blob: f02e334f30e52ec5cb578caf1bd86d5543c407ab [file] [log] [blame]
#include "precompiled.h"
//
// Copyright (c) 2012-2014 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 "common/utilities.h"
#include "libGLESv2/Buffer.h"
#include "libGLESv2/ProgramBinary.h"
#include "libGLESv2/Framebuffer.h"
#include "libGLESv2/RenderBuffer.h"
#include "libGLESv2/renderer/d3d11/Renderer11.h"
#include "libGLESv2/renderer/d3d11/RenderTarget11.h"
#include "libGLESv2/renderer/d3d11/renderer11_utils.h"
#include "libGLESv2/renderer/d3d11/formatutils11.h"
#include "libGLESv2/renderer/d3d11/ShaderExecutable11.h"
#include "libGLESv2/renderer/d3d11/SwapChain11.h"
#include "libGLESv2/renderer/d3d11/Image11.h"
#include "libGLESv2/renderer/d3d11/VertexBuffer11.h"
#include "libGLESv2/renderer/d3d11/IndexBuffer11.h"
#include "libGLESv2/renderer/d3d11/BufferStorage11.h"
#include "libGLESv2/renderer/VertexDataManager.h"
#include "libGLESv2/renderer/IndexDataManager.h"
#include "libGLESv2/renderer/d3d11/TextureStorage11.h"
#include "libGLESv2/renderer/d3d11/Query11.h"
#include "libGLESv2/renderer/d3d11/Fence11.h"
#include "libGLESv2/renderer/d3d11/Blit11.h"
#include "libGLESv2/renderer/d3d11/Clear11.h"
#include "libGLESv2/renderer/d3d11/PixelTransfer11.h"
#include "libEGL/Display.h"
// Enable ANGLE_SKIP_DXGI_1_2_CHECK if there is not a possibility of using cross-process
// HWNDs or the Windows 7 Platform Update (KB2670838) is expected to be installed.
#ifndef ANGLE_SKIP_DXGI_1_2_CHECK
#define ANGLE_SKIP_DXGI_1_2_CHECK 0
#endif
#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;
mBlit = NULL;
mPixelTransfer = NULL;
mClear = 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;
mAppliedVertexShader = NULL;
mAppliedGeometryShader = NULL;
mCurPointGeometryShader = NULL;
mAppliedPixelShader = NULL;
}
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 (!mCompiler.initialize())
{
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 = S_OK;
#ifdef _DEBUG
result = D3D11CreateDevice(NULL,
D3D_DRIVER_TYPE_HARDWARE,
NULL,
D3D11_CREATE_DEVICE_DEBUG,
featureLevels,
ArraySize(featureLevels),
D3D11_SDK_VERSION,
&mDevice,
&mFeatureLevel,
&mDeviceContext);
if (!mDevice || FAILED(result))
{
ERR("Failed creating Debug D3D11 device - falling back to release runtime.\n");
}
if (!mDevice || FAILED(result))
#endif
{
result = D3D11CreateDevice(NULL,
D3D_DRIVER_TYPE_HARDWARE,
NULL,
0,
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
}
}
#if !ANGLE_SKIP_DXGI_1_2_CHECK
// In order to create a swap chain for an HWND owned by another process, DXGI 1.2 is required.
// The easiest way to check is to query for a IDXGIDevice2.
bool requireDXGI1_2 = false;
HWND hwnd = WindowFromDC(mDc);
if (hwnd)
{
DWORD currentProcessId = GetCurrentProcessId();
DWORD wndProcessId;
GetWindowThreadProcessId(hwnd, &wndProcessId);
requireDXGI1_2 = (currentProcessId != wndProcessId);
}
else
{
requireDXGI1_2 = true;
}
if (requireDXGI1_2)
{
IDXGIDevice2 *dxgiDevice2 = NULL;
result = mDevice->QueryInterface(__uuidof(IDXGIDevice2), (void**)&dxgiDevice2);
if (FAILED(result))
{
ERR("DXGI 1.2 required to present to HWNDs owned by another process.\n");
return EGL_NOT_INITIALIZED;
}
SafeRelease(dxgiDevice2);
}
#endif
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;
}
SafeRelease(dxgiDevice);
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_DRAW_RENDERTARGETVIEW_NOT_SET
};
D3D11_INFO_QUEUE_FILTER filter = {0};
filter.DenyList.NumIDs = ArraySize(hideMessages);
filter.DenyList.pIDList = hideMessages;
infoQueue->AddStorageFilterEntries(&filter);
SafeRelease(infoQueue);
}
#endif
mMaxSupportedSamples = 0;
const d3d11::DXGIFormatSet &dxgiFormats = d3d11::GetAllUsedDXGIFormats();
for (d3d11::DXGIFormatSet::const_iterator i = dxgiFormats.begin(); i != dxgiFormats.end(); ++i)
{
MultisampleSupportInfo support = getMultisampleSupportInfo(*i);
mMultisampleSupportMap.insert(std::make_pair(*i, support));
mMaxSupportedSamples = std::max(mMaxSupportedSamples, support.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_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;
}
}
DXGI_FORMAT rgTextureFormats[] =
{
DXGI_FORMAT_R8_UNORM,
DXGI_FORMAT_R8G8_UNORM,
DXGI_FORMAT_R16_FLOAT,
DXGI_FORMAT_R16G16_FLOAT,
DXGI_FORMAT_R32_FLOAT,
DXGI_FORMAT_R32G32_FLOAT,
};
mRGTextureSupport = true;
for (unsigned int i = 0; i < ArraySize(rgTextureFormats); i++)
{
if (SUCCEEDED(mDevice->CheckFormatSupport(rgTextureFormats[i], &formatSupport)))
{
mRGTextureSupport = mRGTextureSupport && (formatSupport & requiredTextureFlags) == requiredTextureFlags;
mRGTextureSupport = mRGTextureSupport && (formatSupport & requiredFilterFlags) == requiredFilterFlags;
mRGTextureSupport = mRGTextureSupport && (formatSupport & requiredRenderableFlags) == requiredRenderableFlags;
}
else
{
