + offscreenTextureDesc.Width = backbufferWidth; + offscreenTextureDesc.Height = backbufferHeight; + offscreenTextureDesc.Format = gl_d3d11::ConvertRenderbufferFormat(mBackBufferFormat); + offscreenTextureDesc.MipLevels = 1; + offscreenTextureDesc.ArraySize = 1; + offscreenTextureDesc.SampleDesc.Count = 1; + offscreenTextureDesc.SampleDesc.Quality = 0; + offscreenTextureDesc.Usage = D3D11_USAGE_DEFAULT; + offscreenTextureDesc.BindFlags = D3D11_BIND_RENDER_TARGET | D3D11_BIND_SHADER_RESOURCE; + offscreenTextureDesc.CPUAccessFlags = 0; + offscreenTextureDesc.MiscFlags = useSharedResource ? D3D11_RESOURCE_MISC_SHARED : 0; + + HRESULT result = device->CreateTexture2D(&offscreenTextureDesc, NULL, &mOffscreenTexture); + + if (FAILED(result)) + { + ERR("Could not create offscreen texture: %08lX", result); + release(); + + if (d3d11::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + + d3d11::SetDebugName(mOffscreenTexture, "Offscreen texture"); + + // EGL_ANGLE_surface_d3d_texture_2d_share_handle requires that we store a share handle for the client + if (useSharedResource) + { + IDXGIResource *offscreenTextureResource = NULL; + result = mOffscreenTexture->QueryInterface(__uuidof(IDXGIResource), (void**)&offscreenTextureResource); + + // Fall back to no share handle on failure + if (FAILED(result)) + { + ERR("Could not query offscreen texture resource: %08lX", result); + } + else + { + result = offscreenTextureResource->GetSharedHandle(&mShareHandle); + + if (FAILED(result)) + { + mShareHandle = NULL; + ERR("Could not get offscreen texture shared handle: %08lX", result); + } + } + } + } + + HRESULT result = device->CreateRenderTargetView(mOffscreenTexture, NULL, &mOffscreenRTView); + + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mOffscreenRTView, "Offscreen render target"); + + result = device->CreateShaderResourceView(mOffscreenTexture, NULL, &mOffscreenSRView); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mOffscreenSRView, "Offscreen shader resource"); + + if (mDepthBufferFormat != GL_NONE) + { + D3D11_TEXTURE2D_DESC depthStencilDesc = {0}; + depthStencilDesc.Width = backbufferWidth; + depthStencilDesc.Height = backbufferHeight; + depthStencilDesc.Format = gl_d3d11::ConvertRenderbufferFormat(mDepthBufferFormat); + depthStencilDesc.MipLevels = 1; + depthStencilDesc.ArraySize = 1; + depthStencilDesc.SampleDesc.Count = 1; + depthStencilDesc.SampleDesc.Quality = 0; + depthStencilDesc.Usage = D3D11_USAGE_DEFAULT; + depthStencilDesc.BindFlags = D3D11_BIND_DEPTH_STENCIL; + depthStencilDesc.CPUAccessFlags = 0; + depthStencilDesc.MiscFlags = 0; + + result = device->CreateTexture2D(&depthStencilDesc, NULL, &mDepthStencilTexture); + if (FAILED(result)) + { + ERR("Could not create depthstencil surface for new swap chain: 0x%08X", result); + release(); + + if (d3d11::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + d3d11::SetDebugName(mDepthStencilTexture, "Depth stencil texture"); + + result = device->CreateDepthStencilView(mDepthStencilTexture, NULL, &mDepthStencilDSView); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mDepthStencilDSView, "Depth stencil view"); + } + + mWidth = backbufferWidth; + mHeight = backbufferHeight; + + if (previousOffscreenTexture != NULL) + { + D3D11_BOX sourceBox = {0}; + sourceBox.left = 0; + sourceBox.right = std::min(previousWidth, mWidth); + sourceBox.top = std::max(previousHeight - mHeight, 0); + sourceBox.bottom = previousHeight; + sourceBox.front = 0; + sourceBox.back = 1; + + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + const int yoffset = std::max(mHeight - previousHeight, 0); + deviceContext->CopySubresourceRegion(mOffscreenTexture, 0, 0, yoffset, 0, previousOffscreenTexture, 0, &sourceBox); + + previousOffscreenTexture->Release(); + + if (mSwapChain) + { + swapRect(0, 0, mWidth, mHeight); + } + } + + return EGL_SUCCESS; +} + +EGLint SwapChain11::resize(EGLint backbufferWidth, EGLint backbufferHeight) +{ + ID3D11Device *device = mRenderer->getDevice(); + + if (device == NULL) + { + return EGL_BAD_ACCESS; + } + + // Can only call resize if we have already created our swap buffer and resources + ASSERT(mSwapChain && mBackBufferTexture && mBackBufferRTView); + + if (mBackBufferTexture) + { + mBackBufferTexture->Release(); + mBackBufferTexture = NULL; + } + + if (mBackBufferRTView) + { + mBackBufferRTView->Release(); + mBackBufferRTView = NULL; + } + + // Resize swap chain + DXGI_FORMAT backbufferDXGIFormat = gl_d3d11::ConvertRenderbufferFormat(mBackBufferFormat); + HRESULT result = mSwapChain->ResizeBuffers(2, backbufferWidth, backbufferHeight, backbufferDXGIFormat, 0); + + if (FAILED(result)) + { + ERR("Error resizing swap chain buffers: 0x%08X", result); + release(); + + if (d3d11::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + + result = mSwapChain->GetBuffer(0, __uuidof(ID3D11Texture2D), (LPVOID*)&mBackBufferTexture); + ASSERT(SUCCEEDED(result)); + if (SUCCEEDED(result)) + { + d3d11::SetDebugName(mBackBufferTexture, "Back buffer texture"); + } + + result = device->CreateRenderTargetView(mBackBufferTexture, NULL, &mBackBufferRTView); + ASSERT(SUCCEEDED(result)); + if (SUCCEEDED(result)) + { + d3d11::SetDebugName(mBackBufferRTView, "Back buffer render target"); + } + + return resetOffscreenTexture(backbufferWidth, backbufferHeight); +} + +EGLint SwapChain11::reset(int backbufferWidth, int backbufferHeight, EGLint swapInterval) +{ + ID3D11Device *device = mRenderer->getDevice(); + + if (device == NULL) + { + return EGL_BAD_ACCESS; + } + + // Release specific resources to free up memory for the new render target, while the + // old render target still exists for the purpose of preserving its contents. + if (mSwapChain) + { + mSwapChain->Release(); + mSwapChain = NULL; + } + + if (mBackBufferTexture) + { + mBackBufferTexture->Release(); + mBackBufferTexture = NULL; + } + + if (mBackBufferRTView) + { + mBackBufferRTView->Release(); + mBackBufferRTView = NULL; + } + + mSwapInterval = static_cast<unsigned int>(swapInterval); + if (mSwapInterval > 4) + { + // IDXGISwapChain::Present documentation states that valid sync intervals are in the [0,4] range + return EGL_BAD_PARAMETER; + } + + // EGL allows creating a surface with 0x0 dimension, however, DXGI does not like 0x0 swapchains + if (backbufferWidth < 1 || backbufferHeight < 1) + { + releaseOffscreenTexture(); + return EGL_SUCCESS; + } + + if (mWindow) + { + // We cannot create a swap chain for an HWND that is owned by a different process + DWORD currentProcessId = GetCurrentProcessId(); + DWORD wndProcessId; + GetWindowThreadProcessId(mWindow, &wndProcessId); + + if (currentProcessId != wndProcessId) + { + ERR("Could not create swap chain, window owned by different process"); + release(); + return EGL_BAD_NATIVE_WINDOW; + } + + IDXGIFactory *factory = mRenderer->getDxgiFactory(); + + DXGI_SWAP_CHAIN_DESC swapChainDesc = {0}; + swapChainDesc.BufferCount = 2; + swapChainDesc.BufferDesc.Format = gl_d3d11::ConvertRenderbufferFormat(mBackBufferFormat); + swapChainDesc.BufferDesc.Width = backbufferWidth; + swapChainDesc.BufferDesc.Height = backbufferHeight; + swapChainDesc.BufferDesc.Scaling = DXGI_MODE_SCALING_UNSPECIFIED; + swapChainDesc.BufferDesc.ScanlineOrdering = DXGI_MODE_SCANLINE_ORDER_UNSPECIFIED; + swapChainDesc.BufferDesc.RefreshRate.Numerator = 0; + swapChainDesc.BufferDesc.RefreshRate.Denominator = 1; + swapChainDesc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT; + swapChainDesc.Flags = 0; + swapChainDesc.OutputWindow = mWindow; + swapChainDesc.SampleDesc.Count = 1; + swapChainDesc.SampleDesc.Quality = 0; + swapChainDesc.Windowed = TRUE; + + HRESULT result = factory->CreateSwapChain(device, &swapChainDesc, &mSwapChain); + + if (FAILED(result)) + { + ERR("Could not create additional swap chains or offscreen surfaces: %08lX", result); + release(); + + if (d3d11::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + + result = mSwapChain->GetBuffer(0, __uuidof(ID3D11Texture2D), (LPVOID*)&mBackBufferTexture); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mBackBufferTexture, "Back buffer texture"); + + result = device->CreateRenderTargetView(mBackBufferTexture, NULL, &mBackBufferRTView); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mBackBufferRTView, "Back buffer render target"); + } + + // If we are resizing the swap chain, we don't wish to recreate all the static resources + if (!mPassThroughResourcesInit) + { + mPassThroughResourcesInit = true; + initPassThroughResources(); + } + + return resetOffscreenTexture(backbufferWidth, backbufferHeight); +} + +void SwapChain11::initPassThroughResources() +{ + ID3D11Device *device = mRenderer->getDevice(); + + ASSERT(device != NULL); + + // Make sure our resources are all not allocated, when we create + ASSERT(mQuadVB == NULL && mPassThroughSampler == NULL); + ASSERT(mPassThroughIL == NULL && mPassThroughVS == NULL && mPassThroughPS == NULL); + + D3D11_BUFFER_DESC vbDesc; + vbDesc.ByteWidth = sizeof(d3d11::PositionTexCoordVertex) * 4; + vbDesc.Usage = D3D11_USAGE_DYNAMIC; + vbDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + vbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + vbDesc.MiscFlags = 0; + vbDesc.StructureByteStride = 0; + + HRESULT result = device->CreateBuffer(&vbDesc, NULL, &mQuadVB); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mQuadVB, "Swap chain quad vertex buffer"); + + D3D11_SAMPLER_DESC samplerDesc; + samplerDesc.Filter = D3D11_FILTER_MIN_MAG_MIP_POINT; + samplerDesc.AddressU = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.AddressV = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.AddressW = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.MipLODBias = 0.0f; + samplerDesc.MaxAnisotropy = 0; + samplerDesc.ComparisonFunc = D3D11_COMPARISON_NEVER; + samplerDesc.BorderColor[0] = 0.0f; + samplerDesc.BorderColor[1] = 0.0f; + samplerDesc.BorderColor[2] = 0.0f; + samplerDesc.BorderColor[3] = 0.0f; + samplerDesc.MinLOD = 0; + samplerDesc.MaxLOD = D3D11_FLOAT32_MAX; + + result = device->CreateSamplerState(&samplerDesc, &mPassThroughSampler); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPassThroughSampler, "Swap chain pass through sampler"); + + D3D11_INPUT_ELEMENT_DESC quadLayout[] = + { + { "POSITION", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + { "TEXCOORD", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 8, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + }; + + result = device->CreateInputLayout(quadLayout, 2, g_VS_Passthrough, sizeof(g_VS_Passthrough), &mPassThroughIL); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPassThroughIL, "Swap chain pass through layout"); + + result = device->CreateVertexShader(g_VS_Passthrough, sizeof(g_VS_Passthrough), NULL, &mPassThroughVS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPassThroughVS, "Swap chain pass through vertex shader"); + + result = device->CreatePixelShader(g_PS_PassthroughRGBA, sizeof(g_PS_PassthroughRGBA), NULL, &mPassThroughPS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPassThroughPS, "Swap chain pass through pixel shader"); +} + +// parameters should be validated/clamped by caller +EGLint SwapChain11::swapRect(EGLint x, EGLint y, EGLint width, EGLint height) +{ + if (!mSwapChain) + { + return EGL_SUCCESS; + } + + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + // Set vertices + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = deviceContext->Map(mQuadVB, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return EGL_BAD_ACCESS; + } + + d3d11::PositionTexCoordVertex *vertices = static_cast<d3d11::PositionTexCoordVertex*>(mappedResource.pData); + + // Create a quad in homogeneous coordinates + float x1 = (x / float(mWidth)) * 2.0f - 1.0f; + float y1 = (y / float(mHeight)) * 2.0f - 1.0f; + float x2 = ((x + width) / float(mWidth)) * 2.0f - 1.0f; + float y2 = ((y + height) / float(mHeight)) * 2.0f - 1.0f; + + float u1 = x / float(mWidth); + float v1 = y / float(mHeight); + float u2 = (x + width) / float(mWidth); + float v2 = (y + height) / float(mHeight); + + d3d11::SetPositionTexCoordVertex(&vertices[0], x1, y1, u1, v1); + d3d11::SetPositionTexCoordVertex(&vertices[1], x1, y2, u1, v2); + d3d11::SetPositionTexCoordVertex(&vertices[2], x2, y1, u2, v1); + d3d11::SetPositionTexCoordVertex(&vertices[3], x2, y2, u2, v2); + + deviceContext->Unmap(mQuadVB, 0); + + static UINT stride = sizeof(d3d11::PositionTexCoordVertex); + static UINT startIdx = 0; + deviceContext->IASetVertexBuffers(0, 1, &mQuadVB, &stride, &startIdx); + + // Apply state + deviceContext->OMSetDepthStencilState(NULL, 0xFFFFFFFF); + + static const float blendFactor[4] = { 1.0f, 1.0f, 1.0f, 1.0f }; + deviceContext->OMSetBlendState(NULL, blendFactor, 0xFFFFFFF); + + deviceContext->RSSetState(NULL); + + // Apply shaders + deviceContext->IASetInputLayout(mPassThroughIL); + deviceContext->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP); + deviceContext->VSSetShader(mPassThroughVS, NULL, 0); + deviceContext->PSSetShader(mPassThroughPS, NULL, 0); + deviceContext->GSSetShader(NULL, NULL, 0); + + // Apply render targets + mRenderer->setOneTimeRenderTarget(mBackBufferRTView); + + // Set the viewport + D3D11_VIEWPORT viewport; + viewport.TopLeftX = 0; + viewport.TopLeftY = 0; + viewport.Width = mWidth; + viewport.Height = mHeight; + viewport.MinDepth = 0.0f; + viewport.MaxDepth = 1.0f; + deviceContext->RSSetViewports(1, &viewport); + + // Apply textures + deviceContext->PSSetShaderResources(0, 1, &mOffscreenSRView); + deviceContext->PSSetSamplers(0, 1, &mPassThroughSampler); + + // Draw + deviceContext->Draw(4, 0); + result = mSwapChain->Present(mSwapInterval, 0); + + if (result == DXGI_ERROR_DEVICE_REMOVED) + { + HRESULT removedReason = device->GetDeviceRemovedReason(); + ERR("Present failed: the D3D11 device was removed: 0x%08X", removedReason); + return EGL_CONTEXT_LOST; + } + else if (result == DXGI_ERROR_DEVICE_RESET) + { + ERR("Present failed: the D3D11 device was reset from a bad command."); + return EGL_CONTEXT_LOST; + } + else if (FAILED(result)) + { + ERR("Present failed with error code 0x%08X", result); + } + + // Unbind + static ID3D11ShaderResourceView *const nullSRV = NULL; + deviceContext->PSSetShaderResources(0, 1, &nullSRV); + + mRenderer->unapplyRenderTargets(); + mRenderer->markAllStateDirty(); + + return EGL_SUCCESS; +} + +// Increments refcount on texture. +// caller must Release() the returned texture +ID3D11Texture2D *SwapChain11::getOffscreenTexture() +{ + if (mOffscreenTexture) + { + mOffscreenTexture->AddRef(); + } + + return mOffscreenTexture; +} + +// Increments refcount on view. +// caller must Release() the returned view +ID3D11RenderTargetView *SwapChain11::getRenderTarget() +{ + if (mOffscreenRTView) + { + mOffscreenRTView->AddRef(); + } + + return mOffscreenRTView; +} + +// Increments refcount on view. +// caller must Release() the returned view +ID3D11ShaderResourceView *SwapChain11::getRenderTargetShaderResource() +{ + if (mOffscreenSRView) + { + mOffscreenSRView->AddRef(); + } + + return mOffscreenSRView; +} + +// Increments refcount on view. +// caller must Release() the returned view +ID3D11DepthStencilView *SwapChain11::getDepthStencil() +{ + if (mDepthStencilDSView) + { + mDepthStencilDSView->AddRef(); + } + + return mDepthStencilDSView; +} + +ID3D11Texture2D *SwapChain11::getDepthStencilTexture() +{ + if (mDepthStencilTexture) + { + mDepthStencilTexture->AddRef(); + } + + return mDepthStencilTexture; +} + +SwapChain11 *SwapChain11::makeSwapChain11(SwapChain *swapChain) +{ + ASSERT(HAS_DYNAMIC_TYPE(rx::SwapChain11*, swapChain)); + return static_cast<rx::SwapChain11*>(swapChain); +} + +void SwapChain11::recreate() +{ + // possibly should use this method instead of reset +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain11.h b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain11.h new file mode 100644 index 0000000000..800104602e --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain11.h @@ -0,0 +1,78 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChain11.h: Defines a back-end specific class for the D3D11 swap chain. + +#ifndef LIBGLESV2_RENDERER_SWAPCHAIN11_H_ +#define LIBGLESV2_RENDERER_SWAPCHAIN11_H_ + +#include "common/angleutils.h" +#include "libGLESv2/renderer/SwapChain.h" + +namespace rx +{ +class Renderer11; + +class SwapChain11 : public SwapChain +{ + public: + SwapChain11(Renderer11 *renderer, HWND window, HANDLE shareHandle, + GLenum backBufferFormat, GLenum depthBufferFormat); + virtual ~SwapChain11(); + + EGLint resize(EGLint backbufferWidth, EGLint backbufferHeight); + virtual EGLint reset(EGLint backbufferWidth, EGLint backbufferHeight, EGLint swapInterval); + virtual EGLint swapRect(EGLint x, EGLint y, EGLint width, EGLint height); + virtual void recreate(); + + virtual ID3D11Texture2D *getOffscreenTexture(); + virtual ID3D11RenderTargetView *getRenderTarget(); + virtual ID3D11ShaderResourceView *getRenderTargetShaderResource(); + + virtual ID3D11Texture2D *getDepthStencilTexture(); + virtual ID3D11DepthStencilView *getDepthStencil(); + + EGLint getWidth() const { return mWidth; } + EGLint getHeight() const { return mHeight; } + + static SwapChain11 *makeSwapChain11(SwapChain *swapChain); + + private: + DISALLOW_COPY_AND_ASSIGN(SwapChain11); + + void release(); + void initPassThroughResources(); + void releaseOffscreenTexture(); + EGLint resetOffscreenTexture(int backbufferWidth, int backbufferHeight); + + Renderer11 *mRenderer; + EGLint mHeight; + EGLint mWidth; + bool mAppCreatedShareHandle; + unsigned int mSwapInterval; + bool mPassThroughResourcesInit; + + IDXGISwapChain *mSwapChain; + + ID3D11Texture2D *mBackBufferTexture; + ID3D11RenderTargetView *mBackBufferRTView; + + ID3D11Texture2D *mOffscreenTexture; + ID3D11RenderTargetView *mOffscreenRTView; + ID3D11ShaderResourceView *mOffscreenSRView; + + ID3D11Texture2D *mDepthStencilTexture; + ID3D11DepthStencilView *mDepthStencilDSView; + + ID3D11Buffer *mQuadVB; + ID3D11SamplerState *mPassThroughSampler; + ID3D11InputLayout *mPassThroughIL; + ID3D11VertexShader *mPassThroughVS; + ID3D11PixelShader *mPassThroughPS; +}; + +} +#endif // LIBGLESV2_RENDERER_SWAPCHAIN11_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain9.cpp new file mode 100644 index 0000000000..f57a874688 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain9.cpp @@ -0,0 +1,449 @@ +#include "precompiled.h" +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChain9.cpp: Implements a back-end specific class for the D3D9 swap chain. + +#include "libGLESv2/renderer/SwapChain9.h" +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/renderer/Renderer9.h" + +namespace rx +{ + +SwapChain9::SwapChain9(Renderer9 *renderer, HWND window, HANDLE shareHandle, + GLenum backBufferFormat, GLenum depthBufferFormat) + : mRenderer(renderer), SwapChain(window, shareHandle, backBufferFormat, depthBufferFormat) +{ + mSwapChain = NULL; + mBackBuffer = NULL; + mDepthStencil = NULL; + mRenderTarget = NULL; + mOffscreenTexture = NULL; + mWidth = -1; + mHeight = -1; + mSwapInterval = -1; +} + +SwapChain9::~SwapChain9() +{ + release(); +} + +void SwapChain9::release() +{ + if (mSwapChain) + { + mSwapChain->Release(); + mSwapChain = NULL; + } + + if (mBackBuffer) + { + mBackBuffer->Release(); + mBackBuffer = NULL; + } + + if (mDepthStencil) + { + mDepthStencil->Release(); + mDepthStencil = NULL; + } + + if (mRenderTarget) + { + mRenderTarget->Release(); + mRenderTarget = NULL; + } + + if (mOffscreenTexture) + { + mOffscreenTexture->Release(); + mOffscreenTexture = NULL; + } + + if (mWindow) + mShareHandle = NULL; +} + +static DWORD convertInterval(EGLint interval) +{ + switch(interval) + { + case 0: return D3DPRESENT_INTERVAL_IMMEDIATE; + case 1: return D3DPRESENT_INTERVAL_ONE; + case 2: return D3DPRESENT_INTERVAL_TWO; + case 3: return D3DPRESENT_INTERVAL_THREE; + case 4: return D3DPRESENT_INTERVAL_FOUR; + default: UNREACHABLE(); + } + + return D3DPRESENT_INTERVAL_DEFAULT; +} + +EGLint SwapChain9::resize(int backbufferWidth, int backbufferHeight) +{ + // D3D9 does not support resizing swap chains without recreating them + return reset(backbufferWidth, backbufferHeight, mSwapInterval); +} + +EGLint SwapChain9::reset(int backbufferWidth, int backbufferHeight, EGLint swapInterval) +{ + IDirect3DDevice9 *device = mRenderer->getDevice(); + + if (device == NULL) + { + return EGL_BAD_ACCESS; + } + + // Evict all non-render target textures to system memory and release all resources + // before reallocating them to free up as much video memory as possible. + device->EvictManagedResources(); + + HRESULT result; + + // Release specific resources to free up memory for the new render target, while the + // old render target still exists for the purpose of preserving its contents. + if (mSwapChain) + { + mSwapChain->Release(); + mSwapChain = NULL; + } + + if (mBackBuffer) + { + mBackBuffer->Release(); + mBackBuffer = NULL; + } + + if (mOffscreenTexture) + { + mOffscreenTexture->Release(); + mOffscreenTexture = NULL; + } + + if (mDepthStencil) + { + mDepthStencil->Release(); + mDepthStencil = NULL; + } + + HANDLE *pShareHandle = NULL; + if (!mWindow && mRenderer->getShareHandleSupport()) + { + pShareHandle = &mShareHandle; + } + + // CreateTexture will fail on zero dimensions, so just release old target + if (!backbufferWidth || !backbufferHeight) + { + if (mRenderTarget) + { + mRenderTarget->Release(); + mRenderTarget = NULL; + } + + mWidth = backbufferWidth; + mHeight = backbufferHeight; + + return EGL_SUCCESS; + } + + result = device->CreateTexture(backbufferWidth, backbufferHeight, 1, D3DUSAGE_RENDERTARGET, + gl_d3d9::ConvertRenderbufferFormat(mBackBufferFormat), D3DPOOL_DEFAULT, + &mOffscreenTexture, pShareHandle); + if (FAILED(result)) + { + ERR("Could not create offscreen texture: %08lX", result); + release(); + + if (d3d9::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + + IDirect3DSurface9 *oldRenderTarget = mRenderTarget; + + result = mOffscreenTexture->GetSurfaceLevel(0, &mRenderTarget); + ASSERT(SUCCEEDED(result)); + + if (oldRenderTarget) + { + RECT rect = + { + 0, 0, + mWidth, mHeight + }; + + if (rect.right > static_cast<LONG>(backbufferWidth)) + { + rect.right = backbufferWidth; + } + + if (rect.bottom > static_cast<LONG>(backbufferHeight)) + { + rect.bottom = backbufferHeight; + } + + mRenderer->endScene(); + + result = device->StretchRect(oldRenderTarget, &rect, mRenderTarget, &rect, D3DTEXF_NONE); + ASSERT(SUCCEEDED(result)); + + oldRenderTarget->Release(); + } + + if (mWindow) + { + D3DPRESENT_PARAMETERS presentParameters = {0}; + presentParameters.AutoDepthStencilFormat = gl_d3d9::ConvertRenderbufferFormat(mDepthBufferFormat); + presentParameters.BackBufferCount = 1; + presentParameters.BackBufferFormat = gl_d3d9::ConvertRenderbufferFormat(mBackBufferFormat); + presentParameters.EnableAutoDepthStencil = FALSE; + presentParameters.Flags = 0; + presentParameters.hDeviceWindow = mWindow; + presentParameters.MultiSampleQuality = 0; // FIXME: Unimplemented + presentParameters.MultiSampleType = D3DMULTISAMPLE_NONE; // FIXME: Unimplemented + presentParameters.PresentationInterval = convertInterval(swapInterval); + presentParameters.SwapEffect = D3DSWAPEFFECT_DISCARD; + presentParameters.Windowed = TRUE; + presentParameters.BackBufferWidth = backbufferWidth; + presentParameters.BackBufferHeight = backbufferHeight; + + // http://crbug.com/140239 + // http://crbug.com/143434 + // + // Some AMD/Intel switchable systems / drivers appear to round swap chain surfaces to a multiple of 64 pixels in width + // when using the integrated Intel. This rounds the width up rather than down. + // + // Some non-switchable AMD GPUs / drivers do not respect the source rectangle to Present. Therefore, when the vendor ID + // is not Intel, the back buffer width must be exactly the same width as the window or horizontal scaling will occur. + if (mRenderer->getAdapterVendor() == VENDOR_ID_INTEL) + { + presentParameters.BackBufferWidth = (presentParameters.BackBufferWidth + 63) / 64 * 64; + } + + result = device->CreateAdditionalSwapChain(&presentParameters, &mSwapChain); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY || result == D3DERR_INVALIDCALL || result == D3DERR_DEVICELOST); + + ERR("Could not create additional swap chains or offscreen surfaces: %08lX", result); + release(); + + if (d3d9::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + + result = mSwapChain->GetBackBuffer(0, D3DBACKBUFFER_TYPE_MONO, &mBackBuffer); + ASSERT(SUCCEEDED(result)); + InvalidateRect(mWindow, NULL, FALSE); + } + + if (mDepthBufferFormat != GL_NONE) + { + result = device->CreateDepthStencilSurface(backbufferWidth, backbufferHeight, + gl_d3d9::ConvertRenderbufferFormat(mDepthBufferFormat), + D3DMULTISAMPLE_NONE, 0, FALSE, &mDepthStencil, NULL); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY || result == D3DERR_INVALIDCALL); + + ERR("Could not create depthstencil surface for new swap chain: 0x%08X", result); + release(); + + if (d3d9::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + } + + mWidth = backbufferWidth; + mHeight = backbufferHeight; + mSwapInterval = swapInterval; + + return EGL_SUCCESS; +} + +// parameters should be validated/clamped by caller +EGLint SwapChain9::swapRect(EGLint x, EGLint y, EGLint width, EGLint height) +{ + if (!mSwapChain) + { + return EGL_SUCCESS; + } + + IDirect3DDevice9 *device = mRenderer->getDevice(); + + // Disable all pipeline operations + device->SetRenderState(D3DRS_ZENABLE, D3DZB_FALSE); + device->SetRenderState(D3DRS_FILLMODE, D3DFILL_SOLID); + device->SetRenderState(D3DRS_ALPHATESTENABLE, FALSE); + device->SetRenderState(D3DRS_ALPHABLENDENABLE, FALSE); + device->SetRenderState(D3DRS_CULLMODE, D3DCULL_NONE); + device->SetRenderState(D3DRS_STENCILENABLE, FALSE); + device->SetRenderState(D3DRS_CLIPPLANEENABLE, 0); + device->SetRenderState(D3DRS_COLORWRITEENABLE, D3DCOLORWRITEENABLE_ALPHA | D3DCOLORWRITEENABLE_BLUE | D3DCOLORWRITEENABLE_GREEN | D3DCOLORWRITEENABLE_RED); + device->SetRenderState(D3DRS_SRGBWRITEENABLE, FALSE); + device->SetRenderState(D3DRS_SCISSORTESTENABLE, FALSE); + device->SetPixelShader(NULL); + device->SetVertexShader(NULL); + + device->SetRenderTarget(0, mBackBuffer); + device->SetDepthStencilSurface(NULL); + + device->SetTexture(0, mOffscreenTexture); + device->SetTextureStageState(0, D3DTSS_COLOROP, D3DTOP_SELECTARG1); + device->SetTextureStageState(0, D3DTSS_COLORARG1, D3DTA_TEXTURE); + device->SetTextureStageState(1, D3DTSS_COLOROP, D3DTOP_DISABLE); + device->SetSamplerState(0, D3DSAMP_MAGFILTER, D3DTEXF_POINT); + device->SetSamplerState(0, D3DSAMP_MINFILTER, D3DTEXF_POINT); + device->SetSamplerState(0, D3DSAMP_ADDRESSU, D3DTADDRESS_CLAMP); + device->SetSamplerState(0, D3DSAMP_ADDRESSV, D3DTADDRESS_CLAMP); + device->SetFVF(D3DFVF_XYZRHW | D3DFVF_TEX1); + + D3DVIEWPORT9 viewport = {0, 0, mWidth, mHeight, 0.0f, 1.0f}; + device->SetViewport(&viewport); + + float x1 = x - 0.5f; + float y1 = (mHeight - y - height) - 0.5f; + float x2 = (x + width) - 0.5f; + float y2 = (mHeight - y) - 0.5f; + + float u1 = x / float(mWidth); + float v1 = y / float(mHeight); + float u2 = (x + width) / float(mWidth); + float v2 = (y + height) / float(mHeight); + + float quad[4][6] = {{x1, y1, 0.0f, 1.0f, u1, v2}, + {x2, y1, 0.0f, 1.0f, u2, v2}, + {x2, y2, 0.0f, 1.0f, u2, v1}, + {x1, y2, 0.0f, 1.0f, u1, v1}}; // x, y, z, rhw, u, v + + mRenderer->startScene(); + device->DrawPrimitiveUP(D3DPT_TRIANGLEFAN, 2, quad, 6 * sizeof(float)); + mRenderer->endScene(); + + device->SetTexture(0, NULL); + + RECT rect = + { + x, mHeight - y - height, + x + width, mHeight - y + }; + + HRESULT result = mSwapChain->Present(&rect, &rect, NULL, NULL, 0); + + mRenderer->markAllStateDirty(); + + if (d3d9::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + + if (result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY || result == D3DERR_DRIVERINTERNALERROR) + { + return EGL_BAD_ALLOC; + } + + ASSERT(SUCCEEDED(result)); + + return EGL_SUCCESS; +} + +// Increments refcount on surface. +// caller must Release() the returned surface +IDirect3DSurface9 *SwapChain9::getRenderTarget() +{ + if (mRenderTarget) + { + mRenderTarget->AddRef(); + } + + return mRenderTarget; +} + +// Increments refcount on surface. +// caller must Release() the returned surface +IDirect3DSurface9 *SwapChain9::getDepthStencil() +{ + if (mDepthStencil) + { + mDepthStencil->AddRef(); + } + + return mDepthStencil; +} + +// Increments refcount on texture. +// caller must Release() the returned texture +IDirect3DTexture9 *SwapChain9::getOffscreenTexture() +{ + if (mOffscreenTexture) + { + mOffscreenTexture->AddRef(); + } + + return mOffscreenTexture; +} + +SwapChain9 *SwapChain9::makeSwapChain9(SwapChain *swapChain) +{ + ASSERT(HAS_DYNAMIC_TYPE(rx::SwapChain9*, swapChain)); + return static_cast<rx::SwapChain9*>(swapChain); +} + +void SwapChain9::recreate() +{ + if (!mSwapChain) + { + return; + } + + IDirect3DDevice9 *device = mRenderer->getDevice(); + if (device == NULL) + { + return; + } + + D3DPRESENT_PARAMETERS presentParameters; + HRESULT result = mSwapChain->GetPresentParameters(&presentParameters); + ASSERT(SUCCEEDED(result)); + + IDirect3DSwapChain9* newSwapChain = NULL; + result = device->CreateAdditionalSwapChain(&presentParameters, &newSwapChain); + if (FAILED(result)) + { + return; + } + + mSwapChain->Release(); + mSwapChain = newSwapChain; + + mBackBuffer->Release(); + result = mSwapChain->GetBackBuffer(0, D3DBACKBUFFER_TYPE_MONO, &mBackBuffer); + ASSERT(SUCCEEDED(result)); +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain9.h b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain9.h new file mode 100644 index 0000000000..16a62bd86f --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/SwapChain9.h @@ -0,0 +1,55 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChain9.h: Defines a back-end specific class for the D3D9 swap chain. + +#ifndef LIBGLESV2_RENDERER_SWAPCHAIN9_H_ +#define LIBGLESV2_RENDERER_SWAPCHAIN9_H_ + +#include "common/angleutils.h" +#include "libGLESv2/renderer/SwapChain.h" + +namespace rx +{ +class Renderer9; + +class SwapChain9 : public SwapChain +{ + public: + SwapChain9(Renderer9 *renderer, HWND window, HANDLE shareHandle, + GLenum backBufferFormat, GLenum depthBufferFormat); + virtual ~SwapChain9(); + + EGLint resize(EGLint backbufferWidth, EGLint backbufferHeight); + virtual EGLint reset(EGLint backbufferWidth, EGLint backbufferHeight, EGLint swapInterval); + virtual EGLint swapRect(EGLint x, EGLint y, EGLint width, EGLint height); + virtual void recreate(); + + virtual IDirect3DSurface9 *getRenderTarget(); + virtual IDirect3DSurface9 *getDepthStencil(); + virtual IDirect3DTexture9 *getOffscreenTexture(); + + static SwapChain9 *makeSwapChain9(SwapChain *swapChain); + + private: + DISALLOW_COPY_AND_ASSIGN(SwapChain9); + + void release(); + + Renderer9 *mRenderer; + EGLint mHeight; + EGLint mWidth; + EGLint mSwapInterval; + + IDirect3DSwapChain9 *mSwapChain; + IDirect3DSurface9 *mBackBuffer; + IDirect3DSurface9 *mRenderTarget; + IDirect3DSurface9 *mDepthStencil; + IDirect3DTexture9* mOffscreenTexture; +}; + +} +#endif // LIBGLESV2_RENDERER_SWAPCHAIN9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage.cpp new file mode 100644 index 0000000000..00b316f1cc --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage.cpp @@ -0,0 +1,122 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// TextureStorage.cpp: Implements the abstract rx::TextureStorageInterface class and its concrete derived +// classes TextureStorageInterface2D and TextureStorageInterfaceCube, which act as the interface to the +// GPU-side texture. + +#include "libGLESv2/renderer/TextureStorage.h" +#include "libGLESv2/renderer/Renderer.h" +#include "libGLESv2/Renderbuffer.h" +#include "libGLESv2/Texture.h" + +#include "common/debug.h" + +namespace rx +{ +unsigned int TextureStorageInterface::mCurrentTextureSerial = 1; + +TextureStorageInterface::TextureStorageInterface() + : mTextureSerial(issueTextureSerial()), + mInstance(NULL) +{ +} + +TextureStorageInterface::~TextureStorageInterface() +{ + delete mInstance; +} + +bool TextureStorageInterface::isRenderTarget() const +{ + return mInstance->isRenderTarget(); +} + + +bool TextureStorageInterface::isManaged() const +{ + return mInstance->isManaged(); +} + +unsigned int TextureStorageInterface::getTextureSerial() const +{ + return mTextureSerial; +} + +unsigned int TextureStorageInterface::issueTextureSerial() +{ + return mCurrentTextureSerial++; +} + +int TextureStorageInterface::getLodOffset() const +{ + return mInstance->getLodOffset(); +} + + +int TextureStorageInterface::levelCount() +{ + return mInstance->levelCount(); +} + +TextureStorageInterface2D::TextureStorageInterface2D(Renderer *renderer, SwapChain *swapchain) + : mRenderTargetSerial(gl::RenderbufferStorage::issueSerial()) +{ + mInstance = renderer->createTextureStorage2D(swapchain); +} + +TextureStorageInterface2D::TextureStorageInterface2D(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height) + : mRenderTargetSerial(gl::RenderbufferStorage::issueSerial()) +{ + mInstance = renderer->createTextureStorage2D(levels, internalformat, usage, forceRenderable, width, height); +} + +TextureStorageInterface2D::~TextureStorageInterface2D() +{ +} + +RenderTarget *TextureStorageInterface2D::getRenderTarget() const +{ + return mInstance->getRenderTarget(); +} + +void TextureStorageInterface2D::generateMipmap(int level) +{ + mInstance->generateMipmap(level); +} + +unsigned int TextureStorageInterface2D::getRenderTargetSerial(GLenum target) const +{ + return mRenderTargetSerial; +} + +TextureStorageInterfaceCube::TextureStorageInterfaceCube(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size) + : mFirstRenderTargetSerial(gl::RenderbufferStorage::issueCubeSerials()) +{ + mInstance = renderer->createTextureStorageCube(levels, internalformat, usage, forceRenderable, size); +} + +TextureStorageInterfaceCube::~TextureStorageInterfaceCube() +{ +} + +RenderTarget *TextureStorageInterfaceCube::getRenderTarget(GLenum faceTarget) const +{ + return mInstance->getRenderTarget(faceTarget); +} + +void TextureStorageInterfaceCube::generateMipmap(int face, int level) +{ + mInstance->generateMipmap(face, level); +} + +unsigned int TextureStorageInterfaceCube::getRenderTargetSerial(GLenum target) const +{ + return mFirstRenderTargetSerial + gl::TextureCubeMap::faceIndex(target); +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage.h b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage.h new file mode 100644 index 0000000000..edddb75f3f --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage.h @@ -0,0 +1,110 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// TextureStorage.h: Defines the abstract rx::TextureStorageInterface class and its concrete derived +// classes TextureStorageInterface2D and TextureStorageInterfaceCube, which act as the interface to the +// GPU-side texture. + +#ifndef LIBGLESV2_RENDERER_TEXTURESTORAGE_H_ +#define LIBGLESV2_RENDERER_TEXTURESTORAGE_H_ + +#include "common/debug.h" + +namespace rx +{ +class Renderer; +class SwapChain; +class RenderTarget; +class Blit; + +class TextureStorage +{ + public: + TextureStorage() {}; + virtual ~TextureStorage() {}; + + virtual int getLodOffset() const = 0; + virtual bool isRenderTarget() const = 0; + virtual bool isManaged() const = 0; + virtual int levelCount() = 0; + + virtual RenderTarget *getRenderTarget() = 0; + virtual RenderTarget *getRenderTarget(GLenum faceTarget) = 0; + virtual void generateMipmap(int level) = 0; + virtual void generateMipmap(int face, int level) = 0; + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorage); + +}; + +class TextureStorageInterface +{ + public: + TextureStorageInterface(); + virtual ~TextureStorageInterface(); + + TextureStorage *getStorageInstance() { return mInstance; } + + unsigned int getTextureSerial() const; + virtual unsigned int getRenderTargetSerial(GLenum target) const = 0; + + virtual int getLodOffset() const; + virtual bool isRenderTarget() const; + virtual bool isManaged() const; + virtual int levelCount(); + + protected: + TextureStorage *mInstance; + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorageInterface); + + const unsigned int mTextureSerial; + static unsigned int issueTextureSerial(); + + static unsigned int mCurrentTextureSerial; +}; + +class TextureStorageInterface2D : public TextureStorageInterface +{ + public: + TextureStorageInterface2D(Renderer *renderer, SwapChain *swapchain); + TextureStorageInterface2D(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height); + virtual ~TextureStorageInterface2D(); + + void generateMipmap(int level); + RenderTarget *getRenderTarget() const; + + virtual unsigned int getRenderTargetSerial(GLenum target) const; + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorageInterface2D); + + const unsigned int mRenderTargetSerial; +}; + +class TextureStorageInterfaceCube : public TextureStorageInterface +{ + public: + TextureStorageInterfaceCube(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size); + virtual ~TextureStorageInterfaceCube(); + + void generateMipmap(int face, int level); + RenderTarget *getRenderTarget(GLenum faceTarget) const; + + virtual unsigned int getRenderTargetSerial(GLenum target) const; + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorageInterfaceCube); + + const unsigned int mFirstRenderTargetSerial; +}; + +} + +#endif // LIBGLESV2_RENDERER_TEXTURESTORAGE_H_ + diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage11.cpp new file mode 100644 index 0000000000..667dbff175 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage11.cpp @@ -0,0 +1,676 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// TextureStorage11.cpp: Implements the abstract rx::TextureStorage11 class and its concrete derived +// classes TextureStorage11_2D and TextureStorage11_Cube, which act as the interface to the D3D11 texture. + +#include "libGLESv2/renderer/TextureStorage11.h" + +#include "libGLESv2/renderer/Renderer11.h" +#include "libGLESv2/renderer/RenderTarget11.h" +#include "libGLESv2/renderer/SwapChain11.h" +#include "libGLESv2/renderer/renderer11_utils.h" + +#include "libGLESv2/utilities.h" +#include "libGLESv2/main.h" + +namespace rx +{ + +TextureStorage11::TextureStorage11(Renderer *renderer, UINT bindFlags) + : mBindFlags(bindFlags), + mLodOffset(0), + mMipLevels(0), + mTexture(NULL), + mTextureFormat(DXGI_FORMAT_UNKNOWN), + mShaderResourceFormat(DXGI_FORMAT_UNKNOWN), + mRenderTargetFormat(DXGI_FORMAT_UNKNOWN), + mDepthStencilFormat(DXGI_FORMAT_UNKNOWN), + mSRV(NULL), + mTextureWidth(0), + mTextureHeight(0) +{ + mRenderer = Renderer11::makeRenderer11(renderer); +} + +TextureStorage11::~TextureStorage11() +{ +} + +TextureStorage11 *TextureStorage11::makeTextureStorage11(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage11*, storage)); + return static_cast<TextureStorage11*>(storage); +} + +DWORD TextureStorage11::GetTextureBindFlags(DXGI_FORMAT format, GLenum glusage, bool forceRenderable) +{ + UINT bindFlags = D3D11_BIND_SHADER_RESOURCE; + + if (d3d11::IsDepthStencilFormat(format)) + { + bindFlags |= D3D11_BIND_DEPTH_STENCIL; + } + else if(forceRenderable || (TextureStorage11::IsTextureFormatRenderable(format) && (glusage == GL_FRAMEBUFFER_ATTACHMENT_ANGLE))) + { + bindFlags |= D3D11_BIND_RENDER_TARGET; + } + return bindFlags; +} + +bool TextureStorage11::IsTextureFormatRenderable(DXGI_FORMAT format) +{ + switch(format) + { + case DXGI_FORMAT_R8G8B8A8_UNORM: + case DXGI_FORMAT_A8_UNORM: + case DXGI_FORMAT_R32G32B32A32_FLOAT: + case DXGI_FORMAT_R32G32B32_FLOAT: + case DXGI_FORMAT_R16G16B16A16_FLOAT: + case DXGI_FORMAT_B8G8R8A8_UNORM: + case DXGI_FORMAT_R8_UNORM: + case DXGI_FORMAT_R8G8_UNORM: + case DXGI_FORMAT_R16_FLOAT: + case DXGI_FORMAT_R16G16_FLOAT: + return true; + case DXGI_FORMAT_BC1_UNORM: + case DXGI_FORMAT_BC2_UNORM: + case DXGI_FORMAT_BC3_UNORM: + return false; + default: + UNREACHABLE(); + return false; + } +} + +UINT TextureStorage11::getBindFlags() const +{ + return mBindFlags; +} + +ID3D11Texture2D *TextureStorage11::getBaseTexture() const +{ + return mTexture; +} + +int TextureStorage11::getLodOffset() const +{ + return mLodOffset; +} + +bool TextureStorage11::isRenderTarget() const +{ + return (mBindFlags & (D3D11_BIND_RENDER_TARGET | D3D11_BIND_DEPTH_STENCIL)) != 0; +} + +bool TextureStorage11::isManaged() const +{ + return false; +} + +int TextureStorage11::levelCount() +{ + int levels = 0; + if (getBaseTexture()) + { + levels = mMipLevels - getLodOffset(); + } + return levels; +} + +UINT TextureStorage11::getSubresourceIndex(int level, int faceIndex) +{ + UINT index = 0; + if (getBaseTexture()) + { + index = D3D11CalcSubresource(level, faceIndex, mMipLevels); + } + return index; +} + +bool TextureStorage11::updateSubresourceLevel(ID3D11Texture2D *srcTexture, unsigned int sourceSubresource, + int level, int face, GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height) +{ + if (srcTexture) + { + // Round up the width and height to the nearest multiple of dimension alignment + unsigned int dimensionAlignment = d3d11::GetTextureFormatDimensionAlignment(mTextureFormat); + width = width + dimensionAlignment - 1 - (width - 1) % dimensionAlignment; + height = height + dimensionAlignment - 1 - (height - 1) % dimensionAlignment; + + D3D11_BOX srcBox; + srcBox.left = xoffset; + srcBox.top = yoffset; + srcBox.right = xoffset + width; + srcBox.bottom = yoffset + height; + srcBox.front = 0; + srcBox.back = 1; + + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + + ASSERT(getBaseTexture()); + context->CopySubresourceRegion(getBaseTexture(), getSubresourceIndex(level + mLodOffset, face), + xoffset, yoffset, 0, srcTexture, sourceSubresource, &srcBox); + return true; + } + + return false; +} + +void TextureStorage11::generateMipmapLayer(RenderTarget11 *source, RenderTarget11 *dest) +{ + if (source && dest) + { + ID3D11ShaderResourceView *sourceSRV = source->getShaderResourceView(); + ID3D11RenderTargetView *destRTV = dest->getRenderTargetView(); + + if (sourceSRV && destRTV) + { + gl::Rectangle sourceArea; + sourceArea.x = 0; + sourceArea.y = 0; + sourceArea.width = source->getWidth(); + sourceArea.height = source->getHeight(); + + gl::Rectangle destArea; + destArea.x = 0; + destArea.y = 0; + destArea.width = dest->getWidth(); + destArea.height = dest->getHeight(); + + mRenderer->copyTexture(sourceSRV, sourceArea, source->getWidth(), source->getHeight(), + destRTV, destArea, dest->getWidth(), dest->getHeight(), + GL_RGBA); + } + + if (sourceSRV) + { + sourceSRV->Release(); + } + if (destRTV) + { + destRTV->Release(); + } + } +} + +TextureStorage11_2D::TextureStorage11_2D(Renderer *renderer, SwapChain11 *swapchain) + : TextureStorage11(renderer, D3D11_BIND_RENDER_TARGET | D3D11_BIND_SHADER_RESOURCE) +{ + mTexture = swapchain->getOffscreenTexture(); + mSRV = swapchain->getRenderTargetShaderResource(); + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + mRenderTarget[i] = NULL; + } + + D3D11_TEXTURE2D_DESC texDesc; + mTexture->GetDesc(&texDesc); + mMipLevels = texDesc.MipLevels; + mTextureFormat = texDesc.Format; + mTextureWidth = texDesc.Width; + mTextureHeight = texDesc.Height; + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + mSRV->GetDesc(&srvDesc); + mShaderResourceFormat = srvDesc.Format; + + ID3D11RenderTargetView* offscreenRTV = swapchain->getRenderTarget(); + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + offscreenRTV->GetDesc(&rtvDesc); + mRenderTargetFormat = rtvDesc.Format; + offscreenRTV->Release(); + + mDepthStencilFormat = DXGI_FORMAT_UNKNOWN; +} + +TextureStorage11_2D::TextureStorage11_2D(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height) + : TextureStorage11(renderer, GetTextureBindFlags(gl_d3d11::ConvertTextureFormat(internalformat), usage, forceRenderable)) +{ + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + mRenderTarget[i] = NULL; + } + + DXGI_FORMAT convertedFormat = gl_d3d11::ConvertTextureFormat(internalformat); + if (d3d11::IsDepthStencilFormat(convertedFormat)) + { + mTextureFormat = d3d11::GetDepthTextureFormat(convertedFormat); + mShaderResourceFormat = d3d11::GetDepthShaderResourceFormat(convertedFormat); + mDepthStencilFormat = convertedFormat; + mRenderTargetFormat = DXGI_FORMAT_UNKNOWN; + } + else + { + mTextureFormat = convertedFormat; + mShaderResourceFormat = convertedFormat; + mDepthStencilFormat = DXGI_FORMAT_UNKNOWN; + mRenderTargetFormat = convertedFormat; + } + + // if the width or height is not positive this should be treated as an incomplete texture + // we handle that here by skipping the d3d texture creation + if (width > 0 && height > 0) + { + // adjust size if needed for compressed textures + gl::MakeValidSize(false, gl::IsCompressed(internalformat), &width, &height, &mLodOffset); + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = width; // Compressed texture size constraints? + desc.Height = height; + desc.MipLevels = (levels > 0) ? levels + mLodOffset : 0; + desc.ArraySize = 1; + desc.Format = mTextureFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = getBindFlags(); + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + + HRESULT result = device->CreateTexture2D(&desc, NULL, &mTexture); + + // this can happen from windows TDR + if (d3d11::isDeviceLostError(result)) + { + mRenderer->notifyDeviceLost(); + gl::error(GL_OUT_OF_MEMORY); + } + else if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + ERR("Creating image failed."); + gl::error(GL_OUT_OF_MEMORY); + } + else + { + mTexture->GetDesc(&desc); + mMipLevels = desc.MipLevels; + mTextureWidth = desc.Width; + mTextureHeight = desc.Height; + } + } +} + +TextureStorage11_2D::~TextureStorage11_2D() +{ + if (mTexture) + { + mTexture->Release(); + mTexture = NULL; + } + + if (mSRV) + { + mSRV->Release(); + mSRV = NULL; + } + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + delete mRenderTarget[i]; + mRenderTarget[i] = NULL; + } +} + +TextureStorage11_2D *TextureStorage11_2D::makeTextureStorage11_2D(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage11_2D*, storage)); + return static_cast<TextureStorage11_2D*>(storage); +} + +RenderTarget *TextureStorage11_2D::getRenderTarget(int level) +{ + if (level >= 0 && level < static_cast<int>(mMipLevels)) + { + if (!mRenderTarget[level]) + { + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result; + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = mShaderResourceFormat; + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D; + srvDesc.Texture2D.MostDetailedMip = level; + srvDesc.Texture2D.MipLevels = 1; + + ID3D11ShaderResourceView *srv; + result = device->CreateShaderResourceView(mTexture, &srvDesc, &srv); + + if (result == E_OUTOFMEMORY) + { + return gl::error(GL_OUT_OF_MEMORY, static_cast<RenderTarget*>(NULL)); + } + ASSERT(SUCCEEDED(result)); + + if (mRenderTargetFormat != DXGI_FORMAT_UNKNOWN) + { + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D; + rtvDesc.Texture2D.MipSlice = level; + + ID3D11RenderTargetView *rtv; + result = device->CreateRenderTargetView(mTexture, &rtvDesc, &rtv); + + if (result == E_OUTOFMEMORY) + { + srv->Release(); + return gl::error(GL_OUT_OF_MEMORY, static_cast<RenderTarget*>(NULL)); + } + ASSERT(SUCCEEDED(result)); + + // RenderTarget11 expects to be the owner of the resources it is given but TextureStorage11 + // also needs to keep a reference to the texture. + mTexture->AddRef(); + + mRenderTarget[level] = new RenderTarget11(mRenderer, rtv, mTexture, srv, + std::max(mTextureWidth >> level, 1U), + std::max(mTextureHeight >> level, 1U)); + } + else if (mDepthStencilFormat != DXGI_FORMAT_UNKNOWN) + { + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; + dsvDesc.Format = mDepthStencilFormat; + dsvDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2D; + dsvDesc.Texture2D.MipSlice = level; + dsvDesc.Flags = 0; + + ID3D11DepthStencilView *dsv; + result = device->CreateDepthStencilView(mTexture, &dsvDesc, &dsv); + + if (result == E_OUTOFMEMORY) + { + srv->Release(); + return gl::error(GL_OUT_OF_MEMORY, static_cast<RenderTarget*>(NULL)); + } + ASSERT(SUCCEEDED(result)); + + // RenderTarget11 expects to be the owner of the resources it is given but TextureStorage11 + // also needs to keep a reference to the texture. + mTexture->AddRef(); + + mRenderTarget[level] = new RenderTarget11(mRenderer, dsv, mTexture, srv, + std::max(mTextureWidth >> level, 1U), + std::max(mTextureHeight >> level, 1U)); + } + else + { + UNREACHABLE(); + } + } + + return mRenderTarget[level]; + } + else + { + return NULL; + } +} + +ID3D11ShaderResourceView *TextureStorage11_2D::getSRV() +{ + if (!mSRV) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = mShaderResourceFormat; + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D; + srvDesc.Texture2D.MipLevels = (mMipLevels == 0 ? -1 : mMipLevels); + srvDesc.Texture2D.MostDetailedMip = 0; + + HRESULT result = device->CreateShaderResourceView(mTexture, &srvDesc, &mSRV); + + if (result == E_OUTOFMEMORY) + { + return gl::error(GL_OUT_OF_MEMORY, static_cast<ID3D11ShaderResourceView*>(NULL)); + } + ASSERT(SUCCEEDED(result)); + } + + return mSRV; +} + +void TextureStorage11_2D::generateMipmap(int level) +{ + RenderTarget11 *source = RenderTarget11::makeRenderTarget11(getRenderTarget(level - 1)); + RenderTarget11 *dest = RenderTarget11::makeRenderTarget11(getRenderTarget(level)); + + generateMipmapLayer(source, dest); +} + +TextureStorage11_Cube::TextureStorage11_Cube(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size) + : TextureStorage11(renderer, GetTextureBindFlags(gl_d3d11::ConvertTextureFormat(internalformat), usage, forceRenderable)) +{ + for (unsigned int i = 0; i < 6; i++) + { + for (unsigned int j = 0; j < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; j++) + { + mRenderTarget[i][j] = NULL; + } + } + + DXGI_FORMAT convertedFormat = gl_d3d11::ConvertTextureFormat(internalformat); + if (d3d11::IsDepthStencilFormat(convertedFormat)) + { + mTextureFormat = d3d11::GetDepthTextureFormat(convertedFormat); + mShaderResourceFormat = d3d11::GetDepthShaderResourceFormat(convertedFormat); + mDepthStencilFormat = convertedFormat; + mRenderTargetFormat = DXGI_FORMAT_UNKNOWN; + } + else + { + mTextureFormat = convertedFormat; + mShaderResourceFormat = convertedFormat; + mDepthStencilFormat = DXGI_FORMAT_UNKNOWN; + mRenderTargetFormat = convertedFormat; + } + + // if the size is not positive this should be treated as an incomplete texture + // we handle that here by skipping the d3d texture creation + if (size > 0) + { + // adjust size if needed for compressed textures + int height = size; + gl::MakeValidSize(false, gl::IsCompressed(internalformat), &size, &height, &mLodOffset); + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = size; + desc.Height = size; + desc.MipLevels = (levels > 0) ? levels + mLodOffset : 0; + desc.ArraySize = 6; + desc.Format = mTextureFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = getBindFlags(); + desc.CPUAccessFlags = 0; + desc.MiscFlags = D3D11_RESOURCE_MISC_TEXTURECUBE; + + HRESULT result = device->CreateTexture2D(&desc, NULL, &mTexture); + + if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + ERR("Creating image failed."); + gl::error(GL_OUT_OF_MEMORY); + } + else + { + mTexture->GetDesc(&desc); + mMipLevels = desc.MipLevels; + mTextureWidth = desc.Width; + mTextureHeight = desc.Height; + } + } +} + +TextureStorage11_Cube::~TextureStorage11_Cube() +{ + if (mTexture) + { + mTexture->Release(); + mTexture = NULL; + } + + if (mSRV) + { + mSRV->Release(); + mSRV = NULL; + } + + for (unsigned int i = 0; i < 6; i++) + { + for (unsigned int j = 0; j < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; j++) + { + delete mRenderTarget[i][j]; + mRenderTarget[i][j] = NULL; + } + } +} + +TextureStorage11_Cube *TextureStorage11_Cube::makeTextureStorage11_Cube(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage11_Cube*, storage)); + return static_cast<TextureStorage11_Cube*>(storage); +} + +RenderTarget *TextureStorage11_Cube::getRenderTarget(GLenum faceTarget, int level) +{ + unsigned int faceIdx = gl::TextureCubeMap::faceIndex(faceTarget); + if (level >= 0 && level < static_cast<int>(mMipLevels)) + { + if (!mRenderTarget[faceIdx][level]) + { + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result; + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = mShaderResourceFormat; + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURECUBE; + srvDesc.Texture2DArray.MostDetailedMip = level; + srvDesc.Texture2DArray.MipLevels = 1; + srvDesc.Texture2DArray.FirstArraySlice = faceIdx; + srvDesc.Texture2DArray.ArraySize = 1; + + ID3D11ShaderResourceView *srv; + result = device->CreateShaderResourceView(mTexture, &srvDesc, &srv); + + if (result == E_OUTOFMEMORY) + { + return gl::error(GL_OUT_OF_MEMORY, static_cast<RenderTarget*>(NULL)); + } + ASSERT(SUCCEEDED(result)); + + if (mRenderTargetFormat != DXGI_FORMAT_UNKNOWN) + { + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = level; + rtvDesc.Texture2DArray.FirstArraySlice = faceIdx; + rtvDesc.Texture2DArray.ArraySize = 1; + + ID3D11RenderTargetView *rtv; + result = device->CreateRenderTargetView(mTexture, &rtvDesc, &rtv); + + if (result == E_OUTOFMEMORY) + { + srv->Release(); + return gl::error(GL_OUT_OF_MEMORY, static_cast<RenderTarget*>(NULL)); + } + ASSERT(SUCCEEDED(result)); + + // RenderTarget11 expects to be the owner of the resources it is given but TextureStorage11 + // also needs to keep a reference to the texture. + mTexture->AddRef(); + + mRenderTarget[faceIdx][level] = new RenderTarget11(mRenderer, rtv, mTexture, srv, + std::max(mTextureWidth >> level, 1U), + std::max(mTextureHeight >> level, 1U)); + } + else if (mDepthStencilFormat != DXGI_FORMAT_UNKNOWN) + { + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; + dsvDesc.Format = mRenderTargetFormat; + dsvDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DARRAY; + dsvDesc.Texture2DArray.MipSlice = level; + dsvDesc.Texture2DArray.FirstArraySlice = faceIdx; + dsvDesc.Texture2DArray.ArraySize = 1; + + ID3D11DepthStencilView *dsv; + result = device->CreateDepthStencilView(mTexture, &dsvDesc, &dsv); + + if (result == E_OUTOFMEMORY) + { + srv->Release(); + return gl::error(GL_OUT_OF_MEMORY, static_cast<RenderTarget*>(NULL)); + } + ASSERT(SUCCEEDED(result)); + + // RenderTarget11 expects to be the owner of the resources it is given but TextureStorage11 + // also needs to keep a reference to the texture. + mTexture->AddRef(); + + mRenderTarget[faceIdx][level] = new RenderTarget11(mRenderer, dsv, mTexture, srv, + std::max(mTextureWidth >> level, 1U), + std::max(mTextureHeight >> level, 1U)); + } + else + { + UNREACHABLE(); + } + } + + return mRenderTarget[faceIdx][level]; + } + else + { + return NULL; + } +} + +ID3D11ShaderResourceView *TextureStorage11_Cube::getSRV() +{ + if (!mSRV) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = mShaderResourceFormat; + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURECUBE; + srvDesc.TextureCube.MipLevels = (mMipLevels == 0 ? -1 : mMipLevels); + srvDesc.TextureCube.MostDetailedMip = 0; + + HRESULT result = device->CreateShaderResourceView(mTexture, &srvDesc, &mSRV); + + if (result == E_OUTOFMEMORY) + { + return gl::error(GL_OUT_OF_MEMORY, static_cast<ID3D11ShaderResourceView*>(NULL)); + } + ASSERT(SUCCEEDED(result)); + } + + return mSRV; +} + +void TextureStorage11_Cube::generateMipmap(int face, int level) +{ + RenderTarget11 *source = RenderTarget11::makeRenderTarget11(getRenderTarget(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face, level - 1)); + RenderTarget11 *dest = RenderTarget11::makeRenderTarget11(getRenderTarget(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face, level)); + + generateMipmapLayer(source, dest); +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage11.h b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage11.h new file mode 100644 index 0000000000..3c5ded05b8 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage11.h @@ -0,0 +1,120 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// TextureStorage11.h: Defines the abstract rx::TextureStorage11 class and its concrete derived +// classes TextureStorage11_2D and TextureStorage11_Cube, which act as the interface to the D3D11 texture. + +#ifndef LIBGLESV2_RENDERER_TEXTURESTORAGE11_H_ +#define LIBGLESV2_RENDERER_TEXTURESTORAGE11_H_ + +#include "libGLESv2/Texture.h" +#include "libGLESv2/renderer/TextureStorage.h" + +namespace rx +{ +class RenderTarget; +class RenderTarget11; +class Renderer; +class Renderer11; +class SwapChain11; + +class TextureStorage11 : public TextureStorage +{ + public: + TextureStorage11(Renderer *renderer, UINT bindFlags); + virtual ~TextureStorage11(); + + static TextureStorage11 *makeTextureStorage11(TextureStorage *storage); + + static DWORD GetTextureBindFlags(DXGI_FORMAT d3dfmt, GLenum glusage, bool forceRenderable); + static bool IsTextureFormatRenderable(DXGI_FORMAT format); + + UINT getBindFlags() const; + + virtual ID3D11Texture2D *getBaseTexture() const; + virtual ID3D11ShaderResourceView *getSRV() = 0; + virtual RenderTarget *getRenderTarget() { return getRenderTarget(0); } + virtual RenderTarget *getRenderTarget(int level) { return NULL; } + virtual RenderTarget *getRenderTarget(GLenum faceTarget) { return getRenderTarget(faceTarget, 0); } + virtual RenderTarget *getRenderTarget(GLenum faceTarget, int level) { return NULL; } + + virtual void generateMipmap(int level) {}; + virtual void generateMipmap(int face, int level) {}; + + virtual int getLodOffset() const; + virtual bool isRenderTarget() const; + virtual bool isManaged() const; + virtual int levelCount(); + UINT getSubresourceIndex(int level, int faceTarget); + + bool updateSubresourceLevel(ID3D11Texture2D *texture, unsigned int sourceSubresource, int level, + int faceTarget, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height); + + protected: + void generateMipmapLayer(RenderTarget11 *source, RenderTarget11 *dest); + + Renderer11 *mRenderer; + int mLodOffset; + unsigned int mMipLevels; + + ID3D11Texture2D *mTexture; + DXGI_FORMAT mTextureFormat; + DXGI_FORMAT mShaderResourceFormat; + DXGI_FORMAT mRenderTargetFormat; + DXGI_FORMAT mDepthStencilFormat; + unsigned int mTextureWidth; + unsigned int mTextureHeight; + + ID3D11ShaderResourceView *mSRV; + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorage11); + + const UINT mBindFlags; +}; + +class TextureStorage11_2D : public TextureStorage11 +{ + public: + TextureStorage11_2D(Renderer *renderer, SwapChain11 *swapchain); + TextureStorage11_2D(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height); + virtual ~TextureStorage11_2D(); + + static TextureStorage11_2D *makeTextureStorage11_2D(TextureStorage *storage); + + virtual ID3D11ShaderResourceView *getSRV(); + virtual RenderTarget *getRenderTarget(int level); + + virtual void generateMipmap(int level); + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorage11_2D); + + RenderTarget11 *mRenderTarget[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; +}; + +class TextureStorage11_Cube : public TextureStorage11 +{ + public: + TextureStorage11_Cube(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size); + virtual ~TextureStorage11_Cube(); + + static TextureStorage11_Cube *makeTextureStorage11_Cube(TextureStorage *storage); + + virtual ID3D11ShaderResourceView *getSRV(); + virtual RenderTarget *getRenderTarget(GLenum faceTarget, int level); + + virtual void generateMipmap(int face, int level); + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorage11_Cube); + + RenderTarget11 *mRenderTarget[6][gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; +}; + +} + +#endif // LIBGLESV2_RENDERER_TEXTURESTORAGE11_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage9.cpp new file mode 100644 index 0000000000..8aa74a7cba --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage9.cpp @@ -0,0 +1,328 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// TextureStorage9.cpp: Implements the abstract rx::TextureStorage9 class and its concrete derived +// classes TextureStorage9_2D and TextureStorage9_Cube, which act as the interface to the +// D3D9 texture. + +#include "libGLESv2/main.h" +#include "libGLESv2/renderer/Renderer9.h" +#include "libGLESv2/renderer/TextureStorage9.h" +#include "libGLESv2/renderer/SwapChain9.h" +#include "libGLESv2/renderer/RenderTarget9.h" +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/Texture.h" + +namespace rx +{ +TextureStorage9::TextureStorage9(Renderer *renderer, DWORD usage) + : mLodOffset(0), + mRenderer(Renderer9::makeRenderer9(renderer)), + mD3DUsage(usage), + mD3DPool(mRenderer->getTexturePool(usage)) +{ +} + +TextureStorage9::~TextureStorage9() +{ +} + +TextureStorage9 *TextureStorage9::makeTextureStorage9(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage9*, storage)); + return static_cast<TextureStorage9*>(storage); +} + +DWORD TextureStorage9::GetTextureUsage(D3DFORMAT d3dfmt, GLenum glusage, bool forceRenderable) +{ + DWORD d3dusage = 0; + + if (d3dfmt == D3DFMT_INTZ) + { + d3dusage |= D3DUSAGE_DEPTHSTENCIL; + } + else if(forceRenderable || (TextureStorage9::IsTextureFormatRenderable(d3dfmt) && (glusage == GL_FRAMEBUFFER_ATTACHMENT_ANGLE))) + { + d3dusage |= D3DUSAGE_RENDERTARGET; + } + return d3dusage; +} + +bool TextureStorage9::IsTextureFormatRenderable(D3DFORMAT format) +{ + if (format == D3DFMT_INTZ) + { + return true; + } + switch(format) + { + case D3DFMT_L8: + case D3DFMT_A8L8: + case D3DFMT_DXT1: + case D3DFMT_DXT3: + case D3DFMT_DXT5: + return false; + case D3DFMT_A8R8G8B8: + case D3DFMT_X8R8G8B8: + case D3DFMT_A16B16G16R16F: + case D3DFMT_A32B32G32R32F: + return true; + default: + UNREACHABLE(); + } + + return false; +} + +bool TextureStorage9::isRenderTarget() const +{ + return (mD3DUsage & (D3DUSAGE_RENDERTARGET | D3DUSAGE_DEPTHSTENCIL)) != 0; +} + +bool TextureStorage9::isManaged() const +{ + return (mD3DPool == D3DPOOL_MANAGED); +} + +D3DPOOL TextureStorage9::getPool() const +{ + return mD3DPool; +} + +DWORD TextureStorage9::getUsage() const +{ + return mD3DUsage; +} + +int TextureStorage9::getLodOffset() const +{ + return mLodOffset; +} + +int TextureStorage9::levelCount() +{ + return getBaseTexture() ? getBaseTexture()->GetLevelCount() - getLodOffset() : 0; +} + +TextureStorage9_2D::TextureStorage9_2D(Renderer *renderer, SwapChain9 *swapchain) : TextureStorage9(renderer, D3DUSAGE_RENDERTARGET) +{ + IDirect3DTexture9 *surfaceTexture = swapchain->getOffscreenTexture(); + mTexture = surfaceTexture; + mRenderTarget = NULL; + + initializeRenderTarget(); +} + +TextureStorage9_2D::TextureStorage9_2D(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height) + : TextureStorage9(renderer, GetTextureUsage(Renderer9::makeRenderer9(renderer)->ConvertTextureInternalFormat(internalformat), usage, forceRenderable)) +{ + mTexture = NULL; + mRenderTarget = NULL; + // if the width or height is not positive this should be treated as an incomplete texture + // we handle that here by skipping the d3d texture creation + if (width > 0 && height > 0) + { + IDirect3DDevice9 *device = mRenderer->getDevice(); + gl::MakeValidSize(false, gl::IsCompressed(internalformat), &width, &height, &mLodOffset); + HRESULT result = device->CreateTexture(width, height, levels ? levels + mLodOffset : 0, getUsage(), + mRenderer->ConvertTextureInternalFormat(internalformat), getPool(), &mTexture, NULL); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + gl::error(GL_OUT_OF_MEMORY); + } + } + + initializeRenderTarget(); +} + +TextureStorage9_2D::~TextureStorage9_2D() +{ + if (mTexture) + { + mTexture->Release(); + } + + delete mRenderTarget; +} + +TextureStorage9_2D *TextureStorage9_2D::makeTextureStorage9_2D(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage9_2D*, storage)); + return static_cast<TextureStorage9_2D*>(storage); +} + +// Increments refcount on surface. +// caller must Release() the returned surface +IDirect3DSurface9 *TextureStorage9_2D::getSurfaceLevel(int level, bool dirty) +{ + IDirect3DSurface9 *surface = NULL; + + if (mTexture) + { + HRESULT result = mTexture->GetSurfaceLevel(level + mLodOffset, &surface); + ASSERT(SUCCEEDED(result)); + + // With managed textures the driver needs to be informed of updates to the lower mipmap levels + if (level + mLodOffset != 0 && isManaged() && dirty) + { + mTexture->AddDirtyRect(NULL); + } + } + + return surface; +} + +RenderTarget *TextureStorage9_2D::getRenderTarget() +{ + return mRenderTarget; +} + +void TextureStorage9_2D::generateMipmap(int level) +{ + IDirect3DSurface9 *upper = getSurfaceLevel(level - 1, false); + IDirect3DSurface9 *lower = getSurfaceLevel(level, true); + + if (upper != NULL && lower != NULL) + { + mRenderer->boxFilter(upper, lower); + } + + if (upper != NULL) upper->Release(); + if (lower != NULL) lower->Release(); +} + +IDirect3DBaseTexture9 *TextureStorage9_2D::getBaseTexture() const +{ + return mTexture; +} + +void TextureStorage9_2D::initializeRenderTarget() +{ + ASSERT(mRenderTarget == NULL); + + if (mTexture != NULL && isRenderTarget()) + { + IDirect3DSurface9 *surface = getSurfaceLevel(0, false); + + mRenderTarget = new RenderTarget9(mRenderer, surface); + } +} + +TextureStorage9_Cube::TextureStorage9_Cube(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size) + : TextureStorage9(renderer, GetTextureUsage(Renderer9::makeRenderer9(renderer)->ConvertTextureInternalFormat(internalformat), usage, forceRenderable)) +{ + mTexture = NULL; + for (int i = 0; i < 6; ++i) + { + mRenderTarget[i] = NULL; + } + + // if the size is not positive this should be treated as an incomplete texture + // we handle that here by skipping the d3d texture creation + if (size > 0) + { + IDirect3DDevice9 *device = mRenderer->getDevice(); + int height = size; + gl::MakeValidSize(false, gl::IsCompressed(internalformat), &size, &height, &mLodOffset); + HRESULT result = device->CreateCubeTexture(size, levels ? levels + mLodOffset : 0, getUsage(), + mRenderer->ConvertTextureInternalFormat(internalformat), getPool(), &mTexture, NULL); + + if (FAILED(result)) + { + ASSERT(result == D3DERR_OUTOFVIDEOMEMORY || result == E_OUTOFMEMORY); + gl::error(GL_OUT_OF_MEMORY); + } + } + + initializeRenderTarget(); +} + +TextureStorage9_Cube::~TextureStorage9_Cube() +{ + if (mTexture) + { + mTexture->Release(); + } + + for (int i = 0; i < 6; ++i) + { + delete mRenderTarget[i]; + } +} + +TextureStorage9_Cube *TextureStorage9_Cube::makeTextureStorage9_Cube(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage9_Cube*, storage)); + return static_cast<TextureStorage9_Cube*>(storage); +} + +// Increments refcount on surface. +// caller must Release() the returned surface +IDirect3DSurface9 *TextureStorage9_Cube::getCubeMapSurface(GLenum faceTarget, int level, bool dirty) +{ + IDirect3DSurface9 *surface = NULL; + + if (mTexture) + { + D3DCUBEMAP_FACES face = gl_d3d9::ConvertCubeFace(faceTarget); + HRESULT result = mTexture->GetCubeMapSurface(face, level + mLodOffset, &surface); + ASSERT(SUCCEEDED(result)); + + // With managed textures the driver needs to be informed of updates to the lower mipmap levels + if (level != 0 && isManaged() && dirty) + { + mTexture->AddDirtyRect(face, NULL); + } + } + + return surface; +} + +RenderTarget *TextureStorage9_Cube::getRenderTarget(GLenum faceTarget) +{ + return mRenderTarget[gl::TextureCubeMap::faceIndex(faceTarget)]; +} + +void TextureStorage9_Cube::generateMipmap(int face, int level) +{ + IDirect3DSurface9 *upper = getCubeMapSurface(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face, level - 1, false); + IDirect3DSurface9 *lower = getCubeMapSurface(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face, level, true); + + if (upper != NULL && lower != NULL) + { + mRenderer->boxFilter(upper, lower); + } + + if (upper != NULL) upper->Release(); + if (lower != NULL) lower->Release(); +} + +IDirect3DBaseTexture9 *TextureStorage9_Cube::getBaseTexture() const +{ + return mTexture; +} + +void TextureStorage9_Cube::initializeRenderTarget() +{ + if (mTexture != NULL && isRenderTarget()) + { + IDirect3DSurface9 *surface = NULL; + + for (int i = 0; i < 6; ++i) + { + ASSERT(mRenderTarget[i] == NULL); + + surface = getCubeMapSurface(GL_TEXTURE_CUBE_MAP_POSITIVE_X + i, 0, false); + + mRenderTarget[i] = new RenderTarget9(mRenderer, surface); + } + } +} + +}