mRGTextureSupport = 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);
ASSERT(!mBlit);
mBlit = new Blit11(this);
ASSERT(!mClear);
mClear = new Clear11(this);
ASSERT(!mPixelTransfer);
mPixelTransfer = new PixelTransfer11(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 depthStencilSupport = 0;
result = mDevice->CheckFormatSupport(depthStencilFormat, &depthStencilSupport);
depthStencilFormatOK = SUCCEEDED(result) && (depthStencilSupport & D3D11_FORMAT_SUPPORT_DEPTH_STENCIL);
}
if (depthStencilFormatOK)
{
// FIXME: parse types from context version
ASSERT(d3d11_gl::GetInternalFormat(renderTargetFormat, 2) == d3d11_gl::GetInternalFormat(renderTargetFormat, 3));
ASSERT(d3d11_gl::GetInternalFormat(depthStencilFormat, 2) == d3d11_gl::GetInternalFormat(depthStencilFormat, 3));
ConfigDesc newConfig;
newConfig.renderTargetFormat = d3d11_gl::GetInternalFormat(renderTargetFormat, getCurrentClientVersion());
newConfig.depthStencilFormat = d3d11_gl::GetInternalFormat(depthStencilFormat, getCurrentClientVersion());
newConfig.multiSample = 0; // FIXME: enumerate multi-sampling
newConfig.fastConfig = true; // Assume all DX11 format conversions to be fast
newConfig.es3Capable = true;
(*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::generateSwizzle(gl::Texture *texture)
{
if (texture)
{
TextureStorageInterface *texStorage = texture->getNativeTexture();
if (texStorage)
{
TextureStorage11 *storage11 = TextureStorage11::makeTextureStorage11(texStorage->getStorageInstance());
storage11->generateSwizzles(texture->getSwizzleRed(), texture->getSwizzleGreen(), texture->getSwizzleBlue(),
texture->getSwizzleAlpha());
}
}
}
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;
bool forceSetTexture = false;
if (texture)
{
TextureStorageInterface *texStorage = texture->getNativeTexture();
if (texStorage)
{
TextureStorage11 *storage11 = TextureStorage11::makeTextureStorage11(texStorage->getStorageInstance());
gl::SamplerState samplerState;
texture->getSamplerState(&samplerState);
textureSRV = storage11->getSRV(samplerState);
}
// 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);
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 || mCurPixelSRVs[index] != textureSRV)
{
mDeviceContext->PSSetShaderResources(index, 1, &textureSRV);
}
mCurPixelSRVs[index] = textureSRV;
}
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 || mCurVertexSRVs[index] != textureSRV)
{
mDeviceContext->VSSetShaderResources(index, 1, &textureSRV);
}
mCurVertexSRVs[index] = textureSRV;
}
else UNREACHABLE();
}
bool Renderer11::setUniformBuffers(const gl::Buffer *vertexUniformBuffers[], const gl::Buffer *fragmentUniformBuffers[])
{
for (unsigned int uniformBufferIndex = 0; uniformBufferIndex < gl::IMPLEMENTATION_MAX_VERTEX_SHADER_UNIFORM_BUFFERS; uniformBufferIndex++)
{
const gl::Buffer *uniformBuffer = vertexUniformBuffers[uniformBufferIndex];
if (uniformBuffer)
{
BufferStorage11 *bufferStorage = BufferStorage11::makeBufferStorage11(uniformBuffer->getStorage());
ID3D11Buffer *constantBuffer = bufferStorage->getBuffer(BUFFER_USAGE_UNIFORM);
if (!constantBuffer)
{
return false;
}
if (mCurrentConstantBufferVS[uniformBufferIndex] != bufferStorage->getSerial())
{
mDeviceContext->VSSetConstantBuffers(getReservedVertexUniformBuffers() + uniformBufferIndex,
1, &constantBuffer);
mCurrentConstantBufferVS[uniformBufferIndex] = bufferStorage->getSerial();
}
}
}
for (unsigned int uniformBufferIndex = 0; uniformBufferIndex < gl::IMPLEMENTATION_MAX_FRAGMENT_SHADER_UNIFORM_BUFFERS; uniformBufferIndex++)
{
const gl::Buffer *uniformBuffer = fragmentUniformBuffers[uniformBufferIndex];
if (uniformBuffer)
{
BufferStorage11 *bufferStorage = BufferStorage11::makeBufferStorage11(uniformBuffer->getStorage());
ID3D11Buffer *constantBuffer = bufferStorage->getBuffer(BUFFER_USAGE_UNIFORM);
if (!constantBuffer)
{
return false;
}
if (mCurrentConstantBufferPS[uniformBufferIndex] != bufferStorage->getSerial())
{
mDeviceContext->PSSetConstantBuffers(getReservedFragmentUniformBuffers() + uniformBufferIndex,
1, &constantBuffer);
mCurrentConstantBufferPS[uniformBufferIndex] = bufferStorage->getSerial();
}
}
}
return true;
}
void Renderer11::setRasterizerState(const gl::RasterizerState &rasterState)
{
if (mForceSetRasterState || memcmp(&rasterState, &mCurRasterState, sizeof(gl::RasterizerState)) != 0)
{
ID3D11RasterizerState *dxRasterState = mStateCache.getRasterizerState(rasterState, mScissorEnabled);
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(gl::Framebuffer *framebuffer, const gl::BlendState &blendState, const gl::ColorF &blendColor,
unsigned int sampleMask)
{
if (mForceSetBlendState ||
memcmp(&blendState, &mCurBlendState, sizeof(gl::BlendState)) != 0 ||
memcmp(&blendColor, &mCurBlendColor, sizeof(gl::ColorF)) != 0 ||
sampleMask != mCurSampleMask)
{