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage9.h b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage9.h new file mode 100644 index 0000000000..86f551a131 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/TextureStorage9.h @@ -0,0 +1,109 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// TextureStorage9.h: Defines the abstract rx::TextureStorage9 class and its concrete derived +// classes TextureStorage9_2D and TextureStorage9_Cube, which act as the interface to the +// D3D9 texture. + +#ifndef LIBGLESV2_RENDERER_TEXTURESTORAGE9_H_ +#define LIBGLESV2_RENDERER_TEXTURESTORAGE9_H_ + +#include "libGLESv2/renderer/TextureStorage.h" +#include "common/debug.h" + +namespace rx +{ +class Renderer9; +class SwapChain9; +class RenderTarget; +class RenderTarget9; +class Blit; + +class TextureStorage9 : public TextureStorage +{ + public: + TextureStorage9(Renderer *renderer, DWORD usage); + virtual ~TextureStorage9(); + + static TextureStorage9 *makeTextureStorage9(TextureStorage *storage); + + static DWORD GetTextureUsage(D3DFORMAT d3dfmt, GLenum glusage, bool forceRenderable); + static bool IsTextureFormatRenderable(D3DFORMAT format); + + D3DPOOL getPool() const; + DWORD getUsage() const; + + virtual IDirect3DBaseTexture9 *getBaseTexture() const = 0; + virtual RenderTarget *getRenderTarget() { return NULL; } + virtual RenderTarget *getRenderTarget(GLenum faceTarget) { return NULL; } + virtual void generateMipmap(int level) {}; + virtual void generateMipmap(int face, int level) {}; + + virtual int getLodOffset() const; + virtual bool isRenderTarget() const; + virtual bool isManaged() const; + virtual int levelCount(); + + protected: + int mLodOffset; + Renderer9 *mRenderer; + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorage9); + + const DWORD mD3DUsage; + const D3DPOOL mD3DPool; +}; + +class TextureStorage9_2D : public TextureStorage9 +{ + public: + TextureStorage9_2D(Renderer *renderer, SwapChain9 *swapchain); + TextureStorage9_2D(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, GLsizei width, GLsizei height); + virtual ~TextureStorage9_2D(); + + static TextureStorage9_2D *makeTextureStorage9_2D(TextureStorage *storage); + + IDirect3DSurface9 *getSurfaceLevel(int level, bool dirty); + virtual RenderTarget *getRenderTarget(); + virtual IDirect3DBaseTexture9 *getBaseTexture() const; + virtual void generateMipmap(int level); + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorage9_2D); + + void initializeRenderTarget(); + + IDirect3DTexture9 *mTexture; + RenderTarget9 *mRenderTarget; +}; + +class TextureStorage9_Cube : public TextureStorage9 +{ + public: + TextureStorage9_Cube(Renderer *renderer, int levels, GLenum internalformat, GLenum usage, bool forceRenderable, int size); + virtual ~TextureStorage9_Cube(); + + static TextureStorage9_Cube *makeTextureStorage9_Cube(TextureStorage *storage); + + IDirect3DSurface9 *getCubeMapSurface(GLenum faceTarget, int level, bool dirty); + virtual RenderTarget *getRenderTarget(GLenum faceTarget); + virtual IDirect3DBaseTexture9 *getBaseTexture() const; + virtual void generateMipmap(int face, int level); + + private: + DISALLOW_COPY_AND_ASSIGN(TextureStorage9_Cube); + + void initializeRenderTarget(); + + IDirect3DCubeTexture9 *mTexture; + RenderTarget9 *mRenderTarget[6]; +}; + +} + +#endif // LIBGLESV2_RENDERER_TEXTURESTORAGE9_H_ + diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer.cpp new file mode 100644 index 0000000000..16e1486511 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer.cpp @@ -0,0 +1,229 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexBuffer.cpp: Defines the abstract VertexBuffer class and VertexBufferInterface +// class with derivations, classes that perform graphics API agnostic vertex buffer operations. + +#include "libGLESv2/renderer/VertexBuffer.h" +#include "libGLESv2/renderer/Renderer.h" +#include "libGLESv2/Context.h" + +namespace rx +{ + +unsigned int VertexBuffer::mNextSerial = 1; + +VertexBuffer::VertexBuffer() +{ + updateSerial(); +} + +VertexBuffer::~VertexBuffer() +{ +} + +void VertexBuffer::updateSerial() +{ + mSerial = mNextSerial++; +} + +unsigned int VertexBuffer::getSerial() const +{ + return mSerial; +} + +VertexBufferInterface::VertexBufferInterface(rx::Renderer *renderer, bool dynamic) : mRenderer(renderer) +{ + mDynamic = dynamic; + mWritePosition = 0; + mReservedSpace = 0; + + mVertexBuffer = renderer->createVertexBuffer(); +} + +VertexBufferInterface::~VertexBufferInterface() +{ + delete mVertexBuffer; +} + +unsigned int VertexBufferInterface::getSerial() const +{ + return mVertexBuffer->getSerial(); +} + +unsigned int VertexBufferInterface::getBufferSize() const +{ + return mVertexBuffer->getBufferSize(); +} + +bool VertexBufferInterface::setBufferSize(unsigned int size) +{ + if (mVertexBuffer->getBufferSize() == 0) + { + return mVertexBuffer->initialize(size, mDynamic); + } + else + { + return mVertexBuffer->setBufferSize(size); + } +} + +unsigned int VertexBufferInterface::getWritePosition() const +{ + return mWritePosition; +} + +void VertexBufferInterface::setWritePosition(unsigned int writePosition) +{ + mWritePosition = writePosition; +} + +bool VertexBufferInterface::discard() +{ + return mVertexBuffer->discard(); +} + +int VertexBufferInterface::storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, GLsizei instances) +{ + if (!reserveSpace(mReservedSpace)) + { + return -1; + } + mReservedSpace = 0; + + if (!mVertexBuffer->storeVertexAttributes(attrib, start, count, instances, mWritePosition)) + { + return -1; + } + + int oldWritePos = static_cast<int>(mWritePosition); + mWritePosition += mVertexBuffer->getSpaceRequired(attrib, count, instances); + + return oldWritePos; +} + +int VertexBufferInterface::storeRawData(const void* data, unsigned int size) +{ + if (!reserveSpace(mReservedSpace)) + { + return -1; + } + mReservedSpace = 0; + + if (!mVertexBuffer->storeRawData(data, size, mWritePosition)) + { + return -1; + } + + int oldWritePos = static_cast<int>(mWritePosition); + mWritePosition += size; + + return oldWritePos; +} + +void VertexBufferInterface::reserveVertexSpace(const gl::VertexAttribute &attribute, GLsizei count, GLsizei instances) +{ + mReservedSpace += mVertexBuffer->getSpaceRequired(attribute, count, instances); +} + +void VertexBufferInterface::reserveRawDataSpace(unsigned int size) +{ + mReservedSpace += size; +} + +VertexBuffer* VertexBufferInterface::getVertexBuffer() const +{ + return mVertexBuffer; +} + + +StreamingVertexBufferInterface::StreamingVertexBufferInterface(rx::Renderer *renderer, std::size_t initialSize) : VertexBufferInterface(renderer, true) +{ + setBufferSize(initialSize); +} + +StreamingVertexBufferInterface::~StreamingVertexBufferInterface() +{ +} + +bool StreamingVertexBufferInterface::reserveSpace(unsigned int size) +{ + bool result = true; + unsigned int curBufferSize = getBufferSize(); + if (size > curBufferSize) + { + result = setBufferSize(std::max(size, 3 * curBufferSize / 2)); + setWritePosition(0); + } + else if (getWritePosition() + size > curBufferSize) + { + if (!discard()) + { + return false; + } + setWritePosition(0); + } + + return result; +} + +StaticVertexBufferInterface::StaticVertexBufferInterface(rx::Renderer *renderer) : VertexBufferInterface(renderer, false) +{ +} + +StaticVertexBufferInterface::~StaticVertexBufferInterface() +{ +} + +int StaticVertexBufferInterface::lookupAttribute(const gl::VertexAttribute &attribute) +{ + for (unsigned int element = 0; element < mCache.size(); element++) + { + if (mCache[element].type == attribute.mType && + mCache[element].size == attribute.mSize && + mCache[element].stride == attribute.stride() && + mCache[element].normalized == attribute.mNormalized) + { + if (mCache[element].attributeOffset == attribute.mOffset % attribute.stride()) + { + return mCache[element].streamOffset; + } + } + } + + return -1; +} + +bool StaticVertexBufferInterface::reserveSpace(unsigned int size) +{ + unsigned int curSize = getBufferSize(); + if (curSize == 0) + { + setBufferSize(size); + return true; + } + else if (curSize >= size) + { + return true; + } + else + { + UNREACHABLE(); // Static vertex buffers can't be resized + return false; + } +} + +int StaticVertexBufferInterface::storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, GLsizei instances) +{ + int attributeOffset = attrib.mOffset % attrib.stride(); + VertexElement element = { attrib.mType, attrib.mSize, attrib.stride(), attrib.mNormalized, attributeOffset, getWritePosition() }; + mCache.push_back(element); + + return VertexBufferInterface::storeVertexAttributes(attrib, start, count, instances); +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer.h b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer.h new file mode 100644 index 0000000000..6ecffd0c0b --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer.h @@ -0,0 +1,138 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexBuffer.h: Defines the abstract VertexBuffer class and VertexBufferInterface +// class with derivations, classes that perform graphics API agnostic vertex buffer operations. + +#ifndef LIBGLESV2_RENDERER_VERTEXBUFFER_H_ +#define LIBGLESV2_RENDERER_VERTEXBUFFER_H_ + +#include "common/angleutils.h" + +namespace gl +{ +class VertexAttribute; +} + +namespace rx +{ +class Renderer; + +class VertexBuffer +{ + public: + VertexBuffer(); + virtual ~VertexBuffer(); + + virtual bool initialize(unsigned int size, bool dynamicUsage) = 0; + + virtual bool storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, + GLsizei instances, unsigned int offset) = 0; + virtual bool storeRawData(const void* data, unsigned int size, unsigned int offset) = 0; + + virtual unsigned int getSpaceRequired(const gl::VertexAttribute &attrib, GLsizei count, + GLsizei instances) const = 0; + + virtual bool requiresConversion(const gl::VertexAttribute &attrib) const = 0; + + virtual unsigned int getBufferSize() const = 0; + virtual bool setBufferSize(unsigned int size) = 0; + virtual bool discard() = 0; + + unsigned int getSerial() const; + + protected: + void updateSerial(); + + private: + DISALLOW_COPY_AND_ASSIGN(VertexBuffer); + + unsigned int mSerial; + static unsigned int mNextSerial; +}; + +class VertexBufferInterface +{ + public: + VertexBufferInterface(rx::Renderer *renderer, bool dynamic); + virtual ~VertexBufferInterface(); + + void reserveVertexSpace(const gl::VertexAttribute &attribute, GLsizei count, GLsizei instances); + void reserveRawDataSpace(unsigned int size); + + unsigned int getBufferSize() const; + + unsigned int getSerial() const; + + virtual int storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, GLsizei instances); + virtual int storeRawData(const void* data, unsigned int size); + + VertexBuffer* getVertexBuffer() const; + + protected: + virtual bool reserveSpace(unsigned int size) = 0; + + unsigned int getWritePosition() const; + void setWritePosition(unsigned int writePosition); + + bool discard(); + + bool setBufferSize(unsigned int size); + + private: + DISALLOW_COPY_AND_ASSIGN(VertexBufferInterface); + + rx::Renderer *const mRenderer; + + VertexBuffer* mVertexBuffer; + + unsigned int mWritePosition; + unsigned int mReservedSpace; + bool mDynamic; +}; + +class StreamingVertexBufferInterface : public VertexBufferInterface +{ + public: + StreamingVertexBufferInterface(rx::Renderer *renderer, std::size_t initialSize); + ~StreamingVertexBufferInterface(); + + protected: + bool reserveSpace(unsigned int size); +}; + +class StaticVertexBufferInterface : public VertexBufferInterface +{ + public: + explicit StaticVertexBufferInterface(rx::Renderer *renderer); + ~StaticVertexBufferInterface(); + + int storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, GLsizei instances); + + // Returns the offset into the vertex buffer, or -1 if not found + int lookupAttribute(const gl::VertexAttribute &attribute); + + protected: + bool reserveSpace(unsigned int size); + + private: + struct VertexElement + { + GLenum type; + GLint size; + GLsizei stride; + bool normalized; + int attributeOffset; + + unsigned int streamOffset; + }; + + std::vector<VertexElement> mCache; +}; + +} + +#endif // LIBGLESV2_RENDERER_VERTEXBUFFER_H_
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer11.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer11.cpp new file mode 100644 index 0000000000..92c8755e22 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer11.cpp @@ -0,0 +1,418 @@ +#include "precompiled.h" +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexBuffer11.cpp: Defines the D3D11 VertexBuffer implementation. + +#include "libGLESv2/renderer/VertexBuffer11.h" +#include "libGLESv2/renderer/BufferStorage.h" + +#include "libGLESv2/Buffer.h" +#include "libGLESv2/renderer/Renderer11.h" +#include "libGLESv2/Context.h" + +namespace rx +{ + +VertexBuffer11::VertexBuffer11(rx::Renderer11 *const renderer) : mRenderer(renderer) +{ + mBuffer = NULL; + mBufferSize = 0; + mDynamicUsage = false; +} + +VertexBuffer11::~VertexBuffer11() +{ + if (mBuffer) + { + mBuffer->Release(); + mBuffer = NULL; + } +} + +bool VertexBuffer11::initialize(unsigned int size, bool dynamicUsage) +{ + if (mBuffer) + { + mBuffer->Release(); + mBuffer = NULL; + } + + updateSerial(); + + if (size > 0) + { + ID3D11Device* dxDevice = mRenderer->getDevice(); + + D3D11_BUFFER_DESC bufferDesc; + bufferDesc.ByteWidth = size; + bufferDesc.Usage = D3D11_USAGE_DYNAMIC; + bufferDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + bufferDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + bufferDesc.MiscFlags = 0; + bufferDesc.StructureByteStride = 0; + + HRESULT result = dxDevice->CreateBuffer(&bufferDesc, NULL, &mBuffer); + if (FAILED(result)) + { + return false; + } + } + + mBufferSize = size; + mDynamicUsage = dynamicUsage; + return true; +} + +VertexBuffer11 *VertexBuffer11::makeVertexBuffer11(VertexBuffer *vetexBuffer) +{ + ASSERT(HAS_DYNAMIC_TYPE(VertexBuffer11*, vetexBuffer)); + return static_cast<VertexBuffer11*>(vetexBuffer); +} + +bool VertexBuffer11::storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, + GLsizei instances, unsigned int offset) +{ + if (mBuffer) + { + gl::Buffer *buffer = attrib.mBoundBuffer.get(); + + int inputStride = attrib.stride(); + const VertexConverter &converter = getVertexConversion(attrib); + + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_NO_OVERWRITE, 0, &mappedResource); + if (FAILED(result)) + { + ERR("Vertex buffer map failed with error 0x%08x", result); + return false; + } + + char* output = reinterpret_cast<char*>(mappedResource.pData) + offset; + + const char *input = NULL; + if (buffer) + { + BufferStorage *storage = buffer->getStorage(); + input = static_cast<const char*>(storage->getData()) + static_cast<int>(attrib.mOffset); + } + else + { + input = static_cast<const char*>(attrib.mPointer); + } + + if (instances == 0 || attrib.mDivisor == 0) + { + input += inputStride * start; + } + + converter.conversionFunc(input, inputStride, count, output); + + dxContext->Unmap(mBuffer, 0); + + return true; + } + else + { + ERR("Vertex buffer not initialized."); + return false; + } +} + +bool VertexBuffer11::storeRawData(const void* data, unsigned int size, unsigned int offset) +{ + if (mBuffer) + { + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_NO_OVERWRITE, 0, &mappedResource); + if (FAILED(result)) + { + ERR("Vertex buffer map failed with error 0x%08x", result); + return false; + } + + char* bufferData = static_cast<char*>(mappedResource.pData); + memcpy(bufferData + offset, data, size); + + dxContext->Unmap(mBuffer, 0); + + return true; + } + else + { + ERR("Vertex buffer not initialized."); + return false; + } +} + +unsigned int VertexBuffer11::getSpaceRequired(const gl::VertexAttribute &attrib, GLsizei count, + GLsizei instances) const +{ + unsigned int elementSize = getVertexConversion(attrib).outputElementSize; + + if (instances == 0 || attrib.mDivisor == 0) + { + return elementSize * count; + } + else + { + return elementSize * ((instances + attrib.mDivisor - 1) / attrib.mDivisor); + } +} + +bool VertexBuffer11::requiresConversion(const gl::VertexAttribute &attrib) const +{ + return !getVertexConversion(attrib).identity; +} + +unsigned int VertexBuffer11::getBufferSize() const +{ + return mBufferSize; +} + +bool VertexBuffer11::setBufferSize(unsigned int size) +{ + if (size > mBufferSize) + { + return initialize(size, mDynamicUsage); + } + else + { + return true; + } +} + +bool VertexBuffer11::discard() +{ + if (mBuffer) + { + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + ERR("Vertex buffer map failed with error 0x%08x", result); + return false; + } + + dxContext->Unmap(mBuffer, 0); + + return true; + } + else + { + ERR("Vertex buffer not initialized."); + return false; + } +} + +unsigned int VertexBuffer11::getVertexSize(const gl::VertexAttribute &attrib) const +{ + return getVertexConversion(attrib).outputElementSize; +} + +DXGI_FORMAT VertexBuffer11::getDXGIFormat(const gl::VertexAttribute &attrib) const +{ + return getVertexConversion(attrib).dxgiFormat; +} + +ID3D11Buffer *VertexBuffer11::getBuffer() const +{ + return mBuffer; +} + +template <typename T, unsigned int componentCount, bool widen, bool normalized> +static void copyVertexData(const void *input, unsigned int stride, unsigned int count, void *output) +{ + unsigned int attribSize = sizeof(T) * componentCount; + + if (attribSize == stride && !widen) + { + memcpy(output, input, count * attribSize); + } + else + { + unsigned int outputStride = widen ? 4 : componentCount; + T defaultVal = normalized ? std::numeric_limits<T>::max() : T(1); + + for (unsigned int i = 0; i < count; i++) + { + const T *offsetInput = reinterpret_cast<const T*>(reinterpret_cast<const char*>(input) + i * stride); + T *offsetOutput = reinterpret_cast<T*>(output) + i * outputStride; + + for (unsigned int j = 0; j < componentCount; j++) + { + offsetOutput[j] = offsetInput[j]; + } + + if (widen) + { + offsetOutput[3] = defaultVal; + } + } + } +} + +template <unsigned int componentCount> +static void copyFixedVertexData(const void* input, unsigned int stride, unsigned int count, void* output) +{ + static const float divisor = 1.0f / (1 << 16); + + for (unsigned int i = 0; i < count; i++) + { + const GLfixed* offsetInput = reinterpret_cast<const GLfixed*>(reinterpret_cast<const char*>(input) + stride * i); + float* offsetOutput = reinterpret_cast<float*>(output) + i * componentCount; + + for (unsigned int j = 0; j < componentCount; j++) + { + offsetOutput[j] = static_cast<float>(offsetInput[j]) * divisor; + } + } +} + +template <typename T, unsigned int componentCount, bool normalized> +static void copyToFloatVertexData(const void* input, unsigned int stride, unsigned int count, void* output) +{ + typedef std::numeric_limits<T> NL; + + for (unsigned int i = 0; i < count; i++) + { + const T *offsetInput = reinterpret_cast<const T*>(reinterpret_cast<const char*>(input) + stride * i); + float *offsetOutput = reinterpret_cast<float*>(output) + i * componentCount; + + for (unsigned int j = 0; j < componentCount; j++) + { + if (normalized) + { + if (NL::is_signed) + { + const float divisor = 1.0f / (2 * static_cast<float>(NL::max()) + 1); + offsetOutput[j] = (2 * static_cast<float>(offsetInput[j]) + 1) * divisor; + } + else + { + offsetOutput[j] = static_cast<float>(offsetInput[j]) / NL::max(); + } + } + else + { + offsetOutput[j] = static_cast<float>(offsetInput[j]); + } + } + } +} + +const VertexBuffer11::VertexConverter VertexBuffer11::mPossibleTranslations[NUM_GL_VERTEX_ATTRIB_TYPES][2][4] = +{ + { // GL_BYTE + { // unnormalized + { ©ToFloatVertexData<GLbyte, 1, false>, false, DXGI_FORMAT_R32_FLOAT, 4 }, + { ©ToFloatVertexData<GLbyte, 2, false>, false, DXGI_FORMAT_R32G32_FLOAT, 8 }, + { ©ToFloatVertexData<GLbyte, 3, false>, false, DXGI_FORMAT_R32G32B32_FLOAT, 12 }, + { ©ToFloatVertexData<GLbyte, 4, false>, false, DXGI_FORMAT_R32G32B32A32_FLOAT, 16 }, + }, + { // normalized + { ©VertexData<GLbyte, 1, false, true>, true, DXGI_FORMAT_R8_SNORM, 1 }, + { ©VertexData<GLbyte, 2, false, true>, true, DXGI_FORMAT_R8G8_SNORM, 2 }, + { ©VertexData<GLbyte, 3, true, true>, false, DXGI_FORMAT_R8G8B8A8_SNORM, 4 }, + { ©VertexData<GLbyte, 4, false, true>, true, DXGI_FORMAT_R8G8B8A8_SNORM, 4 }, + }, + }, + { // GL_UNSIGNED_BYTE + { // unnormalized + { ©ToFloatVertexData<GLubyte, 1, false>, false, DXGI_FORMAT_R32_FLOAT, 4 }, + { ©ToFloatVertexData<GLubyte, 2, false>, false, DXGI_FORMAT_R32G32_FLOAT, 8 }, + { ©ToFloatVertexData<GLubyte, 3, false>, false, DXGI_FORMAT_R32G32B32_FLOAT, 12 }, + { ©ToFloatVertexData<GLubyte, 4, false>, false, DXGI_FORMAT_R32G32B32A32_FLOAT, 16 }, + }, + { // normalized + { ©VertexData<GLubyte, 1, false, true>, true, DXGI_FORMAT_R8_UNORM, 1 }, + { ©VertexData<GLubyte, 2, false, true>, true, DXGI_FORMAT_R8G8_UNORM, 2 }, + { ©VertexData<GLubyte, 3, true, true>, false, DXGI_FORMAT_R8G8B8A8_UNORM, 4 }, + { ©VertexData<GLubyte, 4, false, true>, true, DXGI_FORMAT_R8G8B8A8_UNORM, 4 }, + }, + }, + { // GL_SHORT + { // unnormalized + { ©ToFloatVertexData<GLshort, 1, false>, false, DXGI_FORMAT_R32_FLOAT, 4 }, + { ©ToFloatVertexData<GLshort, 2, false>, false, DXGI_FORMAT_R32G32_FLOAT, 8 }, + { ©ToFloatVertexData<GLshort, 3, false>, false, DXGI_FORMAT_R32G32B32_FLOAT, 12 }, + { ©ToFloatVertexData<GLshort, 4, false>, false, DXGI_FORMAT_R32G32B32A32_FLOAT, 16 }, + }, + { // normalized + { ©VertexData<GLshort, 1, false, true>, true, DXGI_FORMAT_R16_SNORM, 2 }, + { ©VertexData<GLshort, 2, false, true>, true, DXGI_FORMAT_R16G16_SNORM, 4 }, + { ©VertexData<GLshort, 3, true, true>, false, DXGI_FORMAT_R16G16B16A16_SNORM, 8 }, + { ©VertexData<GLshort, 4, false, true>, true, DXGI_FORMAT_R16G16B16A16_SNORM, 8 }, + }, + }, + { // GL_UNSIGNED_SHORT + { // unnormalized + { ©ToFloatVertexData<GLushort, 1, false>, false, DXGI_FORMAT_R32_FLOAT, 4 }, + { ©ToFloatVertexData<GLushort, 2, false>, false, DXGI_FORMAT_R32G32_FLOAT, 8 }, + { ©ToFloatVertexData<GLushort, 3, false>, false, DXGI_FORMAT_R32G32B32_FLOAT, 12 }, + { ©ToFloatVertexData<GLushort, 4, false>, false, DXGI_FORMAT_R32G32B32A32_FLOAT, 16 }, + }, + { // normalized + { ©VertexData<GLushort, 1, false, true>, true, DXGI_FORMAT_R16_UNORM, 2 }, + { ©VertexData<GLushort, 2, false, true>, true, DXGI_FORMAT_R16G16_UNORM, 4 }, + { ©VertexData<GLushort, 3, true, true>, false, DXGI_FORMAT_R16G16B16A16_UNORM, 8 }, + { ©VertexData<GLushort, 4, false, true>, true, DXGI_FORMAT_R16G16B16A16_UNORM, 8 }, + }, + }, + { // GL_FIXED + { // unnormalized + { ©FixedVertexData<1>, false, DXGI_FORMAT_R32_FLOAT, 4 }, + { ©FixedVertexData<2>, false, DXGI_FORMAT_R32G32_FLOAT, 8 }, + { ©FixedVertexData<3>, false, DXGI_FORMAT_R32G32B32_FLOAT, 12 }, + { ©FixedVertexData<4>, false, DXGI_FORMAT_R32G32B32A32_FLOAT, 16 }, + }, + { // normalized + { ©FixedVertexData<1>, false, DXGI_FORMAT_R32_FLOAT, 4 }, + { ©FixedVertexData<2>, false, DXGI_FORMAT_R32G32_FLOAT, 8 }, + { ©FixedVertexData<3>, false, DXGI_FORMAT_R32G32B32_FLOAT, 12 }, + { ©FixedVertexData<4>, false, DXGI_FORMAT_R32G32B32A32_FLOAT, 16 }, + }, + }, + { // GL_FLOAT + { // unnormalized + { ©VertexData<GLfloat, 1, false, false>, true, DXGI_FORMAT_R32_FLOAT, 4 }, + { ©VertexData<GLfloat, 2, false, false>, true, DXGI_FORMAT_R32G32_FLOAT, 8 }, + { ©VertexData<GLfloat, 3, false, false>, true, DXGI_FORMAT_R32G32B32_FLOAT, 12 }, + { ©VertexData<GLfloat, 4, false, false>, true, DXGI_FORMAT_R32G32B32A32_FLOAT, 16 }, + }, + { // normalized + { ©VertexData<GLfloat, 1, false, false>, true, DXGI_FORMAT_R32_FLOAT, 4 }, + { ©VertexData<GLfloat, 2, false, false>, true, DXGI_FORMAT_R32G32_FLOAT, 8 }, + { ©VertexData<GLfloat, 3, false, false>, true, DXGI_FORMAT_R32G32B32_FLOAT, 12 }, + { ©VertexData<GLfloat, 4, false, false>, true, DXGI_FORMAT_R32G32B32A32_FLOAT, 16 }, + }, + }, +}; + +const VertexBuffer11::VertexConverter &VertexBuffer11::getVertexConversion(const gl::VertexAttribute &attribute) +{ + unsigned int typeIndex = 0; + switch (attribute.mType) + { + case GL_BYTE: typeIndex = 0; break; + case GL_UNSIGNED_BYTE: typeIndex = 1; break; + case GL_SHORT: typeIndex = 2; break; + case GL_UNSIGNED_SHORT: typeIndex = 3; break; + case GL_FIXED: typeIndex = 4; break; + case GL_FLOAT: typeIndex = 5; break; + default: UNREACHABLE(); break; + } + + return mPossibleTranslations[typeIndex][attribute.mNormalized ? 1 : 0][attribute.mSize - 1]; +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer11.h b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer11.h new file mode 100644 index 0000000000..75e025075e --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer11.h @@ -0,0 +1,73 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexBuffer11.h: Defines the D3D11 VertexBuffer implementation. + +#ifndef LIBGLESV2_RENDERER_VERTEXBUFFER11_H_ +#define LIBGLESV2_RENDERER_VERTEXBUFFER11_H_ + +#include "libGLESv2/renderer/VertexBuffer.h" + +namespace rx +{ +class Renderer11; + +class VertexBuffer11 : public VertexBuffer +{ + public: + explicit VertexBuffer11(rx::Renderer11 *const renderer); + virtual ~VertexBuffer11(); + + virtual bool initialize(unsigned int size, bool dynamicUsage); + + static VertexBuffer11 *makeVertexBuffer11(VertexBuffer *vetexBuffer); + + virtual bool storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, GLsizei instances, + unsigned int offset); + virtual bool storeRawData(const void* data, unsigned int size, unsigned int offset); + + virtual unsigned int getSpaceRequired(const gl::VertexAttribute &attrib, GLsizei count, GLsizei instances) const; + + virtual bool requiresConversion(const gl::VertexAttribute &attrib) const; + + virtual unsigned int getBufferSize() const; + virtual bool setBufferSize(unsigned int size); + virtual bool discard(); + + unsigned int getVertexSize(const gl::VertexAttribute &attrib) const; + DXGI_FORMAT getDXGIFormat(const gl::VertexAttribute &attrib) const; + + ID3D11Buffer *getBuffer() const; + + private: + DISALLOW_COPY_AND_ASSIGN(VertexBuffer11); + + rx::Renderer11 *const mRenderer; + + ID3D11Buffer *mBuffer; + unsigned int mBufferSize; + bool mDynamicUsage; + + typedef void (*VertexConversionFunction)(const void *, unsigned int, unsigned int, void *); + struct VertexConverter + { + VertexConversionFunction conversionFunc; + bool identity; + DXGI_FORMAT dxgiFormat; + unsigned int outputElementSize; + }; + + enum { NUM_GL_VERTEX_ATTRIB_TYPES = 6 }; + + // This table is used to generate mAttributeTypes. + static const VertexConverter mPossibleTranslations[NUM_GL_VERTEX_ATTRIB_TYPES][2][4]; // [GL types as enumerated by typeIndex()][normalized][size - 1] + + static const VertexConverter &getVertexConversion(const gl::VertexAttribute &attribute); +}; + +} + +#endif // LIBGLESV2_RENDERER_VERTEXBUFFER11_H_