ID3D11BlendState *dxBlendState = mStateCache.getBlendState(framebuffer, blendState);
if (!dxBlendState)
{
ERR("NULL blend state returned by RenderStateCache::getBlendState, setting the default "
"blend state.");
}
float blendColors[4] = {0.0f};
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)
{
blendColors[0] = blendColor.red;
blendColors[1] = blendColor.green;
blendColors[2] = blendColor.blue;
blendColors[3] = blendColor.alpha;
}
else
{
blendColors[0] = blendColor.alpha;
blendColors[1] = blendColor.alpha;
blendColors[2] = blendColor.alpha;
blendColors[3] = 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.");
}
// Max D3D11 stencil reference value is 0xFF, corresponding to the max 8 bits in a stencil buffer
// GL specifies we should clamp the ref value to the nearest bit depth when doing stencil ops
META_ASSERT(D3D11_DEFAULT_STENCIL_READ_MASK == 0xFF);
META_ASSERT(D3D11_DEFAULT_STENCIL_WRITE_MASK == 0xFF);
UINT dxStencilRef = std::min<UINT>(stencilRef, 0xFFu);
mDeviceContext->OMSetDepthStencilState(dxDepthStencilState, dxStencilRef);
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;
GLsizei minCount = 0;
switch (mode)
{
case GL_POINTS: primitiveTopology = D3D11_PRIMITIVE_TOPOLOGY_POINTLIST; minCount = 1; break;
case GL_LINES: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINELIST; minCount = 2; break;
case GL_LINE_LOOP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINESTRIP; minCount = 2; break;
case GL_LINE_STRIP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINESTRIP; minCount = 2; break;
case GL_TRIANGLES: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; minCount = 3; break;
case GL_TRIANGLE_STRIP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP; minCount = 3; break;
// emulate fans via rewriting index buffer
case GL_TRIANGLE_FAN: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; minCount = 3; break;
default:
return gl::error(GL_INVALID_ENUM, false);
}
if (primitiveTopology != mCurrentPrimitiveTopology)
{
mDeviceContext->IASetPrimitiveTopology(primitiveTopology);
mCurrentPrimitiveTopology = primitiveTopology;
}
return count >= minCount;
}
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::FramebufferAttachment *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;
}
// TODO: Detect if this color buffer is already bound as a texture and unbind it first to prevent
// D3D11 warnings.
}
}
// Get the depth stencil render buffer and serials
gl::FramebufferAttachment *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();
}
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();
}
}
// 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;
mForceSetBlendState = true;
if (!mDepthStencilInitialized)
{
mForceSetRasterState = true;
}
for (unsigned int rtIndex = 0; rtIndex < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; rtIndex++)
{
mAppliedRenderTargetSerials[rtIndex] = renderTargetSerials[rtIndex];
}
mAppliedDepthbufferSerial = depthbufferSerial;
mAppliedStencilbufferSerial = stencilbufferSerial;
mRenderTargetDescInitialized = true;
mDepthStencilInitialized = true;
}
invalidateFramebufferSwizzles(framebuffer);
return true;
}
GLenum Renderer11::applyVertexBuffer(gl::ProgramBinary *programBinary, const gl::VertexAttribute vertexAttributes[], gl::VertexAttribCurrentValueData currentValues[],
GLint first, GLsizei count, GLsizei instances)
{
TranslatedAttribute attributes[gl::MAX_VERTEX_ATTRIBS];
GLenum err = mVertexDataManager->prepareVertexData(vertexAttributes, currentValues, 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)
{
ID3D11Buffer *buffer = NULL;
DXGI_FORMAT bufferFormat = (indexInfo->indexType == GL_UNSIGNED_INT) ? DXGI_FORMAT_R32_UINT : DXGI_FORMAT_R16_UINT;
if (indexInfo->storage)
{
BufferStorage11 *storage = BufferStorage11::makeBufferStorage11(indexInfo->storage);
buffer = storage->getBuffer(BUFFER_USAGE_INDEX);
}
else
{
IndexBuffer11* indexBuffer = IndexBuffer11::makeIndexBuffer11(indexInfo->indexBuffer);
buffer = indexBuffer->getBuffer();
}
if (buffer != mAppliedIB || bufferFormat != mAppliedIBFormat || indexInfo->startOffset != mAppliedIBOffset)
{
mDeviceContext->IASetIndexBuffer(buffer, bufferFormat, indexInfo->startOffset);
mAppliedIB = buffer;
mAppliedIBFormat = bufferFormat;
mAppliedIBOffset = indexInfo->startOffset;
}
}
return err;
}
void Renderer11::applyTransformFeedbackBuffers(gl::Buffer *transformFeedbackBuffers[], GLintptr offsets[])
{
ID3D11Buffer* d3dBuffers[gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS];
UINT d3dOffsets[gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS];
bool requiresUpdate = false;
for (size_t i = 0; i < gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS; i++)
{
if (transformFeedbackBuffers[i])
{
BufferStorage11 *storage = BufferStorage11::makeBufferStorage11(transformFeedbackBuffers[i]->getStorage());
ID3D11Buffer *buffer = storage->getBuffer(BUFFER_USAGE_VERTEX_OR_TRANSFORM_FEEDBACK);
d3dBuffers[i] = buffer;
d3dOffsets[i] = (mAppliedTFBuffers[i] != buffer) ? static_cast<UINT>(offsets[i]) : -1;
}
else
{
d3dBuffers[i] = NULL;
d3dOffsets[i] = 0;
}
if (d3dBuffers[i] != mAppliedTFBuffers[i] || offsets[i] != mAppliedTFOffsets[i])
{
requiresUpdate = true;
}
}
if (requiresUpdate)
{
mDeviceContext->SOSetTargets(ArraySize(d3dBuffers), d3dBuffers, d3dOffsets);
for (size_t i = 0; i < gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS; i++)
{
mAppliedTFBuffers[i] = d3dBuffers[i];
mAppliedTFOffsets[i] = offsets[i];
}
}
}
void Renderer11::drawArrays(GLenum mode, GLsizei count, GLsizei instances, bool transformFeedbackActive)
{
if (mode == GL_POINTS && transformFeedbackActive)
{
// Since point sprites are generated with a geometry shader, too many vertices will
// be written if transform feedback is active. To work around this, draw only the points
// with the stream out shader and no pixel shader to feed the stream out buffers and then
// draw again with the point sprite geometry shader to rasterize the point sprites.
mDeviceContext->PSSetShader(NULL, NULL, 0);
if (instances > 0)
{
mDeviceContext->DrawInstanced(count, instances, 0, 0);
}
else
{
mDeviceContext->Draw(count, 0);
}
mDeviceContext->GSSetShader(mCurPointGeometryShader, NULL, 0);
mDeviceContext->PSSetShader(mAppliedPixelShader, NULL, 0);
if (instances > 0)
{
mDeviceContext->DrawInstanced(count, instances, 0, 0);
}
else
{
mDeviceContext->Draw(count, 0);
}
mDeviceContext->GSSetShader(mAppliedGeometryShader, NULL, 0);
}
else 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);
}
}
// Checked by Renderer11::applyPrimitiveType
ASSERT(count >= 0);
if (static_cast<unsigned int>(count) + 1 > (std::numeric_limits<unsigned int>::max() / sizeof(unsigned int)))
{
ERR("Could not create a 32-bit looping index buffer for GL_LINE_LOOP, too many indices required.");
return gl::error(GL_OUT_OF_MEMORY);
}
const unsigned int spaceNeeded = (static_cast<unsigned int>(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;
unsigned int offset;
if (!mLineLoopIB->mapBuffer(spaceNeeded, &mappedMemory, &offset))
{
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 = 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);
}
IndexBuffer11 *indexBuffer = IndexBuffer11::makeIndexBuffer11(mLineLoopIB->getIndexBuffer());
ID3D11Buffer *d3dIndexBuffer = indexBuffer->getBuffer();
DXGI_FORMAT indexFormat = indexBuffer->getIndexFormat();
if (mAppliedIB != d3dIndexBuffer || mAppliedIBFormat != indexFormat || mAppliedIBOffset != indexBufferOffset)
{
mDeviceContext->IASetIndexBuffer(d3dIndexBuffer, indexFormat, indexBufferOffset);
mAppliedIB = d3dIndexBuffer;
mAppliedIBFormat = indexFormat;
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);
}
}
// Checked by Renderer11::applyPrimitiveType
ASSERT(count >= 3);
const unsigned int numTris = count - 2;
if (numTris > (std::numeric_limits<unsigned int>::max() / (sizeof(unsigned int) * 3)))
{
ERR("Could not create a scratch index buffer for GL_TRIANGLE_FAN, too many indices required.");
return gl::error(GL_OUT_OF_MEMORY);
}
const unsigned 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;
unsigned int offset;
if (!mTriangleFanIB->mapBuffer(spaceNeeded, &mappedMemory, &offset))
{
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 = offset;
switch (type)
{
case GL_NONE: // Non-indexed draw
for (unsigned 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 (unsigned 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 (unsigned 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 (unsigned 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);
}
IndexBuffer11 *indexBuffer = IndexBuffer11::makeIndexBuffer11(mTriangleFanIB->getIndexBuffer());
ID3D11Buffer *d3dIndexBuffer = indexBuffer->getBuffer();
DXGI_FORMAT indexFormat = indexBuffer->getIndexFormat();
if (mAppliedIB != d3dIndexBuffer || mAppliedIBFormat != indexFormat || mAppliedIBOffset != indexBufferOffset)
{
mDeviceContext->IASetIndexBuffer(d3dIndexBuffer, indexFormat, indexBufferOffset);
mAppliedIB = d3dIndexBuffer;
mAppliedIBFormat = indexFormat;
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, bool rasterizerDiscard, bool transformFeedbackActive, const gl::VertexFormat inputLayout[])
{
ShaderExecutable *vertexExe = programBinary->getVertexExecutableForInputLayout(inputLayout);
ShaderExecutable *pixelExe = programBinary->getPixelExecutable();
ShaderExecutable *geometryExe = programBinary->getGeometryExecutable();
ID3D11VertexShader *vertexShader = (vertexExe ? ShaderExecutable11::makeShaderExecutable11(vertexExe)->getVertexShader() : NULL);
ID3D11PixelShader *pixelShader = NULL;
// Skip pixel shader if we're doing rasterizer discard.