\ No newline at end of file diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer9.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer9.cpp new file mode 100644 index 0000000000..76dc73e391 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer9.cpp @@ -0,0 +1,486 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexBuffer9.cpp: Defines the D3D9 VertexBuffer implementation. + +#include "libGLESv2/renderer/VertexBuffer9.h" +#include "libGLESv2/renderer/vertexconversion.h" +#include "libGLESv2/renderer/BufferStorage.h" +#include "libGLESv2/Context.h" +#include "libGLESv2/renderer/Renderer9.h" + +#include "libGLESv2/Buffer.h" + +namespace rx +{ + +bool VertexBuffer9::mTranslationsInitialized = false; +VertexBuffer9::FormatConverter VertexBuffer9::mFormatConverters[NUM_GL_VERTEX_ATTRIB_TYPES][2][4]; + +VertexBuffer9::VertexBuffer9(rx::Renderer9 *const renderer) : mRenderer(renderer) +{ + mVertexBuffer = NULL; + mBufferSize = 0; + mDynamicUsage = false; + + if (!mTranslationsInitialized) + { + initializeTranslations(renderer->getCapsDeclTypes()); + mTranslationsInitialized = true; + } +} + +VertexBuffer9::~VertexBuffer9() +{ + if (mVertexBuffer) + { + mVertexBuffer->Release(); + mVertexBuffer = NULL; + } +} + +bool VertexBuffer9::initialize(unsigned int size, bool dynamicUsage) +{ + if (mVertexBuffer) + { + mVertexBuffer->Release(); + mVertexBuffer = NULL; + } + + updateSerial(); + + if (size > 0) + { + DWORD flags = D3DUSAGE_WRITEONLY; + if (dynamicUsage) + { + flags |= D3DUSAGE_DYNAMIC; + } + + HRESULT result = mRenderer->createVertexBuffer(size, flags, &mVertexBuffer); + + if (FAILED(result)) + { + ERR("Out of memory allocating a vertex buffer of size %lu.", size); + return false; + } + } + + mBufferSize = size; + mDynamicUsage = dynamicUsage; + return true; +} + +VertexBuffer9 *VertexBuffer9::makeVertexBuffer9(VertexBuffer *vertexBuffer) +{ + ASSERT(HAS_DYNAMIC_TYPE(VertexBuffer9*, vertexBuffer)); + return static_cast<VertexBuffer9*>(vertexBuffer); +} + +bool VertexBuffer9::storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, + GLsizei instances, unsigned int offset) +{ + if (mVertexBuffer) + { + gl::Buffer *buffer = attrib.mBoundBuffer.get(); + + int inputStride = attrib.stride(); + int elementSize = attrib.typeSize(); + const FormatConverter &converter = formatConverter(attrib); + + DWORD lockFlags = mDynamicUsage ? D3DLOCK_NOOVERWRITE : 0; + + void *mapPtr = NULL; + HRESULT result = mVertexBuffer->Lock(offset, spaceRequired(attrib, count, instances), &mapPtr, lockFlags); + + if (FAILED(result)) + { + ERR("Lock failed with error 0x%08x", result); + return false; + } + + const char *input = NULL; + if (buffer) + { + BufferStorage *storage = buffer->getStorage(); + input = static_cast<const char*>(storage->getData()) + static_cast<int>(attrib.mOffset); + } + else + { + input = static_cast<const char*>(attrib.mPointer); + } + + if (instances == 0 || attrib.mDivisor == 0) + { + input += inputStride * start; + } + + if (converter.identity && inputStride == elementSize) + { + memcpy(mapPtr, input, count * inputStride); + } + else + { + converter.convertArray(input, inputStride, count, mapPtr); + } + + mVertexBuffer->Unlock(); + + return true; + } + else + { + ERR("Vertex buffer not initialized."); + return false; + } +} + +bool VertexBuffer9::storeRawData(const void* data, unsigned int size, unsigned int offset) +{ + if (mVertexBuffer) + { + DWORD lockFlags = mDynamicUsage ? D3DLOCK_NOOVERWRITE : 0; + + void *mapPtr = NULL; + HRESULT result = mVertexBuffer->Lock(offset, size, &mapPtr, lockFlags); + + if (FAILED(result)) + { + ERR("Lock failed with error 0x%08x", result); + return false; + } + + memcpy(mapPtr, data, size); + + mVertexBuffer->Unlock(); + + return true; + } + else + { + ERR("Vertex buffer not initialized."); + return false; + } +} + +unsigned int VertexBuffer9::getSpaceRequired(const gl::VertexAttribute &attrib, GLsizei count, GLsizei instances) const +{ + return spaceRequired(attrib, count, instances); +} + +bool VertexBuffer9::requiresConversion(const gl::VertexAttribute &attrib) const +{ + return formatConverter(attrib).identity; +} + +unsigned int VertexBuffer9::getVertexSize(const gl::VertexAttribute &attrib) const +{ + return spaceRequired(attrib, 1, 0); +} + +D3DDECLTYPE VertexBuffer9::getDeclType(const gl::VertexAttribute &attrib) const +{ + return formatConverter(attrib).d3dDeclType; +} + +unsigned int VertexBuffer9::getBufferSize() const +{ + return mBufferSize; +} + +bool VertexBuffer9::setBufferSize(unsigned int size) +{ + if (size > mBufferSize) + { + return initialize(size, mDynamicUsage); + } + else + { + return true; + } +} + +bool VertexBuffer9::discard() +{ + if (mVertexBuffer) + { + void *dummy; + HRESULT result; + + result = mVertexBuffer->Lock(0, 1, &dummy, D3DLOCK_DISCARD); + if (FAILED(result)) + { + ERR("Discard lock failed with error 0x%08x", result); + return false; + } + + result = mVertexBuffer->Unlock(); + if (FAILED(result)) + { + ERR("Discard unlock failed with error 0x%08x", result); + return false; + } + + return true; + } + else + { + ERR("Vertex buffer not initialized."); + return false; + } +} + +IDirect3DVertexBuffer9 * VertexBuffer9::getBuffer() const +{ + return mVertexBuffer; +} + +// Mapping from OpenGL-ES vertex attrib type to D3D decl type: +// +// BYTE SHORT (Cast) +// BYTE-norm FLOAT (Normalize) (can't be exactly represented as SHORT-norm) +// UNSIGNED_BYTE UBYTE4 (Identity) or SHORT (Cast) +// UNSIGNED_BYTE-norm UBYTE4N (Identity) or FLOAT (Normalize) +// SHORT SHORT (Identity) +// SHORT-norm SHORT-norm (Identity) or FLOAT (Normalize) +// UNSIGNED_SHORT FLOAT (Cast) +// UNSIGNED_SHORT-norm USHORT-norm (Identity) or FLOAT (Normalize) +// FIXED (not in WebGL) FLOAT (FixedToFloat) +// FLOAT FLOAT (Identity) + +// GLToCType maps from GL type (as GLenum) to the C typedef. +template <GLenum GLType> struct GLToCType { }; + +template <> struct GLToCType<GL_BYTE> { typedef GLbyte type; }; +template <> struct GLToCType<GL_UNSIGNED_BYTE> { typedef GLubyte type; }; +template <> struct GLToCType<GL_SHORT> { typedef GLshort type; }; +template <> struct GLToCType<GL_UNSIGNED_SHORT> { typedef GLushort type; }; +template <> struct GLToCType<GL_FIXED> { typedef GLuint type; }; +template <> struct GLToCType<GL_FLOAT> { typedef GLfloat type; }; + +// This differs from D3DDECLTYPE in that it is unsized. (Size expansion is applied last.) +enum D3DVertexType +{ + D3DVT_FLOAT, + D3DVT_SHORT, + D3DVT_SHORT_NORM, + D3DVT_UBYTE, + D3DVT_UBYTE_NORM, + D3DVT_USHORT_NORM +}; + +// D3DToCType maps from D3D vertex type (as enum D3DVertexType) to the corresponding C type. +template <unsigned int D3DType> struct D3DToCType { }; + +template <> struct D3DToCType<D3DVT_FLOAT> { typedef float type; }; +template <> struct D3DToCType<D3DVT_SHORT> { typedef short type; }; +template <> struct D3DToCType<D3DVT_SHORT_NORM> { typedef short type; }; +template <> struct D3DToCType<D3DVT_UBYTE> { typedef unsigned char type; }; +template <> struct D3DToCType<D3DVT_UBYTE_NORM> { typedef unsigned char type; }; +template <> struct D3DToCType<D3DVT_USHORT_NORM> { typedef unsigned short type; }; + +// Encode the type/size combinations that D3D permits. For each type/size it expands to a widener that will provide the appropriate final size. +template <unsigned int type, int size> struct WidenRule { }; + +template <int size> struct WidenRule<D3DVT_FLOAT, size> : NoWiden<size> { }; +template <int size> struct WidenRule<D3DVT_SHORT, size> : WidenToEven<size> { }; +template <int size> struct WidenRule<D3DVT_SHORT_NORM, size> : WidenToEven<size> { }; +template <int size> struct WidenRule<D3DVT_UBYTE, size> : WidenToFour<size> { }; +template <int size> struct WidenRule<D3DVT_UBYTE_NORM, size> : WidenToFour<size> { }; +template <int size> struct WidenRule<D3DVT_USHORT_NORM, size> : WidenToEven<size> { }; + +// VertexTypeFlags encodes the D3DCAPS9::DeclType flag and vertex declaration flag for each D3D vertex type & size combination. +template <unsigned int d3dtype, int size> struct VertexTypeFlags { }; + +template <unsigned int _capflag, unsigned int _declflag> +struct VertexTypeFlagsHelper +{ + enum { capflag = _capflag }; + enum { declflag = _declflag }; +}; + +template <> struct VertexTypeFlags<D3DVT_FLOAT, 1> : VertexTypeFlagsHelper<0, D3DDECLTYPE_FLOAT1> { }; +template <> struct VertexTypeFlags<D3DVT_FLOAT, 2> : VertexTypeFlagsHelper<0, D3DDECLTYPE_FLOAT2> { }; +template <> struct VertexTypeFlags<D3DVT_FLOAT, 3> : VertexTypeFlagsHelper<0, D3DDECLTYPE_FLOAT3> { }; +template <> struct VertexTypeFlags<D3DVT_FLOAT, 4> : VertexTypeFlagsHelper<0, D3DDECLTYPE_FLOAT4> { }; +template <> struct VertexTypeFlags<D3DVT_SHORT, 2> : VertexTypeFlagsHelper<0, D3DDECLTYPE_SHORT2> { }; +template <> struct VertexTypeFlags<D3DVT_SHORT, 4> : VertexTypeFlagsHelper<0, D3DDECLTYPE_SHORT4> { }; +template <> struct VertexTypeFlags<D3DVT_SHORT_NORM, 2> : VertexTypeFlagsHelper<D3DDTCAPS_SHORT2N, D3DDECLTYPE_SHORT2N> { }; +template <> struct VertexTypeFlags<D3DVT_SHORT_NORM, 4> : VertexTypeFlagsHelper<D3DDTCAPS_SHORT4N, D3DDECLTYPE_SHORT4N> { }; +template <> struct VertexTypeFlags<D3DVT_UBYTE, 4> : VertexTypeFlagsHelper<D3DDTCAPS_UBYTE4, D3DDECLTYPE_UBYTE4> { }; +template <> struct VertexTypeFlags<D3DVT_UBYTE_NORM, 4> : VertexTypeFlagsHelper<D3DDTCAPS_UBYTE4N, D3DDECLTYPE_UBYTE4N> { }; +template <> struct VertexTypeFlags<D3DVT_USHORT_NORM, 2> : VertexTypeFlagsHelper<D3DDTCAPS_USHORT2N, D3DDECLTYPE_USHORT2N> { }; +template <> struct VertexTypeFlags<D3DVT_USHORT_NORM, 4> : VertexTypeFlagsHelper<D3DDTCAPS_USHORT4N, D3DDECLTYPE_USHORT4N> { }; + + +// VertexTypeMapping maps GL type & normalized flag to preferred and fallback D3D vertex types (as D3DVertexType enums). +template <GLenum GLtype, bool normalized> struct VertexTypeMapping { }; + +template <D3DVertexType Preferred, D3DVertexType Fallback = Preferred> +struct VertexTypeMappingBase +{ + enum { preferred = Preferred }; + enum { fallback = Fallback }; +}; + +template <> struct VertexTypeMapping<GL_BYTE, false> : VertexTypeMappingBase<D3DVT_SHORT> { }; // Cast +template <> struct VertexTypeMapping<GL_BYTE, true> : VertexTypeMappingBase<D3DVT_FLOAT> { }; // Normalize +template <> struct VertexTypeMapping<GL_UNSIGNED_BYTE, false> : VertexTypeMappingBase<D3DVT_UBYTE, D3DVT_FLOAT> { }; // Identity, Cast +template <> struct VertexTypeMapping<GL_UNSIGNED_BYTE, true> : VertexTypeMappingBase<D3DVT_UBYTE_NORM, D3DVT_FLOAT> { }; // Identity, Normalize +template <> struct VertexTypeMapping<GL_SHORT, false> : VertexTypeMappingBase<D3DVT_SHORT> { }; // Identity +template <> struct VertexTypeMapping<GL_SHORT, true> : VertexTypeMappingBase<D3DVT_SHORT_NORM, D3DVT_FLOAT> { }; // Cast, Normalize +template <> struct VertexTypeMapping<GL_UNSIGNED_SHORT, false> : VertexTypeMappingBase<D3DVT_FLOAT> { }; // Cast +template <> struct VertexTypeMapping<GL_UNSIGNED_SHORT, true> : VertexTypeMappingBase<D3DVT_USHORT_NORM, D3DVT_FLOAT> { }; // Cast, Normalize +template <bool normalized> struct VertexTypeMapping<GL_FIXED, normalized> : VertexTypeMappingBase<D3DVT_FLOAT> { }; // FixedToFloat +template <bool normalized> struct VertexTypeMapping<GL_FLOAT, normalized> : VertexTypeMappingBase<D3DVT_FLOAT> { }; // Identity + + +// Given a GL type & norm flag and a D3D type, ConversionRule provides the type conversion rule (Cast, Normalize, Identity, FixedToFloat). +// The conversion rules themselves are defined in vertexconversion.h. + +// Almost all cases are covered by Cast (including those that are actually Identity since Cast<T,T> knows it's an identity mapping). +template <GLenum fromType, bool normalized, unsigned int toType> +struct ConversionRule : Cast<typename GLToCType<fromType>::type, typename D3DToCType<toType>::type> { }; + +// All conversions from normalized types to float use the Normalize operator. +template <GLenum fromType> struct ConversionRule<fromType, true, D3DVT_FLOAT> : Normalize<typename GLToCType<fromType>::type> { }; + +// Use a full specialization for this so that it preferentially matches ahead of the generic normalize-to-float rules. +template <> struct ConversionRule<GL_FIXED, true, D3DVT_FLOAT> : FixedToFloat<GLint, 16> { }; +template <> struct ConversionRule<GL_FIXED, false, D3DVT_FLOAT> : FixedToFloat<GLint, 16> { }; + +// A 2-stage construction is used for DefaultVertexValues because float must use SimpleDefaultValues (i.e. 0/1) +// whether it is normalized or not. +template <class T, bool normalized> struct DefaultVertexValuesStage2 { }; + +template <class T> struct DefaultVertexValuesStage2<T, true> : NormalizedDefaultValues<T> { }; +template <class T> struct DefaultVertexValuesStage2<T, false> : SimpleDefaultValues<T> { }; + +// Work out the default value rule for a D3D type (expressed as the C type) and +template <class T, bool normalized> struct DefaultVertexValues : DefaultVertexValuesStage2<T, normalized> { }; +template <bool normalized> struct DefaultVertexValues<float, normalized> : SimpleDefaultValues<float> { }; + +// Policy rules for use with Converter, to choose whether to use the preferred or fallback conversion. +// The fallback conversion produces an output that all D3D9 devices must support. +template <class T> struct UsePreferred { enum { type = T::preferred }; }; +template <class T> struct UseFallback { enum { type = T::fallback }; }; + +// Converter ties it all together. Given an OpenGL type/norm/size and choice of preferred/fallback conversion, +// it provides all the members of the appropriate VertexDataConverter, the D3DCAPS9::DeclTypes flag in cap flag +// and the D3DDECLTYPE member needed for the vertex declaration in declflag. +template <GLenum fromType, bool normalized, int size, template <class T> class PreferenceRule> +struct Converter + : VertexDataConverter<typename GLToCType<fromType>::type, + WidenRule<PreferenceRule< VertexTypeMapping<fromType, normalized> >::type, size>, + ConversionRule<fromType, + normalized, + PreferenceRule< VertexTypeMapping<fromType, normalized> >::type>, + DefaultVertexValues<typename D3DToCType<PreferenceRule< VertexTypeMapping<fromType, normalized> >::type>::type, normalized > > +{ +private: + enum { d3dtype = PreferenceRule< VertexTypeMapping<fromType, normalized> >::type }; + enum { d3dsize = WidenRule<d3dtype, size>::finalWidth }; + +public: + enum { capflag = VertexTypeFlags<d3dtype, d3dsize>::capflag }; + enum { declflag = VertexTypeFlags<d3dtype, d3dsize>::declflag }; +}; + +// Initialize a TranslationInfo +#define TRANSLATION(type, norm, size, preferred) \ + { \ + Converter<type, norm, size, preferred>::identity, \ + Converter<type, norm, size, preferred>::finalSize, \ + Converter<type, norm, size, preferred>::convertArray, \ + static_cast<D3DDECLTYPE>(Converter<type, norm, size, preferred>::declflag) \ + } + +#define TRANSLATION_FOR_TYPE_NORM_SIZE(type, norm, size) \ + { \ + Converter<type, norm, size, UsePreferred>::capflag, \ + TRANSLATION(type, norm, size, UsePreferred), \ + TRANSLATION(type, norm, size, UseFallback) \ + } + +#define TRANSLATIONS_FOR_TYPE(type) \ + { \ + { TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 1), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 2), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 3), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 4) }, \ + { TRANSLATION_FOR_TYPE_NORM_SIZE(type, true, 1), TRANSLATION_FOR_TYPE_NORM_SIZE(type, true, 2), TRANSLATION_FOR_TYPE_NORM_SIZE(type, true, 3), TRANSLATION_FOR_TYPE_NORM_SIZE(type, true, 4) }, \ + } + +#define TRANSLATIONS_FOR_TYPE_NO_NORM(type) \ + { \ + { TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 1), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 2), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 3), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 4) }, \ + { TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 1), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 2), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 3), TRANSLATION_FOR_TYPE_NORM_SIZE(type, false, 4) }, \ + } + +const VertexBuffer9::TranslationDescription VertexBuffer9::mPossibleTranslations[NUM_GL_VERTEX_ATTRIB_TYPES][2][4] = // [GL types as enumerated by typeIndex()][normalized][size-1] +{ + TRANSLATIONS_FOR_TYPE(GL_BYTE), + TRANSLATIONS_FOR_TYPE(GL_UNSIGNED_BYTE), + TRANSLATIONS_FOR_TYPE(GL_SHORT), + TRANSLATIONS_FOR_TYPE(GL_UNSIGNED_SHORT), + TRANSLATIONS_FOR_TYPE_NO_NORM(GL_FIXED), + TRANSLATIONS_FOR_TYPE_NO_NORM(GL_FLOAT) +}; + +void VertexBuffer9::initializeTranslations(DWORD declTypes) +{ + for (unsigned int i = 0; i < NUM_GL_VERTEX_ATTRIB_TYPES; i++) + { + for (unsigned int j = 0; j < 2; j++) + { + for (unsigned int k = 0; k < 4; k++) + { + if (mPossibleTranslations[i][j][k].capsFlag == 0 || (declTypes & mPossibleTranslations[i][j][k].capsFlag) != 0) + { + mFormatConverters[i][j][k] = mPossibleTranslations[i][j][k].preferredConversion; + } + else + { + mFormatConverters[i][j][k] = mPossibleTranslations[i][j][k].fallbackConversion; + } + } + } + } +} + +unsigned int VertexBuffer9::typeIndex(GLenum type) +{ + switch (type) + { + case GL_BYTE: return 0; + case GL_UNSIGNED_BYTE: return 1; + case GL_SHORT: return 2; + case GL_UNSIGNED_SHORT: return 3; + case GL_FIXED: return 4; + case GL_FLOAT: return 5; + + default: UNREACHABLE(); return 5; + } +} + +const VertexBuffer9::FormatConverter &VertexBuffer9::formatConverter(const gl::VertexAttribute &attribute) +{ + return mFormatConverters[typeIndex(attribute.mType)][attribute.mNormalized][attribute.mSize - 1]; +} + +unsigned int VertexBuffer9::spaceRequired(const gl::VertexAttribute &attrib, std::size_t count, GLsizei instances) +{ + unsigned int elementSize = formatConverter(attrib).outputElementSize; + + if (instances == 0 || attrib.mDivisor == 0) + { + return elementSize * count; + } + else + { + return elementSize * ((instances + attrib.mDivisor - 1) / attrib.mDivisor); + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer9.h b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer9.h new file mode 100644 index 0000000000..f771635bda --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexBuffer9.h @@ -0,0 +1,90 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexBuffer9.h: Defines the D3D9 VertexBuffer implementation. + +#ifndef LIBGLESV2_RENDERER_VERTEXBUFFER9_H_ +#define LIBGLESV2_RENDERER_VERTEXBUFFER9_H_ + +#include "libGLESv2/renderer/VertexBuffer.h" + +namespace rx +{ +class Renderer9; + +class VertexBuffer9 : public VertexBuffer +{ + public: + explicit VertexBuffer9(rx::Renderer9 *const renderer); + virtual ~VertexBuffer9(); + + virtual bool initialize(unsigned int size, bool dynamicUsage); + + static VertexBuffer9 *makeVertexBuffer9(VertexBuffer *vertexBuffer); + + virtual bool storeVertexAttributes(const gl::VertexAttribute &attrib, GLint start, GLsizei count, GLsizei instances, + unsigned int offset); + virtual bool storeRawData(const void* data, unsigned int size, unsigned int offset); + + virtual unsigned int getSpaceRequired(const gl::VertexAttribute &attrib, GLsizei count, GLsizei instances) const; + + virtual bool requiresConversion(const gl::VertexAttribute &attrib) const; + + unsigned int getVertexSize(const gl::VertexAttribute &attrib) const; + D3DDECLTYPE getDeclType(const gl::VertexAttribute &attrib) const; + + virtual unsigned int getBufferSize() const; + virtual bool setBufferSize(unsigned int size); + virtual bool discard(); + + IDirect3DVertexBuffer9 *getBuffer() const; + + private: + DISALLOW_COPY_AND_ASSIGN(VertexBuffer9); + + rx::Renderer9 *const mRenderer; + + IDirect3DVertexBuffer9 *mVertexBuffer; + unsigned int mBufferSize; + bool mDynamicUsage; + + // Attribute format conversion + enum { NUM_GL_VERTEX_ATTRIB_TYPES = 6 }; + + struct FormatConverter + { + bool identity; + std::size_t outputElementSize; + void (*convertArray)(const void *in, std::size_t stride, std::size_t n, void *out); + D3DDECLTYPE d3dDeclType; + }; + + static bool mTranslationsInitialized; + static void initializeTranslations(DWORD declTypes); + + // [GL types as enumerated by typeIndex()][normalized][size - 1] + static FormatConverter mFormatConverters[NUM_GL_VERTEX_ATTRIB_TYPES][2][4]; + + struct TranslationDescription + { + DWORD capsFlag; + FormatConverter preferredConversion; + FormatConverter fallbackConversion; + }; + + // This table is used to generate mFormatConverters. + // [GL types as enumerated by typeIndex()][normalized][size - 1] + static const TranslationDescription mPossibleTranslations[NUM_GL_VERTEX_ATTRIB_TYPES][2][4]; + + static unsigned int typeIndex(GLenum type); + static const FormatConverter &formatConverter(const gl::VertexAttribute &attribute); + + static unsigned int spaceRequired(const gl::VertexAttribute &attrib, std::size_t count, GLsizei instances); +}; + +} + +#endif // LIBGLESV2_RENDERER_VERTEXBUFFER9_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexDataManager.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/VertexDataManager.cpp new file mode 100644 index 0000000000..ec85857264 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexDataManager.cpp @@ -0,0 +1,258 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexDataManager.h: Defines the VertexDataManager, a class that +// runs the Buffer translation process. + +#include "libGLESv2/renderer/VertexDataManager.h" +#include "libGLESv2/renderer/BufferStorage.h" + +#include "libGLESv2/Buffer.h" +#include "libGLESv2/ProgramBinary.h" +#include "libGLESv2/Context.h" +#include "libGLESv2/renderer/VertexBuffer.h" + +namespace +{ + enum { INITIAL_STREAM_BUFFER_SIZE = 1024*1024 }; + // This has to be at least 4k or else it fails on ATI cards. + enum { CONSTANT_VERTEX_BUFFER_SIZE = 4096 }; +} + +namespace rx +{ + +static int elementsInBuffer(const gl::VertexAttribute &attribute, int size) +{ + int stride = attribute.stride(); + return (size - attribute.mOffset % stride + (stride - attribute.typeSize())) / stride; +} + +VertexDataManager::VertexDataManager(Renderer *renderer) : mRenderer(renderer) +{ + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + mCurrentValue[i][0] = std::numeric_limits<float>::quiet_NaN(); + mCurrentValue[i][1] = std::numeric_limits<float>::quiet_NaN(); + mCurrentValue[i][2] = std::numeric_limits<float>::quiet_NaN(); + mCurrentValue[i][3] = std::numeric_limits<float>::quiet_NaN(); + mCurrentValueBuffer[i] = NULL; + mCurrentValueOffsets[i] = 0; + } + + mStreamingBuffer = new StreamingVertexBufferInterface(renderer, INITIAL_STREAM_BUFFER_SIZE); + + if (!mStreamingBuffer) + { + ERR("Failed to allocate the streaming vertex buffer."); + } +} + +VertexDataManager::~VertexDataManager() +{ + delete mStreamingBuffer; + + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + delete mCurrentValueBuffer[i]; + } +} + +static bool directStoragePossible(VertexBufferInterface* vb, const gl::VertexAttribute& attrib) +{ + gl::Buffer *buffer = attrib.mBoundBuffer.get(); + BufferStorage *storage = buffer ? buffer->getStorage() : NULL; + + const bool isAligned = (attrib.stride() % 4 == 0) && (attrib.mOffset % 4 == 0); + + return storage && storage->supportsDirectBinding() && !vb->getVertexBuffer()->requiresConversion(attrib) && isAligned; +} + +GLenum VertexDataManager::prepareVertexData(const gl::VertexAttribute attribs[], gl::ProgramBinary *programBinary, GLint start, GLsizei count, TranslatedAttribute *translated, GLsizei instances) +{ + if (!mStreamingBuffer) + { + return GL_OUT_OF_MEMORY; + } + + for (int attributeIndex = 0; attributeIndex < gl::MAX_VERTEX_ATTRIBS; attributeIndex++) + { + translated[attributeIndex].active = (programBinary->getSemanticIndex(attributeIndex) != -1); + } + + // Invalidate static buffers that don't contain matching attributes + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + if (translated[i].active && attribs[i].mArrayEnabled) + { + gl::Buffer *buffer = attribs[i].mBoundBuffer.get(); + StaticVertexBufferInterface *staticBuffer = buffer ? buffer->getStaticVertexBuffer() : NULL; + + if (staticBuffer && staticBuffer->getBufferSize() > 0 && staticBuffer->lookupAttribute(attribs[i]) == -1 && + !directStoragePossible(staticBuffer, attribs[i])) + { + buffer->invalidateStaticData(); + } + } + } + + // Reserve the required space in the buffers + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + if (translated[i].active && attribs[i].mArrayEnabled) + { + gl::Buffer *buffer = attribs[i].mBoundBuffer.get(); + StaticVertexBufferInterface *staticBuffer = buffer ? buffer->getStaticVertexBuffer() : NULL; + VertexBufferInterface *vertexBuffer = staticBuffer ? staticBuffer : static_cast<VertexBufferInterface*>(mStreamingBuffer); + + if (!directStoragePossible(vertexBuffer, attribs[i])) + { + if (staticBuffer) + { + if (staticBuffer->getBufferSize() == 0) + { + int totalCount = elementsInBuffer(attribs[i], buffer->size()); + staticBuffer->reserveVertexSpace(attribs[i], totalCount, 0); + } + } + else + { + mStreamingBuffer->reserveVertexSpace(attribs[i], count, instances); + } + } + } + } + + // Perform the vertex data translations + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + if (translated[i].active) + { + if (attribs[i].mArrayEnabled) + { + gl::Buffer *buffer = attribs[i].mBoundBuffer.get(); + + if (!buffer && attribs[i].mPointer == NULL) + { + // This is an application error that would normally result in a crash, but we catch it and return an error + ERR("An enabled vertex array has no buffer and no pointer."); + return GL_INVALID_OPERATION; + } + + StaticVertexBufferInterface *staticBuffer = buffer ? buffer->getStaticVertexBuffer() : NULL; + VertexBufferInterface *vertexBuffer = staticBuffer ? staticBuffer : static_cast<VertexBufferInterface*>(mStreamingBuffer); + + BufferStorage *storage = buffer ? buffer->getStorage() : NULL; + bool directStorage = directStoragePossible(vertexBuffer, attribs[i]); + + std::size_t streamOffset = -1; + unsigned int outputElementSize = 0; + + if (directStorage) + { + outputElementSize = attribs[i].stride(); + streamOffset = attribs[i].mOffset + outputElementSize * start; + storage->markBufferUsage(); + } + else if (staticBuffer) + { + streamOffset = staticBuffer->lookupAttribute(attribs[i]); + outputElementSize = staticBuffer->getVertexBuffer()->getSpaceRequired(attribs[i], 1, 0); + + if (streamOffset == -1) + { + // Convert the entire buffer + int totalCount = elementsInBuffer(attribs[i], storage->getSize()); + int startIndex = attribs[i].mOffset / attribs[i].stride(); + + streamOffset = staticBuffer->storeVertexAttributes(attribs[i], -startIndex, totalCount, 0); + } + + if (streamOffset != -1) + { + streamOffset += (attribs[i].mOffset / attribs[i].stride()) * outputElementSize; + + if (instances == 0 || attribs[i].mDivisor == 0) + { + streamOffset += start * outputElementSize; + } + } + } + else + { + outputElementSize = mStreamingBuffer->getVertexBuffer()->getSpaceRequired(attribs[i], 1, 0); + streamOffset = mStreamingBuffer->storeVertexAttributes(attribs[i], start, count, instances); + } + + if (streamOffset == -1) + { + return GL_OUT_OF_MEMORY; + } + + translated[i].storage = directStorage ? storage : NULL; + translated[i].vertexBuffer = vertexBuffer->getVertexBuffer(); + translated[i].serial = directStorage ? storage->getSerial() : vertexBuffer->getSerial(); + translated[i].divisor = attribs[i].mDivisor; + + translated[i].attribute = &attribs[i]; + translated[i].stride = outputElementSize; + translated[i].offset = streamOffset; + } + else + { + if (!mCurrentValueBuffer[i]) + { + mCurrentValueBuffer[i] = new StreamingVertexBufferInterface(mRenderer, CONSTANT_VERTEX_BUFFER_SIZE); + } + + StreamingVertexBufferInterface *buffer = mCurrentValueBuffer[i]; + + if (mCurrentValue[i][0] != attribs[i].mCurrentValue[0] || + mCurrentValue[i][1] != attribs[i].mCurrentValue[1] || + mCurrentValue[i][2] != attribs[i].mCurrentValue[2] || + mCurrentValue[i][3] != attribs[i].mCurrentValue[3]) + { + unsigned int requiredSpace = sizeof(float) * 4; + buffer->reserveRawDataSpace(requiredSpace); + int streamOffset = buffer->storeRawData(attribs[i].mCurrentValue, requiredSpace); + if (streamOffset == -1) + { + return GL_OUT_OF_MEMORY; + } + + mCurrentValueOffsets[i] = streamOffset; + } + + translated[i].storage = NULL; + translated[i].vertexBuffer = mCurrentValueBuffer[i]->getVertexBuffer(); + translated[i].serial = mCurrentValueBuffer[i]->getSerial(); + translated[i].divisor = 0; + + translated[i].attribute = &attribs[i]; + translated[i].stride = 0; + translated[i].offset = mCurrentValueOffsets[i]; + } + } + } + + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + if (translated[i].active && attribs[i].mArrayEnabled) + { + gl::Buffer *buffer = attribs[i].mBoundBuffer.get(); + + if (buffer) + { + buffer->promoteStaticUsage(count * attribs[i].typeSize()); + } + } + } + + return GL_NO_ERROR; +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexDataManager.h b/src/3rdparty/angle/src/libGLESv2/renderer/VertexDataManager.h new file mode 100644 index 0000000000..28387e6baf --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexDataManager.h @@ -0,0 +1,65 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexDataManager.h: Defines the VertexDataManager, a class that +// runs the Buffer translation process. + +#ifndef LIBGLESV2_RENDERER_VERTEXDATAMANAGER_H_ +#define LIBGLESV2_RENDERER_VERTEXDATAMANAGER_H_ + +#include "libGLESv2/Constants.h" +#include "common/angleutils.h" + +namespace gl +{ +class VertexAttribute; +class ProgramBinary; +} + +namespace rx +{ +class BufferStorage; +class StreamingVertexBufferInterface; +class VertexBuffer; +class Renderer; + +struct TranslatedAttribute +{ + bool active; + + const gl::VertexAttribute *attribute; + UINT offset; + UINT stride; // 0 means not to advance the read pointer at all + + VertexBuffer *vertexBuffer; + BufferStorage *storage; + unsigned int serial; + unsigned int divisor; +}; + +class VertexDataManager +{ + public: + VertexDataManager(rx::Renderer *renderer); + virtual ~VertexDataManager(); + + GLenum prepareVertexData(const gl::VertexAttribute attribs[], gl::ProgramBinary *programBinary, GLint start, GLsizei count, TranslatedAttribute *outAttribs, GLsizei instances); + + private: + DISALLOW_COPY_AND_ASSIGN(VertexDataManager); + + rx::Renderer *const mRenderer; + + StreamingVertexBufferInterface *mStreamingBuffer; + + float mCurrentValue[gl::MAX_VERTEX_ATTRIBS][4]; + StreamingVertexBufferInterface *mCurrentValueBuffer[gl::MAX_VERTEX_ATTRIBS]; + std::size_t mCurrentValueOffsets[gl::MAX_VERTEX_ATTRIBS]; +}; + +} + +#endif // LIBGLESV2_RENDERER_VERTEXDATAMANAGER_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexDeclarationCache.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/VertexDeclarationCache.cpp new file mode 100644 index 0000000000..9b83a6476e --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexDeclarationCache.cpp @@ -0,0 +1,217 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexDeclarationCache.cpp: Implements a helper class to construct and cache vertex declarations. + +#include "libGLESv2/ProgramBinary.h" +#include "libGLESv2/Context.h" +#include "libGLESv2/renderer/VertexBuffer9.h" +#include "libGLESv2/renderer/VertexDeclarationCache.h" + +namespace rx +{ + +VertexDeclarationCache::VertexDeclarationCache() : mMaxLru(0) +{ + for (int i = 0; i < NUM_VERTEX_DECL_CACHE_ENTRIES; i++) + { + mVertexDeclCache[i].vertexDeclaration = NULL; + mVertexDeclCache[i].lruCount = 0; + } + + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + mAppliedVBs[i].serial = 0; + } + + mLastSetVDecl = NULL; + mInstancingEnabled = true; +} + +VertexDeclarationCache::~VertexDeclarationCache() +{ + for (int i = 0; i < NUM_VERTEX_DECL_CACHE_ENTRIES; i++) + { + if (mVertexDeclCache[i].vertexDeclaration) + { + mVertexDeclCache[i].vertexDeclaration->Release(); + } + } +} + +GLenum VertexDeclarationCache::applyDeclaration(IDirect3DDevice9 *device, TranslatedAttribute attributes[], gl::ProgramBinary *programBinary, GLsizei instances, GLsizei *repeatDraw) +{ + *repeatDraw = 1; + + int indexedAttribute = gl::MAX_VERTEX_ATTRIBS; + int instancedAttribute = gl::MAX_VERTEX_ATTRIBS; + + if (instances > 0) + { + // Find an indexed attribute to be mapped to D3D stream 0 + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + if (attributes[i].active) + { + if (indexedAttribute == gl::MAX_VERTEX_ATTRIBS && attributes[i].divisor == 0) + { + indexedAttribute = i; + } + else if (instancedAttribute == gl::MAX_VERTEX_ATTRIBS && attributes[i].divisor != 0) + { + instancedAttribute = i; + } + if (indexedAttribute != gl::MAX_VERTEX_ATTRIBS && instancedAttribute != gl::MAX_VERTEX_ATTRIBS) + break; // Found both an indexed and instanced attribute + } + } + + if (indexedAttribute == gl::MAX_VERTEX_ATTRIBS) + { + return GL_INVALID_OPERATION; + } + } + + D3DVERTEXELEMENT9 elements[gl::MAX_VERTEX_ATTRIBS + 1]; + D3DVERTEXELEMENT9 *element = &elements[0]; + + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + if (attributes[i].active) + { + // Directly binding the storage buffer is not supported for d3d9 + ASSERT(attributes[i].storage == NULL); + + int stream = i; + + if (instances > 0) + { + // Due to a bug on ATI cards we can't enable instancing when none of the attributes are instanced. + if (instancedAttribute == gl::MAX_VERTEX_ATTRIBS) + { + *repeatDraw = instances; + } + else + { + if (i == indexedAttribute) + { + stream = 0; + } + else if (i == 0) + { + stream = indexedAttribute; + } + + UINT frequency = 1; + + if (attributes[i].divisor == 0) + { + frequency = D3DSTREAMSOURCE_INDEXEDDATA | instances; + } + else + { + frequency = D3DSTREAMSOURCE_INSTANCEDATA | attributes[i].divisor; + } + + device->SetStreamSourceFreq(stream, frequency); + mInstancingEnabled = true; + } + } + + VertexBuffer9 *vertexBuffer = VertexBuffer9::makeVertexBuffer9(attributes[i].vertexBuffer); + + if (mAppliedVBs[stream].serial != attributes[i].serial || + mAppliedVBs[stream].stride != attributes[i].stride || + mAppliedVBs[stream].offset != attributes[i].offset) + { + device->SetStreamSource(stream, vertexBuffer->getBuffer(), attributes[i].offset, attributes[i].stride); + mAppliedVBs[stream].serial = attributes[i].serial; + mAppliedVBs[stream].stride = attributes[i].stride; + mAppliedVBs[stream].offset = attributes[i].offset; + } + + element->Stream = stream; + element->Offset = 0; + element->Type = attributes[i].attribute->mArrayEnabled ? vertexBuffer->getDeclType(*attributes[i].attribute) : D3DDECLTYPE_FLOAT4; + element->Method = D3DDECLMETHOD_DEFAULT; + element->Usage = D3DDECLUSAGE_TEXCOORD; + element->UsageIndex = programBinary->getSemanticIndex(i); + element++; + } + } + + if (instances == 0 || instancedAttribute == gl::MAX_VERTEX_ATTRIBS) + { + if (mInstancingEnabled) + { + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + device->SetStreamSourceFreq(i, 1); + } + + mInstancingEnabled = false; + } + } + + static const D3DVERTEXELEMENT9 end = D3DDECL_END(); + *(element++) = end; + + for (int i = 0; i < NUM_VERTEX_DECL_CACHE_ENTRIES; i++) + { + VertexDeclCacheEntry *entry = &mVertexDeclCache[i]; + if (memcmp(entry->cachedElements, elements, (element - elements) * sizeof(D3DVERTEXELEMENT9)) == 0 && entry->vertexDeclaration) + { + entry->lruCount = ++mMaxLru; + if(entry->vertexDeclaration != mLastSetVDecl) + { + device->SetVertexDeclaration(entry->vertexDeclaration); + mLastSetVDecl = entry->vertexDeclaration; + } + + return GL_NO_ERROR; + } + } + + VertexDeclCacheEntry *lastCache = mVertexDeclCache; + + for (int i = 0; i < NUM_VERTEX_DECL_CACHE_ENTRIES; i++) + { + if (mVertexDeclCache[i].lruCount < lastCache->lruCount) + { + lastCache = &mVertexDeclCache[i]; + } + } + + if (lastCache->vertexDeclaration != NULL) + { + lastCache->vertexDeclaration->Release(); + lastCache->vertexDeclaration = NULL; + // mLastSetVDecl is set to the replacement, so we don't have to worry + // about it. + } + + memcpy(lastCache->cachedElements, elements, (element - elements) * sizeof(D3DVERTEXELEMENT9)); + device->CreateVertexDeclaration(elements, &lastCache->vertexDeclaration); + device->SetVertexDeclaration(lastCache->vertexDeclaration); + mLastSetVDecl = lastCache->vertexDeclaration; + lastCache->lruCount = ++mMaxLru; + + return GL_NO_ERROR; +} + +void VertexDeclarationCache::markStateDirty() +{ + for (int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + mAppliedVBs[i].serial = 0; + } + + mLastSetVDecl = NULL; + mInstancingEnabled = true; // Forces it to be disabled when not used +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/VertexDeclarationCache.h b/src/3rdparty/angle/src/libGLESv2/renderer/VertexDeclarationCache.h new file mode 100644 index 0000000000..3fc024a9ba --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/VertexDeclarationCache.h @@ -0,0 +1,58 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexDeclarationCache.h: Defines a helper class to construct and cache vertex declarations. + +#ifndef LIBGLESV2_RENDERER_VERTEXDECLARATIONCACHE_H_ +#define LIBGLESV2_RENDERER_VERTEXDECLARATIONCACHE_H_ + +#include "libGLESv2/renderer/VertexDataManager.h" + +namespace gl +{ +class VertexDataManager; +} + +namespace rx +{ + +class VertexDeclarationCache +{ + public: + VertexDeclarationCache(); + ~VertexDeclarationCache(); + + GLenum applyDeclaration(IDirect3DDevice9 *device, TranslatedAttribute attributes[], gl::ProgramBinary *programBinary, GLsizei instances, GLsizei *repeatDraw); + + void markStateDirty(); + + private: + UINT mMaxLru; + + enum { NUM_VERTEX_DECL_CACHE_ENTRIES = 32 }; + + struct VBData + { + unsigned int serial; + unsigned int stride; + unsigned int offset; + }; + + VBData mAppliedVBs[gl::MAX_VERTEX_ATTRIBS]; + IDirect3DVertexDeclaration9 *mLastSetVDecl; + bool mInstancingEnabled; + + struct VertexDeclCacheEntry + { + D3DVERTEXELEMENT9 cachedElements[gl::MAX_VERTEX_ATTRIBS + 1]; + UINT lruCount; + IDirect3DVertexDeclaration9 *vertexDeclaration; + } mVertexDeclCache[NUM_VERTEX_DECL_CACHE_ENTRIES]; +}; + +} + +#endif // LIBGLESV2_RENDERER_VERTEXDECLARATIONCACHE_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/generatemip.h b/src/3rdparty/angle/src/libGLESv2/renderer/generatemip.h new file mode 100644 index 0000000000..8e1973605b --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/generatemip.h @@ -0,0 +1,203 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// generatemip.h: Defines the GenerateMip function, templated on the format +// type of the image for which mip levels are being generated. + +#ifndef LIBGLESV2_RENDERER_GENERATEMIP_H_ +#define LIBGLESV2_RENDERER_GENERATEMIP_H_ + +#include "libGLESv2/mathutil.h" + +namespace rx +{ +struct L8 +{ + unsigned char L; + + static void average(L8 *dst, const L8 *src1, const L8 *src2) + { + dst->L = ((src1->L ^ src2->L) >> 1) + (src1->L & src2->L); + } +}; + +typedef L8 R8; // R8 type is functionally equivalent for mip purposes +typedef L8 A8; // A8 type is functionally equivalent for mip purposes + +struct A8L8 +{ + unsigned char L; + unsigned char A; + + static void average(A8L8 *dst, const A8L8 *src1, const A8L8 *src2) + { + *(unsigned short*)dst = (((*(unsigned short*)src1 ^ *(unsigned short*)src2) & 0xFEFE) >> 1) + (*(unsigned short*)src1 & *(unsigned short*)src2); + } +}; + +typedef A8L8 R8G8; // R8G8 type is functionally equivalent for mip purposes + +struct A8R8G8B8 +{ + unsigned char B; + unsigned char G; + unsigned char R; + unsigned char A; + + static void average(A8R8G8B8 *dst, const A8R8G8B8 *src1, const A8R8G8B8 *src2) + { + *(unsigned int*)dst = (((*(unsigned int*)src1 ^ *(unsigned int*)src2) & 0xFEFEFEFE) >> 1) + (*(unsigned int*)src1 & *(unsigned int*)src2); + } +}; + +typedef A8R8G8B8 R8G8B8A8; // R8G8B8A8 type is functionally equivalent for mip purposes + +struct A16B16G16R16F +{ + unsigned short R; + unsigned short G; + unsigned short B; + unsigned short A; + + static void average(A16B16G16R16F *dst, const A16B16G16R16F *src1, const A16B16G16R16F *src2) + { + dst->R = gl::float32ToFloat16((gl::float16ToFloat32(src1->R) + gl::float16ToFloat32(src2->R)) * 0.5f); + dst->G = gl::float32ToFloat16((gl::float16ToFloat32(src1->G) + gl::float16ToFloat32(src2->G)) * 0.5f); + dst->B = gl::float32ToFloat16((gl::float16ToFloat32(src1->B) + gl::float16ToFloat32(src2->B)) * 0.5f); + dst->A = gl::float32ToFloat16((gl::float16ToFloat32(src1->A) + gl::float16ToFloat32(src2->A)) * 0.5f); + } +}; + +struct R16F +{ + unsigned short R; + + static void average(R16F *dst, const R16F *src1, const R16F *src2) + { + dst->R = gl::float32ToFloat16((gl::float16ToFloat32(src1->R) + gl::float16ToFloat32(src2->R)) * 0.5f); + } +}; + +struct R16G16F +{ + unsigned short R; + unsigned short G; + + static void average(R16G16F *dst, const R16G16F *src1, const R16G16F *src2) + { + dst->R = gl::float32ToFloat16((gl::float16ToFloat32(src1->R) + gl::float16ToFloat32(src2->R)) * 0.5f); + dst->G = gl::float32ToFloat16((gl::float16ToFloat32(src1->G) + gl::float16ToFloat32(src2->G)) * 0.5f); + } +}; + +struct A32B32G32R32F +{ + float R; + float G; + float B; + float A; + + static void average(A32B32G32R32F *dst, const A32B32G32R32F *src1, const A32B32G32R32F *src2) + { + dst->R = (src1->R + src2->R) * 0.5f; + dst->G = (src1->G + src2->G) * 0.5f; + dst->B = (src1->B + src2->B) * 0.5f; + dst->A = (src1->A + src2->A) * 0.5f; + } +}; + +struct R32F +{ + float R; + + static void average(R32F *dst, const R32F *src1, const R32F *src2) + { + dst->R = (src1->R + src2->R) * 0.5f; + } +}; + +struct R32G32F +{ + float R; + float G; + + static void average(R32G32F *dst, const R32G32F *src1, const R32G32F *src2) + { + dst->R = (src1->R + src2->R) * 0.5f; + dst->G = (src1->G + src2->G) * 0.5f; + } +}; + +struct R32G32B32F +{ + float R; + float G; + float B; + + static void average(R32G32B32F *dst, const R32G32B32F *src1, const R32G32B32F *src2) + { + dst->R = (src1->R + src2->R) * 0.5f; + dst->G = (src1->G + src2->G) * 0.5f; + dst->B = (src1->B + src2->B) * 0.5f; + } +}; + +template <typename T> +static void GenerateMip(unsigned int sourceWidth, unsigned int sourceHeight, + const unsigned char *sourceData, int sourcePitch, + unsigned char *destData, int destPitch) +{ + unsigned int mipWidth = std::max(1U, sourceWidth >> 1); + unsigned int mipHeight = std::max(1U, sourceHeight >> 1); + + if (sourceHeight == 1) + { + ASSERT(sourceWidth != 1); + + const T *src = (const T*)sourceData; + T *dst = (T*)destData; + + for (unsigned int x = 0; x < mipWidth; x++) + { + T::average(&dst[x], &src[x * 2], &src[x * 2 + 1]); + } + } + else if (sourceWidth == 1) + { + ASSERT(sourceHeight != 1); + + for (unsigned int y = 0; y < mipHeight; y++) + { + const T *src0 = (const T*)(sourceData + y * 2 * sourcePitch); + const T *src1 = (const T*)(sourceData + y * 2 * sourcePitch + sourcePitch); + T *dst = (T*)(destData + y * destPitch); + + T::average(dst, src0, src1); + } + } + else + { + for (unsigned int y = 0; y < mipHeight; y++) + { + const T *src0 = (const T*)(sourceData + y * 2 * sourcePitch); + const T *src1 = (const T*)(sourceData + y * 2 * sourcePitch + sourcePitch); + T *dst = (T*)(destData + y * destPitch); + + for (unsigned int x = 0; x < mipWidth; x++) + { + T tmp0; + T tmp1; + + T::average(&tmp0, &src0[x * 2], &src0[x * 2 + 1]); + T::average(&tmp1, &src1[x * 2], &src1[x * 2 + 1]); + T::average(&dst[x], &tmp0, &tmp1); + } + } + } +} +} + +#endif // LIBGLESV2_RENDERER_GENERATEMIP_H_ diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/renderer11_utils.