if (!rasterizerDiscard)
{
pixelShader = (pixelExe ? ShaderExecutable11::makeShaderExecutable11(pixelExe)->getPixelShader() : NULL);
}
ID3D11GeometryShader *geometryShader = NULL;
if (transformFeedbackActive)
{
geometryShader = (vertexExe ? ShaderExecutable11::makeShaderExecutable11(vertexExe)->getStreamOutShader() : NULL);
}
else if (mCurRasterState.pointDrawMode)
{
geometryShader = (geometryExe ? ShaderExecutable11::makeShaderExecutable11(geometryExe)->getGeometryShader() : NULL);
}
bool dirtyUniforms = false;
if (vertexShader != mAppliedVertexShader)
{
mDeviceContext->VSSetShader(vertexShader, NULL, 0);
mAppliedVertexShader = vertexShader;
dirtyUniforms = true;
}
if (geometryShader != mAppliedGeometryShader)
{
mDeviceContext->GSSetShader(geometryShader, NULL, 0);
mAppliedGeometryShader = geometryShader;
dirtyUniforms = true;
}
if (geometryExe && mCurRasterState.pointDrawMode)
{
mCurPointGeometryShader = ShaderExecutable11::makeShaderExecutable11(geometryExe)->getGeometryShader();
}
else
{
mCurPointGeometryShader = NULL;
}
if (pixelShader != mAppliedPixelShader)
{
mDeviceContext->PSSetShader(pixelShader, NULL, 0);
mAppliedPixelShader = pixelShader;
dirtyUniforms = true;
}
if (dirtyUniforms)
{
programBinary->dirtyAllUniforms();
}
}
void Renderer11::applyUniforms(const gl::ProgramBinary &programBinary)
{
const std::vector<gl::LinkedUniform*> &uniformArray = programBinary.getUniforms();
unsigned int totalRegisterCountVS = 0;
unsigned int totalRegisterCountPS = 0;
bool vertexUniformsDirty = false;
bool pixelUniformsDirty = false;
for (size_t uniformIndex = 0; uniformIndex < uniformArray.size(); uniformIndex++)
{
const gl::LinkedUniform &uniform = *uniformArray[uniformIndex];
if (uniform.isReferencedByVertexShader() && !uniform.isSampler())
{
totalRegisterCountVS += uniform.registerCount;
vertexUniformsDirty = (vertexUniformsDirty || uniform.dirty);
}
if (uniform.isReferencedByFragmentShader() && !uniform.isSampler())
{
totalRegisterCountPS += uniform.registerCount;
pixelUniformsDirty = (pixelUniformsDirty || uniform.dirty);
}
}
const UniformStorage11 *vertexUniformStorage = UniformStorage11::makeUniformStorage11(&programBinary.getVertexUniformStorage());
const UniformStorage11 *fragmentUniformStorage = UniformStorage11::makeUniformStorage11(&programBinary.getFragmentUniformStorage());
ASSERT(vertexUniformStorage);
ASSERT(fragmentUniformStorage);
ID3D11Buffer *vertexConstantBuffer = vertexUniformStorage->getConstantBuffer();
ID3D11Buffer *pixelConstantBuffer = fragmentUniformStorage->getConstantBuffer();
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);
UNUSED_ASSERTION_VARIABLE(result);
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);
UNUSED_ASSERTION_VARIABLE(result);
ASSERT(SUCCEEDED(result));
mapPS = (float(*)[4])map.pData;
}
for (size_t uniformIndex = 0; uniformIndex < uniformArray.size(); uniformIndex++)
{
gl::LinkedUniform *uniform = uniformArray[uniformIndex];
if (!uniform->isSampler())
{
unsigned int componentCount = (4 - uniform->registerElement);
// we assume that uniforms from structs are arranged in struct order in our uniforms list. otherwise we would
// overwrite previously written regions of memory.
if (uniform->isReferencedByVertexShader() && mapVS)
{
memcpy(&mapVS[uniform->vsRegisterIndex][uniform->registerElement], uniform->data, uniform->registerCount * sizeof(float) * componentCount);
}
if (uniform->isReferencedByFragmentShader() && mapPS)
{
memcpy(&mapPS[uniform->psRegisterIndex][uniform->registerElement], uniform->data, uniform->registerCount * sizeof(float) * componentCount);
}
}
}
if (mapVS)
{
mDeviceContext->Unmap(vertexConstantBuffer, 0);
}
if (mapPS)
{
mDeviceContext->Unmap(pixelConstantBuffer, 0);
}
if (mCurrentVertexConstantBuffer != vertexConstantBuffer)
{
mDeviceContext->VSSetConstantBuffers(0, 1, &vertexConstantBuffer);
mCurrentVertexConstantBuffer = vertexConstantBuffer;
}
if (mCurrentPixelConstantBuffer != pixelConstantBuffer)
{
mDeviceContext->PSSetConstantBuffers(0, 1, &pixelConstantBuffer);
mCurrentPixelConstantBuffer = 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);
UNUSED_ASSERTION_VARIABLE(result);
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);
UNUSED_ASSERTION_VARIABLE(result);
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
if (mCurrentGeometryConstantBuffer != mDriverConstantBufferPS)
{
mDeviceContext->GSSetConstantBuffers(0, 1, &mDriverConstantBufferPS);
mCurrentGeometryConstantBuffer = mDriverConstantBufferPS;
}
}
void Renderer11::clear(const gl::ClearParameters &clearParams, gl::Framebuffer *frameBuffer)
{
mClear->clearFramebuffer(clearParams, frameBuffer);
invalidateFramebufferSwizzles(frameBuffer);
}
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;
mCurVertexSRVs[i] = NULL;
}
for (int i = 0; i < gl::MAX_TEXTURE_IMAGE_UNITS; i++)
{
mForceSetPixelSamplerStates[i] = true;
mCurPixelSRVs[i] = NULL;
}
mForceSetBlendState = true;
mForceSetRasterState = true;
mForceSetDepthStencilState = true;
mForceSetScissor = true;
mForceSetViewport = true;
mAppliedIB = NULL;
mAppliedIBFormat = DXGI_FORMAT_UNKNOWN;
mAppliedIBOffset = 0;
mAppliedVertexShader = NULL;
mAppliedGeometryShader = NULL;
mCurPointGeometryShader = NULL;
mAppliedPixelShader = NULL;
for (size_t i = 0; i < gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS; i++)
{
mAppliedTFBuffers[i] = NULL;
mAppliedTFOffsets[i] = 0;
}
memset(&mAppliedVertexConstants, 0, sizeof(dx_VertexConstants));
memset(&mAppliedPixelConstants, 0, sizeof(dx_PixelConstants));
mInputLayoutCache.markDirty();
for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_VERTEX_SHADER_UNIFORM_BUFFERS; i++)
{
mCurrentConstantBufferVS[i] = -1;
mCurrentConstantBufferPS[i] = -1;
}
mCurrentVertexConstantBuffer = NULL;
mCurrentPixelConstantBuffer = NULL;
mCurrentGeometryConstantBuffer = NULL;
mCurrentPrimitiveTopology = D3D_PRIMITIVE_TOPOLOGY_UNDEFINED;
}
void Renderer11::releaseDeviceResources()
{
mStateCache.clear();
mInputLayoutCache.clear();
SafeDelete(mVertexDataManager);
SafeDelete(mIndexDataManager);
SafeDelete(mLineLoopIB);
SafeDelete(mTriangleFanIB);
SafeDelete(mBlit);
SafeDelete(mClear);
SafeDelete(mPixelTransfer);
SafeRelease(mDriverConstantBufferVS);
SafeRelease(mDriverConstantBufferPS);
SafeRelease(mSyncQuery);
}
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;
}
SafeRelease(dummyContext);
SafeRelease(dummyDevice);
return true;
}
void Renderer11::release()
{
releaseDeviceResources();
SafeRelease(mDxgiFactory);
SafeRelease(mDxgiAdapter);
if (mDeviceContext)
{
mDeviceContext->ClearState();
mDeviceContext->Flush();
SafeRelease(mDeviceContext);
}
SafeRelease(mDevice);
if (mD3d11Module)
{
FreeLibrary(mD3d11Module);
mD3d11Module = NULL;
}
if (mDxgiModule)
{
FreeLibrary(mDxgiModule);
mDxgiModule = NULL;
}
mCompiler.release();
}
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() const
{
return mDXT1TextureSupport;
}
bool Renderer11::getDXT3TextureSupport() const
{
return mDXT3TextureSupport;
}
bool Renderer11::getDXT5TextureSupport() const
{
return mDXT5TextureSupport;
}
bool Renderer11::getDepthTextureSupport() const
{
return mDepthTextureSupport;
}
bool Renderer11::getFloat32TextureSupport() const
{
return mFloat32TextureSupport;
}
bool Renderer11::getFloat32TextureFilteringSupport() const
{
return mFloat32FilterSupport;
}
bool Renderer11::getFloat32TextureRenderingSupport() const
{
return mFloat32RenderSupport;
}
bool Renderer11::getFloat16TextureSupport() const
{
return mFloat16TextureSupport;
}
bool Renderer11::getFloat16TextureFilteringSupport() const
{
return mFloat16FilterSupport;
}
bool Renderer11::getFloat16TextureRenderingSupport() const
{
return mFloat16RenderSupport;
}
bool Renderer11::getRGB565TextureSupport() const
{
return false;
}
bool Renderer11::getLuminanceTextureSupport() const
{
return false;
}
bool Renderer11::getLuminanceAlphaTextureSupport() const
{
return false;
}
bool Renderer11::getRGTextureSupport() const