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/renderer11_utils.cpp new file mode 100644 index 0000000000..5f01dc12ed --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/renderer11_utils.cpp @@ -0,0 +1,688 @@ +#include "precompiled.h" +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// renderer11_utils.cpp: Conversion functions and other utility routines +// specific to the D3D11 renderer. + +#include "libGLESv2/renderer/renderer11_utils.h" + +#include "common/debug.h" + +namespace gl_d3d11 +{ + +D3D11_BLEND ConvertBlendFunc(GLenum glBlend, bool isAlpha) +{ + D3D11_BLEND d3dBlend = D3D11_BLEND_ZERO; + + switch (glBlend) + { + case GL_ZERO: d3dBlend = D3D11_BLEND_ZERO; break; + case GL_ONE: d3dBlend = D3D11_BLEND_ONE; break; + case GL_SRC_COLOR: d3dBlend = (isAlpha ? D3D11_BLEND_SRC_ALPHA : D3D11_BLEND_SRC_COLOR); break; + case GL_ONE_MINUS_SRC_COLOR: d3dBlend = (isAlpha ? D3D11_BLEND_INV_SRC_ALPHA : D3D11_BLEND_INV_SRC_COLOR); break; + case GL_DST_COLOR: d3dBlend = (isAlpha ? D3D11_BLEND_DEST_ALPHA : D3D11_BLEND_DEST_COLOR); break; + case GL_ONE_MINUS_DST_COLOR: d3dBlend = (isAlpha ? D3D11_BLEND_INV_DEST_ALPHA : D3D11_BLEND_INV_DEST_COLOR); break; + case GL_SRC_ALPHA: d3dBlend = D3D11_BLEND_SRC_ALPHA; break; + case GL_ONE_MINUS_SRC_ALPHA: d3dBlend = D3D11_BLEND_INV_SRC_ALPHA; break; + case GL_DST_ALPHA: d3dBlend = D3D11_BLEND_DEST_ALPHA; break; + case GL_ONE_MINUS_DST_ALPHA: d3dBlend = D3D11_BLEND_INV_DEST_ALPHA; break; + case GL_CONSTANT_COLOR: d3dBlend = D3D11_BLEND_BLEND_FACTOR; break; + case GL_ONE_MINUS_CONSTANT_COLOR: d3dBlend = D3D11_BLEND_INV_BLEND_FACTOR; break; + case GL_CONSTANT_ALPHA: d3dBlend = D3D11_BLEND_BLEND_FACTOR; break; + case GL_ONE_MINUS_CONSTANT_ALPHA: d3dBlend = D3D11_BLEND_INV_BLEND_FACTOR; break; + case GL_SRC_ALPHA_SATURATE: d3dBlend = D3D11_BLEND_SRC_ALPHA_SAT; break; + default: UNREACHABLE(); + } + + return d3dBlend; +} + +D3D11_BLEND_OP ConvertBlendOp(GLenum glBlendOp) +{ + D3D11_BLEND_OP d3dBlendOp = D3D11_BLEND_OP_ADD; + + switch (glBlendOp) + { + case GL_FUNC_ADD: d3dBlendOp = D3D11_BLEND_OP_ADD; break; + case GL_FUNC_SUBTRACT: d3dBlendOp = D3D11_BLEND_OP_SUBTRACT; break; + case GL_FUNC_REVERSE_SUBTRACT: d3dBlendOp = D3D11_BLEND_OP_REV_SUBTRACT; break; + default: UNREACHABLE(); + } + + return d3dBlendOp; +} + +UINT8 ConvertColorMask(bool red, bool green, bool blue, bool alpha) +{ + UINT8 mask = 0; + if (red) + { + mask |= D3D11_COLOR_WRITE_ENABLE_RED; + } + if (green) + { + mask |= D3D11_COLOR_WRITE_ENABLE_GREEN; + } + if (blue) + { + mask |= D3D11_COLOR_WRITE_ENABLE_BLUE; + } + if (alpha) + { + mask |= D3D11_COLOR_WRITE_ENABLE_ALPHA; + } + return mask; +} + +D3D11_CULL_MODE ConvertCullMode(bool cullEnabled, GLenum cullMode) +{ + D3D11_CULL_MODE cull = D3D11_CULL_NONE; + + if (cullEnabled) + { + switch (cullMode) + { + case GL_FRONT: cull = D3D11_CULL_FRONT; break; + case GL_BACK: cull = D3D11_CULL_BACK; break; + case GL_FRONT_AND_BACK: cull = D3D11_CULL_NONE; break; + default: UNREACHABLE(); + } + } + else + { + cull = D3D11_CULL_NONE; + } + + return cull; +} + +D3D11_COMPARISON_FUNC ConvertComparison(GLenum comparison) +{ + D3D11_COMPARISON_FUNC d3dComp = D3D11_COMPARISON_NEVER; + switch (comparison) + { + case GL_NEVER: d3dComp = D3D11_COMPARISON_NEVER; break; + case GL_ALWAYS: d3dComp = D3D11_COMPARISON_ALWAYS; break; + case GL_LESS: d3dComp = D3D11_COMPARISON_LESS; break; + case GL_LEQUAL: d3dComp = D3D11_COMPARISON_LESS_EQUAL; break; + case GL_EQUAL: d3dComp = D3D11_COMPARISON_EQUAL; break; + case GL_GREATER: d3dComp = D3D11_COMPARISON_GREATER; break; + case GL_GEQUAL: d3dComp = D3D11_COMPARISON_GREATER_EQUAL; break; + case GL_NOTEQUAL: d3dComp = D3D11_COMPARISON_NOT_EQUAL; break; + default: UNREACHABLE(); + } + + return d3dComp; +} + +D3D11_DEPTH_WRITE_MASK ConvertDepthMask(bool depthWriteEnabled) +{ + return depthWriteEnabled ? D3D11_DEPTH_WRITE_MASK_ALL : D3D11_DEPTH_WRITE_MASK_ZERO; +} + +UINT8 ConvertStencilMask(GLuint stencilmask) +{ + return static_cast<UINT8>(stencilmask); +} + +D3D11_STENCIL_OP ConvertStencilOp(GLenum stencilOp) +{ + D3D11_STENCIL_OP d3dStencilOp = D3D11_STENCIL_OP_KEEP; + + switch (stencilOp) + { + case GL_ZERO: d3dStencilOp = D3D11_STENCIL_OP_ZERO; break; + case GL_KEEP: d3dStencilOp = D3D11_STENCIL_OP_KEEP; break; + case GL_REPLACE: d3dStencilOp = D3D11_STENCIL_OP_REPLACE; break; + case GL_INCR: d3dStencilOp = D3D11_STENCIL_OP_INCR_SAT; break; + case GL_DECR: d3dStencilOp = D3D11_STENCIL_OP_DECR_SAT; break; + case GL_INVERT: d3dStencilOp = D3D11_STENCIL_OP_INVERT; break; + case GL_INCR_WRAP: d3dStencilOp = D3D11_STENCIL_OP_INCR; break; + case GL_DECR_WRAP: d3dStencilOp = D3D11_STENCIL_OP_DECR; break; + default: UNREACHABLE(); + } + + return d3dStencilOp; +} + +D3D11_FILTER ConvertFilter(GLenum minFilter, GLenum magFilter, float maxAnisotropy) +{ + if (maxAnisotropy > 1.0f) + { + return D3D11_ENCODE_ANISOTROPIC_FILTER(false); + } + else + { + D3D11_FILTER_TYPE dxMin = D3D11_FILTER_TYPE_POINT; + D3D11_FILTER_TYPE dxMip = D3D11_FILTER_TYPE_POINT; + switch (minFilter) + { + case GL_NEAREST: dxMin = D3D11_FILTER_TYPE_POINT; dxMip = D3D11_FILTER_TYPE_POINT; break; + case GL_LINEAR: dxMin = D3D11_FILTER_TYPE_LINEAR; dxMip = D3D11_FILTER_TYPE_POINT; break; + case GL_NEAREST_MIPMAP_NEAREST: dxMin = D3D11_FILTER_TYPE_POINT; dxMip = D3D11_FILTER_TYPE_POINT; break; + case GL_LINEAR_MIPMAP_NEAREST: dxMin = D3D11_FILTER_TYPE_LINEAR; dxMip = D3D11_FILTER_TYPE_POINT; break; + case GL_NEAREST_MIPMAP_LINEAR: dxMin = D3D11_FILTER_TYPE_POINT; dxMip = D3D11_FILTER_TYPE_LINEAR; break; + case GL_LINEAR_MIPMAP_LINEAR: dxMin = D3D11_FILTER_TYPE_LINEAR; dxMip = D3D11_FILTER_TYPE_LINEAR; break; + default: UNREACHABLE(); + } + + D3D11_FILTER_TYPE dxMag = D3D11_FILTER_TYPE_POINT; + switch (magFilter) + { + case GL_NEAREST: dxMag = D3D11_FILTER_TYPE_POINT; break; + case GL_LINEAR: dxMag = D3D11_FILTER_TYPE_LINEAR; break; + default: UNREACHABLE(); + } + + return D3D11_ENCODE_BASIC_FILTER(dxMin, dxMag, dxMip, false); + } +} + +D3D11_TEXTURE_ADDRESS_MODE ConvertTextureWrap(GLenum wrap) +{ + switch (wrap) + { + case GL_REPEAT: return D3D11_TEXTURE_ADDRESS_WRAP; + case GL_CLAMP_TO_EDGE: return D3D11_TEXTURE_ADDRESS_CLAMP; + case GL_MIRRORED_REPEAT: return D3D11_TEXTURE_ADDRESS_MIRROR; + default: UNREACHABLE(); + } + + return D3D11_TEXTURE_ADDRESS_WRAP; +} + +FLOAT ConvertMinLOD(GLenum minFilter, unsigned int lodOffset) +{ + return (minFilter == GL_NEAREST || minFilter == GL_LINEAR) ? static_cast<float>(lodOffset) : -FLT_MAX; +} + +FLOAT ConvertMaxLOD(GLenum minFilter, unsigned int lodOffset) +{ + return (minFilter == GL_NEAREST || minFilter == GL_LINEAR) ? static_cast<float>(lodOffset) : FLT_MAX; +} + +} + +namespace d3d11_gl +{ + +GLenum ConvertBackBufferFormat(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_R8G8B8A8_UNORM: return GL_RGBA8_OES; + case DXGI_FORMAT_B8G8R8A8_UNORM: return GL_BGRA8_EXT; + default: + UNREACHABLE(); + } + + return GL_RGBA8_OES; +} + +GLenum ConvertDepthStencilFormat(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_UNKNOWN: return GL_NONE; + case DXGI_FORMAT_D16_UNORM: return GL_DEPTH_COMPONENT16; + case DXGI_FORMAT_D24_UNORM_S8_UINT: return GL_DEPTH24_STENCIL8_OES; + default: + UNREACHABLE(); + } + + return GL_DEPTH24_STENCIL8_OES; +} + +GLenum ConvertRenderbufferFormat(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_B8G8R8A8_UNORM: + return GL_BGRA8_EXT; + case DXGI_FORMAT_R8G8B8A8_UNORM: + return GL_RGBA8_OES; + case DXGI_FORMAT_D16_UNORM: + return GL_DEPTH_COMPONENT16; + case DXGI_FORMAT_D24_UNORM_S8_UINT: + return GL_DEPTH24_STENCIL8_OES; + default: + UNREACHABLE(); + } + + return GL_RGBA8_OES; +} + +GLenum ConvertTextureInternalFormat(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_R8G8B8A8_UNORM: + return GL_RGBA8_OES; + case DXGI_FORMAT_A8_UNORM: + return GL_ALPHA8_EXT; + case DXGI_FORMAT_BC1_UNORM: + return GL_COMPRESSED_RGBA_S3TC_DXT1_EXT; + case DXGI_FORMAT_BC2_UNORM: + return GL_COMPRESSED_RGBA_S3TC_DXT3_ANGLE; + case DXGI_FORMAT_BC3_UNORM: + return GL_COMPRESSED_RGBA_S3TC_DXT5_ANGLE; + case DXGI_FORMAT_R32G32B32A32_FLOAT: + return GL_RGBA32F_EXT; + case DXGI_FORMAT_R32G32B32_FLOAT: + return GL_RGB32F_EXT; + case DXGI_FORMAT_R16G16B16A16_FLOAT: + return GL_RGBA16F_EXT; + case DXGI_FORMAT_B8G8R8A8_UNORM: + return GL_BGRA8_EXT; + case DXGI_FORMAT_R8_UNORM: + return GL_R8_EXT; + case DXGI_FORMAT_R8G8_UNORM: + return GL_RG8_EXT; + case DXGI_FORMAT_R16_FLOAT: + return GL_R16F_EXT; + case DXGI_FORMAT_R16G16_FLOAT: + return GL_RG16F_EXT; + case DXGI_FORMAT_D16_UNORM: + return GL_DEPTH_COMPONENT16; + case DXGI_FORMAT_D24_UNORM_S8_UINT: + return GL_DEPTH24_STENCIL8_OES; + case DXGI_FORMAT_UNKNOWN: + return GL_NONE; + default: + UNREACHABLE(); + } + + return GL_RGBA8_OES; +} + +} + +namespace gl_d3d11 +{ + +DXGI_FORMAT ConvertRenderbufferFormat(GLenum format) +{ + switch (format) + { + case GL_RGBA4: + case GL_RGB5_A1: + case GL_RGBA8_OES: + case GL_RGB565: + case GL_RGB8_OES: + return DXGI_FORMAT_R8G8B8A8_UNORM; + case GL_BGRA8_EXT: + return DXGI_FORMAT_B8G8R8A8_UNORM; + case GL_DEPTH_COMPONENT16: + return DXGI_FORMAT_D16_UNORM; + case GL_STENCIL_INDEX8: + case GL_DEPTH24_STENCIL8_OES: + return DXGI_FORMAT_D24_UNORM_S8_UINT; + default: + UNREACHABLE(); + } + + return DXGI_FORMAT_R8G8B8A8_UNORM; +} + +DXGI_FORMAT ConvertTextureFormat(GLenum internalformat) +{ + switch (internalformat) + { + case GL_RGB565: + case GL_RGBA4: + case GL_RGB5_A1: + case GL_RGB8_OES: + case GL_RGBA8_OES: + case GL_LUMINANCE8_EXT: + case GL_LUMINANCE8_ALPHA8_EXT: + return DXGI_FORMAT_R8G8B8A8_UNORM; + case GL_ALPHA8_EXT: + return DXGI_FORMAT_A8_UNORM; + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return DXGI_FORMAT_BC1_UNORM; + case GL_COMPRESSED_RGBA_S3TC_DXT3_ANGLE: + return DXGI_FORMAT_BC2_UNORM; + case GL_COMPRESSED_RGBA_S3TC_DXT5_ANGLE: + return DXGI_FORMAT_BC3_UNORM; + case GL_RGBA32F_EXT: + case GL_ALPHA32F_EXT: + case GL_LUMINANCE_ALPHA32F_EXT: + return DXGI_FORMAT_R32G32B32A32_FLOAT; + case GL_RGB32F_EXT: + case GL_LUMINANCE32F_EXT: + return DXGI_FORMAT_R32G32B32_FLOAT; + case GL_RGBA16F_EXT: + case GL_ALPHA16F_EXT: + case GL_LUMINANCE_ALPHA16F_EXT: + case GL_RGB16F_EXT: + case GL_LUMINANCE16F_EXT: + return DXGI_FORMAT_R16G16B16A16_FLOAT; + case GL_BGRA8_EXT: + return DXGI_FORMAT_B8G8R8A8_UNORM; + case GL_R8_EXT: + return DXGI_FORMAT_R8_UNORM; + case GL_RG8_EXT: + return DXGI_FORMAT_R8G8_UNORM; + case GL_R16F_EXT: + return DXGI_FORMAT_R16_FLOAT; + case GL_RG16F_EXT: + return DXGI_FORMAT_R16G16_FLOAT; + case GL_DEPTH_COMPONENT16: + return DXGI_FORMAT_D16_UNORM; + case GL_DEPTH_COMPONENT32_OES: + case GL_DEPTH24_STENCIL8_OES: + return DXGI_FORMAT_D24_UNORM_S8_UINT; + case GL_NONE: + return DXGI_FORMAT_UNKNOWN; + default: + UNREACHABLE(); + } + + return DXGI_FORMAT_R8G8B8A8_UNORM; +} + +} + +namespace d3d11 +{ + +void SetPositionTexCoordVertex(PositionTexCoordVertex* vertex, float x, float y, float u, float v) +{ + vertex->x = x; + vertex->y = y; + vertex->u = u; + vertex->v = v; +} + +void SetPositionDepthColorVertex(PositionDepthColorVertex* vertex, float x, float y, float z, + const gl::Color &color) +{ + vertex->x = x; + vertex->y = y; + vertex->z = z; + vertex->r = color.red; + vertex->g = color.green; + vertex->b = color.blue; + vertex->a = color.alpha; +} + +size_t ComputePixelSizeBits(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_R1_UNORM: + return 1; + + case DXGI_FORMAT_A8_UNORM: + case DXGI_FORMAT_R8_SINT: + case DXGI_FORMAT_R8_SNORM: + case DXGI_FORMAT_R8_TYPELESS: + case DXGI_FORMAT_R8_UINT: + case DXGI_FORMAT_R8_UNORM: + return 8; + + case DXGI_FORMAT_B5G5R5A1_UNORM: + case DXGI_FORMAT_B5G6R5_UNORM: + case DXGI_FORMAT_D16_UNORM: + case DXGI_FORMAT_R16_FLOAT: + case DXGI_FORMAT_R16_SINT: + case DXGI_FORMAT_R16_SNORM: + case DXGI_FORMAT_R16_TYPELESS: + case DXGI_FORMAT_R16_UINT: + case DXGI_FORMAT_R16_UNORM: + case DXGI_FORMAT_R8G8_SINT: + case DXGI_FORMAT_R8G8_SNORM: + case DXGI_FORMAT_R8G8_TYPELESS: + case DXGI_FORMAT_R8G8_UINT: + case DXGI_FORMAT_R8G8_UNORM: + return 16; + + case DXGI_FORMAT_B8G8R8X8_TYPELESS: + case DXGI_FORMAT_B8G8R8X8_UNORM: + case DXGI_FORMAT_B8G8R8X8_UNORM_SRGB: + case DXGI_FORMAT_D24_UNORM_S8_UINT: + case DXGI_FORMAT_D32_FLOAT: + case DXGI_FORMAT_D32_FLOAT_S8X24_UINT: + case DXGI_FORMAT_G8R8_G8B8_UNORM: + case DXGI_FORMAT_R10G10B10_XR_BIAS_A2_UNORM: + case DXGI_FORMAT_R10G10B10A2_TYPELESS: + case DXGI_FORMAT_R10G10B10A2_UINT: + case DXGI_FORMAT_R10G10B10A2_UNORM: + case DXGI_FORMAT_R11G11B10_FLOAT: + case DXGI_FORMAT_R16G16_FLOAT: + case DXGI_FORMAT_R16G16_SINT: + case DXGI_FORMAT_R16G16_SNORM: + case DXGI_FORMAT_R16G16_TYPELESS: + case DXGI_FORMAT_R16G16_UINT: + case DXGI_FORMAT_R16G16_UNORM: + case DXGI_FORMAT_R24_UNORM_X8_TYPELESS: + case DXGI_FORMAT_R24G8_TYPELESS: + case DXGI_FORMAT_R32_FLOAT: + case DXGI_FORMAT_R32_FLOAT_X8X24_TYPELESS: + case DXGI_FORMAT_R32_SINT: + case DXGI_FORMAT_R32_TYPELESS: + case DXGI_FORMAT_R32_UINT: + case DXGI_FORMAT_R8G8_B8G8_UNORM: + case DXGI_FORMAT_R8G8B8A8_SINT: + case DXGI_FORMAT_R8G8B8A8_SNORM: + case DXGI_FORMAT_R8G8B8A8_TYPELESS: + case DXGI_FORMAT_R8G8B8A8_UINT: + case DXGI_FORMAT_R8G8B8A8_UNORM: + case DXGI_FORMAT_R8G8B8A8_UNORM_SRGB: + case DXGI_FORMAT_B8G8R8A8_TYPELESS: + case DXGI_FORMAT_B8G8R8A8_UNORM: + case DXGI_FORMAT_B8G8R8A8_UNORM_SRGB: + case DXGI_FORMAT_R9G9B9E5_SHAREDEXP: + case DXGI_FORMAT_X24_TYPELESS_G8_UINT: + case DXGI_FORMAT_X32_TYPELESS_G8X24_UINT: + return 32; + + case DXGI_FORMAT_R16G16B16A16_FLOAT: + case DXGI_FORMAT_R16G16B16A16_SINT: + case DXGI_FORMAT_R16G16B16A16_SNORM: + case DXGI_FORMAT_R16G16B16A16_TYPELESS: + case DXGI_FORMAT_R16G16B16A16_UINT: + case DXGI_FORMAT_R16G16B16A16_UNORM: + case DXGI_FORMAT_R32G32_FLOAT: + case DXGI_FORMAT_R32G32_SINT: + case DXGI_FORMAT_R32G32_TYPELESS: + case DXGI_FORMAT_R32G32_UINT: + case DXGI_FORMAT_R32G8X24_TYPELESS: + return 64; + + case DXGI_FORMAT_R32G32B32_FLOAT: + case DXGI_FORMAT_R32G32B32_SINT: + case DXGI_FORMAT_R32G32B32_TYPELESS: + case DXGI_FORMAT_R32G32B32_UINT: + return 96; + + case DXGI_FORMAT_R32G32B32A32_FLOAT: + case DXGI_FORMAT_R32G32B32A32_SINT: + case DXGI_FORMAT_R32G32B32A32_TYPELESS: + case DXGI_FORMAT_R32G32B32A32_UINT: + return 128; + + case DXGI_FORMAT_BC1_TYPELESS: + case DXGI_FORMAT_BC1_UNORM: + case DXGI_FORMAT_BC1_UNORM_SRGB: + case DXGI_FORMAT_BC4_SNORM: + case DXGI_FORMAT_BC4_TYPELESS: + case DXGI_FORMAT_BC4_UNORM: + return 4; + + case DXGI_FORMAT_BC2_TYPELESS: + case DXGI_FORMAT_BC2_UNORM: + case DXGI_FORMAT_BC2_UNORM_SRGB: + case DXGI_FORMAT_BC3_TYPELESS: + case DXGI_FORMAT_BC3_UNORM: + case DXGI_FORMAT_BC3_UNORM_SRGB: + case DXGI_FORMAT_BC5_SNORM: + case DXGI_FORMAT_BC5_TYPELESS: + case DXGI_FORMAT_BC5_UNORM: + case DXGI_FORMAT_BC6H_SF16: + case DXGI_FORMAT_BC6H_TYPELESS: + case DXGI_FORMAT_BC6H_UF16: + case DXGI_FORMAT_BC7_TYPELESS: + case DXGI_FORMAT_BC7_UNORM: + case DXGI_FORMAT_BC7_UNORM_SRGB: + return 8; + + default: + return 0; + } +} + +size_t ComputeBlockSizeBits(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_BC1_TYPELESS: + case DXGI_FORMAT_BC1_UNORM: + case DXGI_FORMAT_BC1_UNORM_SRGB: + case DXGI_FORMAT_BC4_SNORM: + case DXGI_FORMAT_BC4_TYPELESS: + case DXGI_FORMAT_BC4_UNORM: + case DXGI_FORMAT_BC2_TYPELESS: + case DXGI_FORMAT_BC2_UNORM: + case DXGI_FORMAT_BC2_UNORM_SRGB: + case DXGI_FORMAT_BC3_TYPELESS: + case DXGI_FORMAT_BC3_UNORM: + case DXGI_FORMAT_BC3_UNORM_SRGB: + case DXGI_FORMAT_BC5_SNORM: + case DXGI_FORMAT_BC5_TYPELESS: + case DXGI_FORMAT_BC5_UNORM: + case DXGI_FORMAT_BC6H_SF16: + case DXGI_FORMAT_BC6H_TYPELESS: + case DXGI_FORMAT_BC6H_UF16: + case DXGI_FORMAT_BC7_TYPELESS: + case DXGI_FORMAT_BC7_UNORM: + case DXGI_FORMAT_BC7_UNORM_SRGB: + return ComputePixelSizeBits(format) * 16; + default: + UNREACHABLE(); + return 0; + } +} + +bool IsCompressed(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_BC1_TYPELESS: + case DXGI_FORMAT_BC1_UNORM: + case DXGI_FORMAT_BC1_UNORM_SRGB: + case DXGI_FORMAT_BC4_SNORM: + case DXGI_FORMAT_BC4_TYPELESS: + case DXGI_FORMAT_BC4_UNORM: + case DXGI_FORMAT_BC2_TYPELESS: + case DXGI_FORMAT_BC2_UNORM: + case DXGI_FORMAT_BC2_UNORM_SRGB: + case DXGI_FORMAT_BC3_TYPELESS: + case DXGI_FORMAT_BC3_UNORM: + case DXGI_FORMAT_BC3_UNORM_SRGB: + case DXGI_FORMAT_BC5_SNORM: + case DXGI_FORMAT_BC5_TYPELESS: + case DXGI_FORMAT_BC5_UNORM: + case DXGI_FORMAT_BC6H_SF16: + case DXGI_FORMAT_BC6H_TYPELESS: + case DXGI_FORMAT_BC6H_UF16: + case DXGI_FORMAT_BC7_TYPELESS: + case DXGI_FORMAT_BC7_UNORM: + case DXGI_FORMAT_BC7_UNORM_SRGB: + return true; + case DXGI_FORMAT_UNKNOWN: + UNREACHABLE(); + return false; + default: + return false; + } +} + +unsigned int GetTextureFormatDimensionAlignment(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_BC1_TYPELESS: + case DXGI_FORMAT_BC1_UNORM: + case DXGI_FORMAT_BC1_UNORM_SRGB: + case DXGI_FORMAT_BC4_SNORM: + case DXGI_FORMAT_BC4_TYPELESS: + case DXGI_FORMAT_BC4_UNORM: + case DXGI_FORMAT_BC2_TYPELESS: + case DXGI_FORMAT_BC2_UNORM: + case DXGI_FORMAT_BC2_UNORM_SRGB: + case DXGI_FORMAT_BC3_TYPELESS: + case DXGI_FORMAT_BC3_UNORM: + case DXGI_FORMAT_BC3_UNORM_SRGB: + case DXGI_FORMAT_BC5_SNORM: + case DXGI_FORMAT_BC5_TYPELESS: + case DXGI_FORMAT_BC5_UNORM: + case DXGI_FORMAT_BC6H_SF16: + case DXGI_FORMAT_BC6H_TYPELESS: + case DXGI_FORMAT_BC6H_UF16: + case DXGI_FORMAT_BC7_TYPELESS: + case DXGI_FORMAT_BC7_UNORM: + case DXGI_FORMAT_BC7_UNORM_SRGB: + return 4; + case DXGI_FORMAT_UNKNOWN: + UNREACHABLE(); + return 1; + default: + return 1; + } +} + +bool IsDepthStencilFormat(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_D32_FLOAT_S8X24_UINT: + case DXGI_FORMAT_D32_FLOAT: + case DXGI_FORMAT_D24_UNORM_S8_UINT: + case DXGI_FORMAT_D16_UNORM: + return true; + default: + return false; + } +} + +DXGI_FORMAT GetDepthTextureFormat(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_D32_FLOAT_S8X24_UINT: return DXGI_FORMAT_R32G8X24_TYPELESS; + case DXGI_FORMAT_D32_FLOAT: return DXGI_FORMAT_R32_TYPELESS; + case DXGI_FORMAT_D24_UNORM_S8_UINT: return DXGI_FORMAT_R24G8_TYPELESS; + case DXGI_FORMAT_D16_UNORM: return DXGI_FORMAT_R16_TYPELESS; + default: UNREACHABLE(); return DXGI_FORMAT_UNKNOWN; + } +} + +DXGI_FORMAT GetDepthShaderResourceFormat(DXGI_FORMAT format) +{ + switch (format) + { + case DXGI_FORMAT_D32_FLOAT_S8X24_UINT: return DXGI_FORMAT_R32_FLOAT_X8X24_TYPELESS; + case DXGI_FORMAT_D32_FLOAT: return DXGI_FORMAT_R32_UINT; + case DXGI_FORMAT_D24_UNORM_S8_UINT: return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + case DXGI_FORMAT_D16_UNORM: return DXGI_FORMAT_R16_UNORM; + default: UNREACHABLE(); return DXGI_FORMAT_UNKNOWN; + } +} + +HRESULT SetDebugName(ID3D11DeviceChild *resource, const char *name) +{ +#if defined(_DEBUG) + return resource->SetPrivateData(WKPDID_D3DDebugObjectName, strlen(name), name); +#else + return S_OK; +#endif +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/renderer11_utils.h b/src/3rdparty/angle/src/libGLESv2/renderer/renderer11_utils.h new file mode 100644 index 0000000000..1bc48c1a13 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/renderer11_utils.h @@ -0,0 +1,95 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// renderer11_utils.h: Conversion functions and other utility routines +// specific to the D3D11 renderer. + +#ifndef LIBGLESV2_RENDERER_RENDERER11_UTILS_H +#define LIBGLESV2_RENDERER_RENDERER11_UTILS_H + +#include "libGLESv2/angletypes.h" + +namespace gl_d3d11 +{ + +D3D11_BLEND ConvertBlendFunc(GLenum glBlend, bool isAlpha); +D3D11_BLEND_OP ConvertBlendOp(GLenum glBlendOp); +UINT8 ConvertColorMask(bool maskRed, bool maskGreen, bool maskBlue, bool maskAlpha); + +D3D11_CULL_MODE ConvertCullMode(bool cullEnabled, GLenum cullMode); + +D3D11_COMPARISON_FUNC ConvertComparison(GLenum comparison); +D3D11_DEPTH_WRITE_MASK ConvertDepthMask(bool depthWriteEnabled); +UINT8 ConvertStencilMask(GLuint stencilmask); +D3D11_STENCIL_OP ConvertStencilOp(GLenum stencilOp); + +D3D11_FILTER ConvertFilter(GLenum minFilter, GLenum magFilter, float maxAnisotropy); +D3D11_TEXTURE_ADDRESS_MODE ConvertTextureWrap(GLenum wrap); +FLOAT ConvertMinLOD(GLenum minFilter, unsigned int lodOffset); +FLOAT ConvertMaxLOD(GLenum minFilter, unsigned int lodOffset); + +DXGI_FORMAT ConvertRenderbufferFormat(GLenum format); +DXGI_FORMAT ConvertTextureFormat(GLenum format); +} + +namespace d3d11_gl +{ + +GLenum ConvertBackBufferFormat(DXGI_FORMAT format); +GLenum ConvertDepthStencilFormat(DXGI_FORMAT format); +GLenum ConvertRenderbufferFormat(DXGI_FORMAT format); +GLenum ConvertTextureInternalFormat(DXGI_FORMAT format); + +} + +namespace d3d11 +{ + +struct PositionTexCoordVertex +{ + float x, y; + float u, v; +}; +void SetPositionTexCoordVertex(PositionTexCoordVertex* vertex, float x, float y, float u, float v); + +struct PositionDepthColorVertex +{ + float x, y, z; + float r, g, b, a; +}; +void SetPositionDepthColorVertex(PositionDepthColorVertex* vertex, float x, float y, float z, + const gl::Color &color); + +size_t ComputePixelSizeBits(DXGI_FORMAT format); +size_t ComputeBlockSizeBits(DXGI_FORMAT format); + +bool IsCompressed(DXGI_FORMAT format); +unsigned int GetTextureFormatDimensionAlignment(DXGI_FORMAT format); + +bool IsDepthStencilFormat(DXGI_FORMAT format); +DXGI_FORMAT GetDepthTextureFormat(DXGI_FORMAT format); +DXGI_FORMAT GetDepthShaderResourceFormat(DXGI_FORMAT format); + +HRESULT SetDebugName(ID3D11DeviceChild *resource, const char *name); + +inline bool isDeviceLostError(HRESULT errorCode) +{ + switch (errorCode) + { + case DXGI_ERROR_DEVICE_HUNG: + case DXGI_ERROR_DEVICE_REMOVED: + case DXGI_ERROR_DEVICE_RESET: + case DXGI_ERROR_DRIVER_INTERNAL_ERROR: + case DXGI_ERROR_NOT_CURRENTLY_AVAILABLE: + return true; + default: + return false; + } +} + +} + +#endif // LIBGLESV2_RENDERER_RENDERER11_UTILS_H diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/renderer9_utils.cpp b/src/3rdparty/angle/src/libGLESv2/renderer/renderer9_utils.cpp new file mode 100644 index 0000000000..da75d465e3 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/renderer9_utils.cpp @@ -0,0 +1,500 @@ +#include "precompiled.h" +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// renderer9_utils.cpp: Conversion functions and other utility routines +// specific to the D3D9 renderer. + +#include "libGLESv2/renderer/renderer9_utils.h" +#include "libGLESv2/mathutil.h" +#include "libGLESv2/Context.h" + +#include "common/debug.h" + +namespace gl_d3d9 +{ + +D3DCMPFUNC ConvertComparison(GLenum comparison) +{ + D3DCMPFUNC d3dComp = D3DCMP_ALWAYS; + switch (comparison) + { + case GL_NEVER: d3dComp = D3DCMP_NEVER; break; + case GL_ALWAYS: d3dComp = D3DCMP_ALWAYS; break; + case GL_LESS: d3dComp = D3DCMP_LESS; break; + case GL_LEQUAL: d3dComp = D3DCMP_LESSEQUAL; break; + case GL_EQUAL: d3dComp = D3DCMP_EQUAL; break; + case GL_GREATER: d3dComp = D3DCMP_GREATER; break; + case GL_GEQUAL: d3dComp = D3DCMP_GREATEREQUAL; break; + case GL_NOTEQUAL: d3dComp = D3DCMP_NOTEQUAL; break; + default: UNREACHABLE(); + } + + return d3dComp; +} + +D3DCOLOR ConvertColor(gl::Color color) +{ + return D3DCOLOR_RGBA(gl::unorm<8>(color.red), + gl::unorm<8>(color.green), + gl::unorm<8>(color.blue), + gl::unorm<8>(color.alpha)); +} + +D3DBLEND ConvertBlendFunc(GLenum blend) +{ + D3DBLEND d3dBlend = D3DBLEND_ZERO; + + switch (blend) + { + case GL_ZERO: d3dBlend = D3DBLEND_ZERO; break; + case GL_ONE: d3dBlend = D3DBLEND_ONE; break; + case GL_SRC_COLOR: d3dBlend = D3DBLEND_SRCCOLOR; break; + case GL_ONE_MINUS_SRC_COLOR: d3dBlend = D3DBLEND_INVSRCCOLOR; break; + case GL_DST_COLOR: d3dBlend = D3DBLEND_DESTCOLOR; break; + case GL_ONE_MINUS_DST_COLOR: d3dBlend = D3DBLEND_INVDESTCOLOR; break; + case GL_SRC_ALPHA: d3dBlend = D3DBLEND_SRCALPHA; break; + case GL_ONE_MINUS_SRC_ALPHA: d3dBlend = D3DBLEND_INVSRCALPHA; break; + case GL_DST_ALPHA: d3dBlend = D3DBLEND_DESTALPHA; break; + case GL_ONE_MINUS_DST_ALPHA: d3dBlend = D3DBLEND_INVDESTALPHA; break; + case GL_CONSTANT_COLOR: d3dBlend = D3DBLEND_BLENDFACTOR; break; + case GL_ONE_MINUS_CONSTANT_COLOR: d3dBlend = D3DBLEND_INVBLENDFACTOR; break; + case GL_CONSTANT_ALPHA: d3dBlend = D3DBLEND_BLENDFACTOR; break; + case GL_ONE_MINUS_CONSTANT_ALPHA: d3dBlend = D3DBLEND_INVBLENDFACTOR; break; + case GL_SRC_ALPHA_SATURATE: d3dBlend = D3DBLEND_SRCALPHASAT; break; + default: UNREACHABLE(); + } + + return d3dBlend; +} + +D3DBLENDOP ConvertBlendOp(GLenum blendOp) +{ + D3DBLENDOP d3dBlendOp = D3DBLENDOP_ADD; + + switch (blendOp) + { + case GL_FUNC_ADD: d3dBlendOp = D3DBLENDOP_ADD; break; + case GL_FUNC_SUBTRACT: d3dBlendOp = D3DBLENDOP_SUBTRACT; break; + case GL_FUNC_REVERSE_SUBTRACT: d3dBlendOp = D3DBLENDOP_REVSUBTRACT; break; + default: UNREACHABLE(); + } + + return d3dBlendOp; +} + +D3DSTENCILOP ConvertStencilOp(GLenum stencilOp) +{ + D3DSTENCILOP d3dStencilOp = D3DSTENCILOP_KEEP; + + switch (stencilOp) + { + case GL_ZERO: d3dStencilOp = D3DSTENCILOP_ZERO; break; + case GL_KEEP: d3dStencilOp = D3DSTENCILOP_KEEP; break; + case GL_REPLACE: d3dStencilOp = D3DSTENCILOP_REPLACE; break; + case GL_INCR: d3dStencilOp = D3DSTENCILOP_INCRSAT; break; + case GL_DECR: d3dStencilOp = D3DSTENCILOP_DECRSAT; break; + case GL_INVERT: d3dStencilOp = D3DSTENCILOP_INVERT; break; + case GL_INCR_WRAP: d3dStencilOp = D3DSTENCILOP_INCR; break; + case GL_DECR_WRAP: d3dStencilOp = D3DSTENCILOP_DECR; break; + default: UNREACHABLE(); + } + + return d3dStencilOp; +} + +D3DTEXTUREADDRESS ConvertTextureWrap(GLenum wrap) +{ + D3DTEXTUREADDRESS d3dWrap = D3DTADDRESS_WRAP; + + switch (wrap) + { + case GL_REPEAT: d3dWrap = D3DTADDRESS_WRAP; break; + case GL_CLAMP_TO_EDGE: d3dWrap = D3DTADDRESS_CLAMP; break; + case GL_MIRRORED_REPEAT: d3dWrap = D3DTADDRESS_MIRROR; break; + default: UNREACHABLE(); + } + + return d3dWrap; +} + +D3DCULL ConvertCullMode(GLenum cullFace, GLenum frontFace) +{ + D3DCULL cull = D3DCULL_CCW; + switch (cullFace) + { + case GL_FRONT: + cull = (frontFace == GL_CCW ? D3DCULL_CW : D3DCULL_CCW); + break; + case GL_BACK: + cull = (frontFace == GL_CCW ? D3DCULL_CCW : D3DCULL_CW); + break; + case GL_FRONT_AND_BACK: + cull = D3DCULL_NONE; // culling will be handled during draw + break; + default: UNREACHABLE(); + } + + return cull; +} + +D3DCUBEMAP_FACES ConvertCubeFace(GLenum cubeFace) +{ + D3DCUBEMAP_FACES face = D3DCUBEMAP_FACE_POSITIVE_X; + + switch (cubeFace) + { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X: + face = D3DCUBEMAP_FACE_POSITIVE_X; + break; + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: + face = D3DCUBEMAP_FACE_NEGATIVE_X; + break; + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: + face = D3DCUBEMAP_FACE_POSITIVE_Y; + break; + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: + face = D3DCUBEMAP_FACE_NEGATIVE_Y; + break; + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: + face = D3DCUBEMAP_FACE_POSITIVE_Z; + break; + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: + face = D3DCUBEMAP_FACE_NEGATIVE_Z; + break; + default: UNREACHABLE(); + } + + return face; +} + +DWORD ConvertColorMask(bool red, bool green, bool blue, bool alpha) +{ + return (red ? D3DCOLORWRITEENABLE_RED : 0) | + (green ? D3DCOLORWRITEENABLE_GREEN : 0) | + (blue ? D3DCOLORWRITEENABLE_BLUE : 0) | + (alpha ? D3DCOLORWRITEENABLE_ALPHA : 0); +} + +D3DTEXTUREFILTERTYPE ConvertMagFilter(GLenum magFilter, float maxAnisotropy) +{ + if (maxAnisotropy > 1.0f) + { + return D3DTEXF_ANISOTROPIC; + } + + D3DTEXTUREFILTERTYPE d3dMagFilter = D3DTEXF_POINT; + switch (magFilter) + { + case GL_NEAREST: d3dMagFilter = D3DTEXF_POINT; break; + case GL_LINEAR: d3dMagFilter = D3DTEXF_LINEAR; break; + default: UNREACHABLE(); + } + + return d3dMagFilter; +} + +void ConvertMinFilter(GLenum minFilter, D3DTEXTUREFILTERTYPE *d3dMinFilter, D3DTEXTUREFILTERTYPE *d3dMipFilter, float maxAnisotropy) +{ + switch (minFilter) + { + case GL_NEAREST: + *d3dMinFilter = D3DTEXF_POINT; + *d3dMipFilter = D3DTEXF_NONE; + break; + case GL_LINEAR: + *d3dMinFilter = D3DTEXF_LINEAR; + *d3dMipFilter = D3DTEXF_NONE; + break; + case GL_NEAREST_MIPMAP_NEAREST: + *d3dMinFilter = D3DTEXF_POINT; + *d3dMipFilter = D3DTEXF_POINT; + break; + case GL_LINEAR_MIPMAP_NEAREST: + *d3dMinFilter = D3DTEXF_LINEAR; + *d3dMipFilter = D3DTEXF_POINT; + break; + case GL_NEAREST_MIPMAP_LINEAR: + *d3dMinFilter = D3DTEXF_POINT; + *d3dMipFilter = D3DTEXF_LINEAR; + break; + case GL_LINEAR_MIPMAP_LINEAR: + *d3dMinFilter = D3DTEXF_LINEAR; + *d3dMipFilter = D3DTEXF_LINEAR; + break; + default: + *d3dMinFilter = D3DTEXF_POINT; + *d3dMipFilter = D3DTEXF_NONE; + UNREACHABLE(); + } + + if (maxAnisotropy > 1.0f) + { + *d3dMinFilter = D3DTEXF_ANISOTROPIC; + } +} + +D3DFORMAT ConvertRenderbufferFormat(GLenum format) +{ + switch (format) + { + case GL_NONE: return D3DFMT_NULL; + case GL_RGBA4: + case GL_RGB5_A1: + case GL_RGBA8_OES: return D3DFMT_A8R8G8B8; + case GL_RGB565: return D3DFMT_R5G6B5; + case GL_RGB8_OES: return D3DFMT_X8R8G8B8; + case GL_DEPTH_COMPONENT16: + case GL_STENCIL_INDEX8: + case GL_DEPTH24_STENCIL8_OES: return D3DFMT_D24S8; + default: UNREACHABLE(); return D3DFMT_A8R8G8B8; + } +} + +D3DMULTISAMPLE_TYPE GetMultisampleTypeFromSamples(GLsizei samples) +{ + if (samples <= 1) + return D3DMULTISAMPLE_NONE; + else + return (D3DMULTISAMPLE_TYPE)samples; +} + +} + +namespace d3d9_gl +{ + +unsigned int GetStencilSize(D3DFORMAT stencilFormat) +{ + if (stencilFormat == D3DFMT_INTZ) + { + return 8; + } + switch(stencilFormat) + { + case D3DFMT_D24FS8: + case D3DFMT_D24S8: + return 8; + case D3DFMT_D24X4S4: + return 4; + case D3DFMT_D15S1: + return 1; + case D3DFMT_D16_LOCKABLE: + case D3DFMT_D32: + case D3DFMT_D24X8: + case D3DFMT_D32F_LOCKABLE: + case D3DFMT_D16: + return 0; + //case D3DFMT_D32_LOCKABLE: return 0; // DirectX 9Ex only + //case D3DFMT_S8_LOCKABLE: return 8; // DirectX 9Ex only + default: + return 0; + } +} + +unsigned int GetAlphaSize(D3DFORMAT colorFormat) +{ + switch (colorFormat) + { + case D3DFMT_A16B16G16R16F: + return 16; + case D3DFMT_A32B32G32R32F: + return 32; + case D3DFMT_A2R10G10B10: + return 2; + case D3DFMT_A8R8G8B8: + return 8; + case D3DFMT_A1R5G5B5: + return 1; + case D3DFMT_X8R8G8B8: + case D3DFMT_R5G6B5: + return 0; + default: + return 0; + } +} + +GLsizei GetSamplesFromMultisampleType(D3DMULTISAMPLE_TYPE type) +{ + if (type == D3DMULTISAMPLE_NONMASKABLE) + return 0; + else + return type; +} + +bool IsFormatChannelEquivalent(D3DFORMAT d3dformat, GLenum format) +{ + switch (d3dformat) + { + case D3DFMT_L8: + return (format == GL_LUMINANCE); + case D3DFMT_A8L8: + return (format == GL_LUMINANCE_ALPHA); + case D3DFMT_DXT1: + return (format == GL_COMPRESSED_RGBA_S3TC_DXT1_EXT || format == GL_COMPRESSED_RGB_S3TC_DXT1_EXT); + case D3DFMT_DXT3: + return (format == GL_COMPRESSED_RGBA_S3TC_DXT3_ANGLE); + case D3DFMT_DXT5: + return (format == GL_COMPRESSED_RGBA_S3TC_DXT5_ANGLE); + case D3DFMT_A8R8G8B8: + case D3DFMT_A16B16G16R16F: + case D3DFMT_A32B32G32R32F: + return (format == GL_RGBA || format == GL_BGRA_EXT); + case D3DFMT_X8R8G8B8: + return (format == GL_RGB); + default: + if (d3dformat == D3DFMT_INTZ && gl::IsDepthTexture(format)) + return true; + return false; + } +} + +GLenum ConvertBackBufferFormat(D3DFORMAT format) +{ + switch (format) + { + case D3DFMT_A4R4G4B4: return GL_RGBA4; + case D3DFMT_A8R8G8B8: return GL_RGBA8_OES; + case D3DFMT_A1R5G5B5: return GL_RGB5_A1; + case D3DFMT_R5G6B5: return GL_RGB565; + case D3DFMT_X8R8G8B8: return GL_RGB8_OES; + default: + UNREACHABLE(); + } + + return GL_RGBA4; +} + +GLenum ConvertDepthStencilFormat(D3DFORMAT format) +{ + if (format == D3DFMT_INTZ) + { + return GL_DEPTH24_STENCIL8_OES; + } + switch (format) + { + case D3DFMT_D16: + case D3DFMT_D24X8: + return GL_DEPTH_COMPONENT16; + case D3DFMT_D24S8: + return GL_DEPTH24_STENCIL8_OES; + case D3DFMT_UNKNOWN: + return GL_NONE; + default: + UNREACHABLE(); + } + + return GL_DEPTH24_STENCIL8_OES; +} + +GLenum ConvertRenderTargetFormat(D3DFORMAT format) +{ + if (format == D3DFMT_INTZ) + { + return GL_DEPTH24_STENCIL8_OES; + } + + switch (format) + { + case D3DFMT_A4R4G4B4: return GL_RGBA4; + case D3DFMT_A8R8G8B8: return GL_RGBA8_OES; + case D3DFMT_A1R5G5B5: return GL_RGB5_A1; + case D3DFMT_R5G6B5: return GL_RGB565; + case D3DFMT_X8R8G8B8: return GL_RGB8_OES; + case D3DFMT_D16: + case D3DFMT_D24X8: + return GL_DEPTH_COMPONENT16; + case D3DFMT_D24S8: + return GL_DEPTH24_STENCIL8_OES; + case D3DFMT_UNKNOWN: + return GL_NONE; + default: + UNREACHABLE(); + } + + return GL_RGBA4; +} + +GLenum GetEquivalentFormat(D3DFORMAT format) +{ + if (format == D3DFMT_INTZ) + return GL_DEPTH24_STENCIL8_OES; + if (format == D3DFMT_NULL) + return GL_NONE; + + switch (format) + { + case D3DFMT_A4R4G4B4: return GL_RGBA4; + case D3DFMT_A8R8G8B8: return GL_RGBA8_OES; + case D3DFMT_A1R5G5B5: return GL_RGB5_A1; + case D3DFMT_R5G6B5: return GL_RGB565; + case D3DFMT_X8R8G8B8: return GL_RGB8_OES; + case D3DFMT_D16: return GL_DEPTH_COMPONENT16; + case D3DFMT_D24S8: return GL_DEPTH24_STENCIL8_OES; + case D3DFMT_UNKNOWN: return GL_NONE; + case D3DFMT_DXT1: return GL_COMPRESSED_RGBA_S3TC_DXT1_EXT; + case D3DFMT_DXT3: return GL_COMPRESSED_RGBA_S3TC_DXT3_ANGLE; + case D3DFMT_DXT5: return GL_COMPRESSED_RGBA_S3TC_DXT5_ANGLE; + case D3DFMT_A32B32G32R32F: return GL_RGBA32F_EXT; + case D3DFMT_A16B16G16R16F: return GL_RGBA16F_EXT; + case D3DFMT_L8: return GL_LUMINANCE8_EXT; + case D3DFMT_A8L8: return GL_LUMINANCE8_ALPHA8_EXT; + default: UNREACHABLE(); + return GL_NONE; + } +} + +} + +namespace d3d9 +{ + +bool IsCompressedFormat(D3DFORMAT surfaceFormat) +{ + switch(surfaceFormat) + { + case D3DFMT_DXT1: + case D3DFMT_DXT2: + case D3DFMT_DXT3: + case D3DFMT_DXT4: + case D3DFMT_DXT5: + return true; + default: + return false; + } +} + +size_t ComputeRowSize(D3DFORMAT format, unsigned int width) +{ + if (format == D3DFMT_INTZ) + { + return 4 * width; + } + switch (format) + { + case D3DFMT_L8: + return 1 * width; + case D3DFMT_A8L8: + return 2 * width; + case D3DFMT_X8R8G8B8: + case D3DFMT_A8R8G8B8: + return 4 * width; + case D3DFMT_A16B16G16R16F: + return 8 * width; + case D3DFMT_A32B32G32R32F: + return 16 * width; + case D3DFMT_DXT1: + return 8 * ((width + 3) / 4); + case D3DFMT_DXT3: + case D3DFMT_DXT5: + return 16 * ((width + 3) / 4); + default: + UNREACHABLE(); + return 0; + } +} + +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/renderer9_utils.h b/src/3rdparty/angle/src/libGLESv2/renderer/renderer9_utils.h new file mode 100644 index 0000000000..bf6cdf1ea6 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/renderer9_utils.h @@ -0,0 +1,74 @@ +// +// Copyright (c) 2002-2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// renderer9_utils.h: Conversion functions and other utility routines +// specific to the D3D9 renderer + +#ifndef LIBGLESV2_RENDERER_RENDERER9_UTILS_H +#define LIBGLESV2_RENDERER_RENDERER9_UTILS_H + +#include "libGLESv2/utilities.h" + +const D3DFORMAT D3DFMT_INTZ = ((D3DFORMAT)(MAKEFOURCC('I','N','T','Z'))); +const D3DFORMAT D3DFMT_NULL = ((D3DFORMAT)(MAKEFOURCC('N','U','L','L'))); + +namespace gl_d3d9 +{ + +D3DCMPFUNC ConvertComparison(GLenum comparison); +D3DCOLOR ConvertColor(gl::Color color); +D3DBLEND ConvertBlendFunc(GLenum blend); +D3DBLENDOP ConvertBlendOp(GLenum blendOp); +D3DSTENCILOP ConvertStencilOp(GLenum stencilOp); +D3DTEXTUREADDRESS ConvertTextureWrap(GLenum wrap); +D3DCULL ConvertCullMode(GLenum cullFace, GLenum frontFace); +D3DCUBEMAP_FACES ConvertCubeFace(GLenum cubeFace); +DWORD ConvertColorMask(bool red, bool green, bool blue, bool alpha); +D3DTEXTUREFILTERTYPE ConvertMagFilter(GLenum magFilter, float maxAnisotropy); +void ConvertMinFilter(GLenum minFilter, D3DTEXTUREFILTERTYPE *d3dMinFilter, D3DTEXTUREFILTERTYPE *d3dMipFilter, float maxAnisotropy); +D3DFORMAT ConvertRenderbufferFormat(GLenum format); +D3DMULTISAMPLE_TYPE GetMultisampleTypeFromSamples(GLsizei samples); + +} + +namespace d3d9_gl +{ + +GLuint GetAlphaSize(D3DFORMAT colorFormat); +GLuint GetStencilSize(D3DFORMAT stencilFormat); + +GLsizei GetSamplesFromMultisampleType(D3DMULTISAMPLE_TYPE type); + +bool IsFormatChannelEquivalent(D3DFORMAT d3dformat, GLenum format); +GLenum ConvertBackBufferFormat(D3DFORMAT format); +GLenum ConvertDepthStencilFormat(D3DFORMAT format); +GLenum ConvertRenderTargetFormat(D3DFORMAT format); +GLenum GetEquivalentFormat(D3DFORMAT format); + +} + +namespace d3d9 +{ +bool IsCompressedFormat(D3DFORMAT format); +size_t ComputeRowSize(D3DFORMAT format, unsigned int width); + +inline bool isDeviceLostError(HRESULT errorCode) +{ + switch (errorCode) + { + case D3DERR_DRIVERINTERNALERROR: + case D3DERR_DEVICELOST: + case D3DERR_DEVICEHUNG: + case D3DERR_DEVICEREMOVED: + return true; + default: + return false; + } +} + +} + +#endif // LIBGLESV2_RENDERER_RENDERER9_UTILS_H diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Blit.ps b/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Blit.ps new file mode 100644 index 0000000000..dcb3bd0e76 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Blit.ps @@ -0,0 +1,39 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +sampler2D tex : s0; + +uniform float4 mode : c0; + +// Passthrough Pixel Shader +// Outputs texture 0 sampled at texcoord 0. +float4 passthroughps(float4 texcoord : TEXCOORD0) : COLOR +{ + return tex2D(tex, texcoord.xy); +}; + +// Luminance Conversion Pixel Shader +// Outputs sample(tex0, tc0).rrra. +// For LA output (pass A) set C0.X = 1, C0.Y = 0. +// For L output (A = 1) set C0.X = 0, C0.Y = 1. +float4 luminanceps(float4 texcoord : TEXCOORD0) : COLOR +{ + float4 tmp = tex2D(tex, texcoord.xy); + tmp.w = tmp.w * mode.x + mode.y; + return tmp.xxxw; +}; + +// RGB/A Component Mask Pixel Shader +// Outputs sample(tex0, tc0) with options to force RGB = 0 and/or A = 1. +// To force RGB = 0, set C0.X = 0, otherwise C0.X = 1. +// To force A = 1, set C0.Z = 0, C0.W = 1, otherwise C0.Z = 1, C0.W = 0. +float4 componentmaskps(float4 texcoord : TEXCOORD0) : COLOR +{ + float4 tmp = tex2D(tex, texcoord.xy); + tmp.xyz = tmp.xyz * mode.x; + tmp.w = tmp.w * mode.z + mode.w; + return tmp; +}; diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Blit.vs b/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Blit.vs new file mode 100644 index 0000000000..3a36980b93 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Blit.vs @@ -0,0 +1,43 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +struct VS_OUTPUT +{ + float4 position : POSITION; + float4 texcoord : TEXCOORD0; +}; + +uniform float4 halfPixelSize : c0; + +// Standard Vertex Shader +// Input 0 is the homogenous position. +// Outputs the homogenous position as-is. +// Outputs a tex coord with (0,0) in the upper-left corner of the screen and (1,1) in the bottom right. +// C0.X must be negative half-pixel width, C0.Y must be half-pixel height. C0.ZW must be 0. +VS_OUTPUT standardvs(in float4 position : POSITION) +{ + VS_OUTPUT Out; + + Out.position = position + halfPixelSize; + Out.texcoord = position * float4(0.5, -0.5, 1.0, 1.0) + float4(0.5, 0.5, 0, 0); + + return Out; +}; + +// Flip Y Vertex Shader +// Input 0 is the homogenous position. +// Outputs the homogenous position as-is. +// Outputs a tex coord with (0,1) in the upper-left corner of the screen and (1,0) in the bottom right. +// C0.XY must be the half-pixel width and height. C0.ZW must be 0. +VS_OUTPUT flipyvs(in float4 position : POSITION) +{ + VS_OUTPUT Out; + + Out.position = position + halfPixelSize; + Out.texcoord = position * float4(0.5, 0.5, 1.0, 1.0) + float4(0.5, 0.5, 0, 0); + + return Out; +}; diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Clear11.hlsl b/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Clear11.hlsl new file mode 100644 index 0000000000..d2752601e9 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Clear11.hlsl @@ -0,0 +1,33 @@ +void VS_Clear( in float3 inPosition : POSITION, in float4 inColor : COLOR, + out float4 outPosition : SV_POSITION, out float4 outColor : COLOR) +{ + outPosition = float4(inPosition, 1.0f); + outColor = inColor; +} + +// Assume we are in SM4+, which has 8 color outputs +struct PS_Output +{ + float4 color0 : SV_TARGET0; + float4 color1 : SV_TARGET1; + float4 color2 : SV_TARGET2; + float4 color3 : SV_TARGET3; + float4 color4 : SV_TARGET4; + float4 color5 : SV_TARGET5; + float4 color6 : SV_TARGET6; + float4 color7 : SV_TARGET7; +}; + +PS_Output PS_Clear(in float4 inPosition : SV_POSITION, in float4 inColor : COLOR) +{ + PS_Output outColor; + outColor.color0 = inColor; + outColor.color1 = inColor; + outColor.color2 = inColor; + outColor.color3 = inColor; + outColor.color4 = inColor; + outColor.color5 = inColor; + outColor.color6 = inColor; + outColor.color7 = inColor; + return outColor; +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Passthrough11.hlsl b/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Passthrough11.hlsl new file mode 100644 index 0000000000..43b7801efc --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/shaders/Passthrough11.hlsl @@ -0,0 +1,29 @@ +Texture2D Texture : register(t0); +SamplerState Sampler : register(s0); + +void VS_Passthrough( in float2 inPosition : POSITION, in float2 inTexCoord : TEXCOORD0, + out float4 outPosition : SV_POSITION, out float2 outTexCoord : TEXCOORD0) +{ + outPosition = float4(inPosition, 0.0f, 1.0f); + outTexCoord = inTexCoord; +} + +float4 PS_PassthroughRGBA(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return Texture.Sample(Sampler, inTexCoord).rgba; +} + +float4 PS_PassthroughRGB(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return float4(Texture.Sample(Sampler, inTexCoord).rgb, 1.0f); +} + +float4 PS_PassthroughLum(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return float4(Texture.Sample(Sampler, inTexCoord).rrr, 1.0f); +} + +float4 PS_PassthroughLumAlpha(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return Texture.Sample(Sampler, inTexCoord).rrra; +} diff --git a/src/3rdparty/angle/src/libGLESv2/renderer/vertexconversion.h b/src/3rdparty/angle/src/libGLESv2/renderer/vertexconversion.h new file mode 100644 index 0000000000..590b9d48a3 --- /dev/null +++ b/src/3rdparty/angle/src/libGLESv2/renderer/vertexconversion.h @@ -0,0 +1,203 @@ +// +// Copyright (c) 2002-2010 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// vertexconversion.h: A library of vertex conversion classes that can be used to build +// the FormatConverter objects used by the buffer conversion system. + +#ifndef LIBGLESV2_VERTEXCONVERSION_H_ +#define LIBGLESV2_VERTEXCONVERSION_H_ + +namespace rx +{ + +// Conversion types: +// static const bool identity: true if this is an identity transform, false otherwise +// static U convert(T): convert a single element from the input type to the output type +// typedef ... OutputType: the type produced by this conversion + +template <class T> +struct Identity +{ + static const bool identity = true; + + typedef T OutputType; + + static T convert(T x) + { + return x; + } +}; + +template <class FromT, class ToT> +struct Cast +{ + static const bool identity = false; + + typedef ToT OutputType; + + static ToT convert(FromT x) + { + return static_cast<ToT>(x); + } +}; + +template <class T> +struct Cast<T, T> +{ + static const bool identity = true; + + typedef T OutputType; + + static T convert(T x) + { + return static_cast<T>(x); + } +}; + +template <class T> +struct Normalize +{ + static const bool identity = false; + + typedef float OutputType; + + static float convert(T x) + { + typedef std::numeric_limits<T> NL; + float f = static_cast<float>(x); + + if (NL::is_signed) + { + // const float => VC2008 computes it at compile time + // static const float => VC2008 computes it the first time we get here, stores it to memory with static guard and all that. + const float divisor = 1.0f/(2*static_cast<float>(NL::max())+1); + return (2*f+1)*divisor; + } + else + { + return f/NL::max(); + } + } +}; + +template <class FromType, std::size_t ScaleBits> +struct FixedToFloat +{ + static const bool identity = false; + + typedef float OutputType; + + static float convert(FromType x) + { + const float divisor = 1.0f / static_cast<float>(static_cast<FromType>(1) << ScaleBits); + return static_cast<float>(x) * divisor; + } +}; + +// Widen types: +// static const unsigned int initialWidth: number of components before conversion +// static const unsigned int finalWidth: number of components after conversion + +// Float is supported at any size. +template <std::size_t N> +struct NoWiden +{ + static const std::size_t initialWidth = N; + static const std::size_t finalWidth = N; +}; + +// SHORT, norm-SHORT, norm-UNSIGNED_SHORT are supported but only with 2 or 4 components +template <std::size_t N> +struct WidenToEven +{ + static const std::size_t initialWidth = N; + static const std::size_t finalWidth = N+(N&1); +}; + +template <std::size_t N> +struct WidenToFour +{ + static const std::size_t initialWidth = N; + static const std::size_t finalWidth = 4; +}; + +// Most types have 0 and 1 that are just that. +template <class T> +struct SimpleDefaultValues +{ + static T zero() { return static_cast<T>(0); } + static T one() { return static_cast<T>(1); } +}; + +// But normalised types only store [0,1] or [-1,1] so 1.0 is represented by the max value. +template <class T> +struct NormalizedDefaultValues +{ + static T zero() { return static_cast<T>(0); } + static T one() { return std::numeric_limits<T>::max(); } +}; + +// Converter: +// static const bool identity: true if this is an identity transform (with no widening) +// static const std::size_t finalSize: number of bytes per output vertex +// static void convertArray(const void *in, std::size_t stride, std::size_t n, void *out): convert an array of vertices. Input may be strided, but output will be unstrided. + +template <class InT, class WidenRule, class Converter, class DefaultValueRule = SimpleDefaultValues<InT> > +struct VertexDataConverter +{ + typedef typename Converter::OutputType OutputType; + typedef InT InputType; + + static const bool identity = (WidenRule::initialWidth == WidenRule::finalWidth) && Converter::identity; + static const std::size_t finalSize = WidenRule::finalWidth * sizeof(OutputType); + + static void convertArray(const InputType *in, std::size_t stride, std::size_t n, OutputType *out) + { + for (std::size_t i = 0; i < n; i++) + { + const InputType *ein = pointerAddBytes(in, i * stride); + + copyComponent(out, ein, 0, static_cast<OutputType>(DefaultValueRule::zero())); + copyComponent(out, ein, 1, static_cast<OutputType>(DefaultValueRule::zero())); + copyComponent(out, ein, 2, static_cast<OutputType>(DefaultValueRule::zero())); + copyComponent(out, ein, 3, static_cast<OutputType>(DefaultValueRule::one())); + + out += WidenRule::finalWidth; + } + } + + static void convertArray(const void *in, std::size_t stride, std::size_t n, void *out) + { + return convertArray(static_cast<const InputType*>(in), stride, n, static_cast<OutputType*>(out)); + } + + private: + // Advance the given pointer by a number of bytes (not pointed-to elements). + template <class T> + static T *pointerAddBytes(T *basePtr, std::size_t numBytes) + { + return reinterpret_cast<T *>(reinterpret_cast<uintptr_t>(basePtr) + numBytes); + } + + static void copyComponent(OutputType *out, const InputType *in, std::size_t elementindex, OutputType defaultvalue) + { + if (WidenRule::finalWidth > elementindex) + { + if (WidenRule::initialWidth > elementindex) + { + out[elementindex] = Converter::convert(in[elementindex]); + } + else + { + out[elementindex] = defaultvalue; + } + } + } +}; + +} + +#endif // LIBGLESV2_VERTEXCONVERSION_H_ |