{
return mRGTextureSupport;
}
bool Renderer11::getTextureFilterAnisotropySupport() const
{
return true;
}
bool Renderer11::getPBOSupport() 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() const
{
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);
META_ASSERT(D3D11_VS_OUTPUT_REGISTER_COUNT <= D3D11_PS_INPUT_REGISTER_COUNT);
META_ASSERT(D3D10_VS_OUTPUT_REGISTER_COUNT <= D3D10_PS_INPUT_REGISTER_COUNT);
switch (mFeatureLevel)
{
case D3D_FEATURE_LEVEL_11_0:
return D3D11_VS_OUTPUT_REGISTER_COUNT - getReservedVaryings();
case D3D_FEATURE_LEVEL_10_1:
return D3D10_1_VS_OUTPUT_REGISTER_COUNT - getReservedVaryings();
case D3D_FEATURE_LEVEL_10_0:
return D3D10_VS_OUTPUT_REGISTER_COUNT - getReservedVaryings();
default: UNREACHABLE();
return 0;
}
}
unsigned int Renderer11::getMaxVertexShaderUniformBuffers() const
{
META_ASSERT(gl::IMPLEMENTATION_MAX_VERTEX_SHADER_UNIFORM_BUFFERS >= D3D10_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT &&
gl::IMPLEMENTATION_MAX_VERTEX_SHADER_UNIFORM_BUFFERS >= D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT);
switch (mFeatureLevel)
{
case D3D_FEATURE_LEVEL_11_0:
return D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT - getReservedVertexUniformBuffers();
case D3D_FEATURE_LEVEL_10_1:
case D3D_FEATURE_LEVEL_10_0:
return D3D10_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT - getReservedVertexUniformBuffers();
default: UNREACHABLE();
return 0;
}
}
unsigned int Renderer11::getMaxFragmentShaderUniformBuffers() const
{
META_ASSERT(gl::IMPLEMENTATION_MAX_FRAGMENT_SHADER_UNIFORM_BUFFERS >= D3D10_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT &&
gl::IMPLEMENTATION_MAX_FRAGMENT_SHADER_UNIFORM_BUFFERS >= D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT);
switch (mFeatureLevel)
{
case D3D_FEATURE_LEVEL_11_0:
return D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT - getReservedFragmentUniformBuffers();
case D3D_FEATURE_LEVEL_10_1:
case D3D_FEATURE_LEVEL_10_0:
return D3D10_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT - getReservedFragmentUniformBuffers();
default: UNREACHABLE();
return 0;
}
}
unsigned int Renderer11::getReservedVertexUniformBuffers() const
{
// we reserve one buffer for the application uniforms, and one for driver uniforms
return 2;
}
unsigned int Renderer11::getReservedFragmentUniformBuffers() const
{
// we reserve one buffer for the application uniforms, and one for driver uniforms
return 2;
}
unsigned int Renderer11::getReservedVaryings() const
{
// We potentially reserve varyings for gl_Position, _dx_Position, gl_FragCoord and gl_PointSize
return 4;
}
unsigned int Renderer11::getMaxTransformFeedbackBuffers() const
{
META_ASSERT(gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS >= D3D11_SO_BUFFER_SLOT_COUNT &&
gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS >= D3D10_SO_BUFFER_SLOT_COUNT);
switch (mFeatureLevel)
{
case D3D_FEATURE_LEVEL_11_0:
return D3D11_SO_BUFFER_SLOT_COUNT;
case D3D_FEATURE_LEVEL_10_1:
return D3D10_1_SO_BUFFER_SLOT_COUNT;
case D3D_FEATURE_LEVEL_10_0:
return D3D10_SO_BUFFER_SLOT_COUNT;
default: UNREACHABLE();
return 0;
}
}
unsigned int Renderer11::getMaxTransformFeedbackSeparateComponents() const
{
switch (mFeatureLevel)
{
case D3D_FEATURE_LEVEL_11_0:
return getMaxTransformFeedbackInterleavedComponents() / getMaxTransformFeedbackBuffers();
case D3D_FEATURE_LEVEL_10_1:
case D3D_FEATURE_LEVEL_10_0:
// D3D 10 and 10.1 only allow one output per output slot if an output slot other than zero
// is used.
return 4;
default: UNREACHABLE();
return 0;
}
}
unsigned int Renderer11::getMaxTransformFeedbackInterleavedComponents() const
{
return (getMaxVaryingVectors() * 4);
}
unsigned int Renderer11::getMaxUniformBufferSize() const
{
// Each component is a 4-element vector of 4-byte units (floats)
const unsigned int bytesPerComponent = 4 * sizeof(float);
switch (mFeatureLevel)
{
case D3D_FEATURE_LEVEL_11_0:
return D3D11_REQ_CONSTANT_BUFFER_ELEMENT_COUNT * bytesPerComponent;
case D3D_FEATURE_LEVEL_10_1:
case D3D_FEATURE_LEVEL_10_0:
return D3D10_REQ_CONSTANT_BUFFER_ELEMENT_COUNT * bytesPerComponent;
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::getMaxRecommendedElementsIndices() const
{
META_ASSERT(D3D11_REQ_DRAWINDEXED_INDEX_COUNT_2_TO_EXP == 32);
META_ASSERT(D3D10_REQ_DRAWINDEXED_INDEX_COUNT_2_TO_EXP == 32);
// D3D11 allows up to 2^32 elements, but we report max signed int for convenience.
return std::numeric_limits<GLint>::max();
}
int Renderer11::getMaxRecommendedElementsVertices() const
{
META_ASSERT(D3D11_REQ_DRAW_VERTEX_COUNT_2_TO_EXP == 32);
META_ASSERT(D3D10_REQ_DRAW_VERTEX_COUNT_2_TO_EXP == 32);
// D3D11 allows up to 2^32 elements, but we report max signed int for convenience.
return std::numeric_limits<GLint>::max();
}
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;
}
}
int Renderer11::getMaxTextureDepth() const
{
switch (mFeatureLevel)
{
case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_TEXTURE3D_U_V_OR_W_DIMENSION; // 2048
case D3D_FEATURE_LEVEL_10_1:
case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_TEXTURE3D_U_V_OR_W_DIMENSION; // 2048
default: UNREACHABLE(); return 0;
}
}
int Renderer11::getMaxTextureArrayLayers() const
{
switch (mFeatureLevel)
{
case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_TEXTURE2D_ARRAY_AXIS_DIMENSION; // 2048
case D3D_FEATURE_LEVEL_10_1:
case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_TEXTURE2D_ARRAY_AXIS_DIMENSION; // 512
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;
}
GLsizei Renderer11::getMaxSupportedFormatSamples(GLenum internalFormat) const
{
DXGI_FORMAT format = gl_d3d11::GetRenderableFormat(internalFormat, getCurrentClientVersion());
MultisampleSupportMap::const_iterator iter = mMultisampleSupportMap.find(format);
return (iter != mMultisampleSupportMap.end()) ? iter->second.maxSupportedSamples : 0;
}
GLsizei Renderer11::getNumSampleCounts(GLenum internalFormat) const
{
unsigned int numCounts = 0;
// D3D11 supports multisampling for signed and unsigned format, but ES 3.0 does not
GLenum componentType = gl::GetComponentType(internalFormat, getCurrentClientVersion());
if (componentType != GL_INT && componentType != GL_UNSIGNED_INT)
{
DXGI_FORMAT format = gl_d3d11::GetRenderableFormat(internalFormat, getCurrentClientVersion());
MultisampleSupportMap::const_iterator iter = mMultisampleSupportMap.find(format);
if (iter != mMultisampleSupportMap.end())
{
const MultisampleSupportInfo& info = iter->second;
for (int i = 0; i < D3D11_MAX_MULTISAMPLE_SAMPLE_COUNT; i++)
{
if (info.qualityLevels[i] > 0)
{
numCounts++;
}
}
}
}
return numCounts;
}
void Renderer11::getSampleCounts(GLenum internalFormat, GLsizei bufSize, GLint *params) const
{
// D3D11 supports multisampling for signed and unsigned format, but ES 3.0 does not
GLenum componentType = gl::GetComponentType(internalFormat, getCurrentClientVersion());
if (componentType == GL_INT || componentType == GL_UNSIGNED_INT)
{
return;
}
DXGI_FORMAT format = gl_d3d11::GetRenderableFormat(internalFormat, getCurrentClientVersion());
MultisampleSupportMap::const_iterator iter = mMultisampleSupportMap.find(format);
if (iter != mMultisampleSupportMap.end())
{
const MultisampleSupportInfo& info = iter->second;
int bufPos = 0;
for (int i = D3D11_MAX_MULTISAMPLE_SAMPLE_COUNT - 1; i >= 0 && bufPos < bufSize; i--)
{
if (info.qualityLevels[i] > 0)
{
params[bufPos++] = i + 1;
}
}
}
}
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:
// Feature level 10.0 and 10.1 cards perform very poorly when the pixel shader
// outputs to multiple RTs that are not bound.
// TODO: Remove pixel shader outputs for render targets that are not bound.
return 1;
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->getResource(), source11->getResource());
dest11->invalidateSwizzleCache();
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->getResource(), source11->getResource());
dest11->invalidateSwizzleCache();
return true;
}
return false;
}
bool Renderer11::copyToRenderTarget(TextureStorageInterface3D *dest, TextureStorageInterface3D *source)
{
if (source && dest)
{
TextureStorage11_3D *source11 = TextureStorage11_3D::makeTextureStorage11_3D(source->getStorageInstance());
TextureStorage11_3D *dest11 = TextureStorage11_3D::makeTextureStorage11_3D(dest->getStorageInstance());
mDeviceContext->CopyResource(dest11->getResource(), source11->getResource());
dest11->invalidateSwizzleCache();
return true;
}
return false;
}
bool Renderer11::copyToRenderTarget(TextureStorageInterface2DArray *dest, TextureStorageInterface2DArray *source)
{
if (source && dest)
{
TextureStorage11_2DArray *source11 = TextureStorage11_2DArray::makeTextureStorage11_2DArray(source->getStorageInstance());
TextureStorage11_2DArray *dest11 = TextureStorage11_2DArray::makeTextureStorage11_2DArray(dest->getStorageInstance());
mDeviceContext->CopyResource(dest11->getResource(), source11->getResource());
dest11->invalidateSwizzleCache();
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::FramebufferAttachment *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)
{
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)
{
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)
{
ERR("Failed to retrieve the render target view from the destination render target.");
return gl::error(GL_OUT_OF_MEMORY, false);
}
gl::Box sourceArea(sourceRect.x, sourceRect.y, 0, sourceRect.width, sourceRect.height, 1);
gl::Extents sourceSize(sourceRenderTarget->getWidth(), sourceRenderTarget->getHeight(), 1);
gl::Box destArea(xoffset, yoffset, 0, sourceRect.width, sourceRect.height, 1);
gl::Extents destSize(destRenderTarget->getWidth(), destRenderTarget->getHeight(), 1);
// Use nearest filtering because source and destination are the same size for the direct
// copy
bool ret = mBlit->copyTexture(source, sourceArea, sourceSize, dest, destArea, destSize, NULL,
destFormat, GL_NEAREST);
storage11->invalidateSwizzleCacheLevel(level);
return ret;
}
bool Renderer11::copyImage(gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat,
GLint xoffset, GLint yoffset, TextureStorageInterfaceCube *storage, GLenum target, GLint level)
{