gecko-dev/gfx/layers/ImageDataSerializer.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

296 lines
10 KiB
C++
Raw Normal View History

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "ImageDataSerializer.h"
#include "gfx2DGlue.h" // for SurfaceFormatToImageFormat
#include "mozilla/gfx/Point.h" // for IntSize
#include "mozilla/Assertions.h" // for MOZ_ASSERT, etc
#include "mozilla/gfx/2D.h" // for DataSourceSurface, Factory
#include "mozilla/gfx/Logging.h" // for gfxDebug
#include "mozilla/gfx/Tools.h" // for GetAlignedStride, etc
#include "mozilla/gfx/Types.h"
#include "mozilla/mozalloc.h" // for operator delete, etc
#include "YCbCrUtils.h" // for YCbCr conversions
namespace mozilla {
namespace layers {
namespace ImageDataSerializer {
using namespace gfx;
int32_t ComputeRGBStride(SurfaceFormat aFormat, int32_t aWidth) {
Bug 1479145 - Give RGB textures a 32-byte aligned stride on macOS in order to improve texture upload efficiency on certain drivers. r=mattwoodrow In particular, it looks like this alignment is required by the Intel driver on macOS if you want to avoid CPU copies. It was already known that the efficiency gains from client storage only materialize if you follow certain restrictions: - The textures need to use the TEXTURE_RECTANGLE_ARB texture target. - The textures' format, internalFormat and type need to be chosen from a small list of supported configurations. Unsupported configurations will trigger format conversions on the CPU. - The GL_TEXTURE_STORAGE_HINT_APPLE may need to be set to shared or cached. - glTextureRangeAPPLE may or may not make a difference. It now appears that the stride alignment is another requirement: When uploading textures which otherwise comply with the above requirements, the Intel driver will still make copies using the CPU if the texture's stride is not 32-byte aligned. These CPU copies are reflected in a high CPU usage (as observed in Activity Monitor) and they show up in profiles as time spent inside _platform_memmove under glrUpdateTexture. However, when uploading 32-byte stride aligned textures which comply with the above requirements, this CPU usage goes away. There might still be hardware copies behind the scenes, but they no longer take up CPU time. Differential Revision: https://phabricator.services.mozilla.com/D25316 --HG-- extra : moz-landing-system : lando
2019-03-29 20:11:12 +00:00
#ifdef XP_MACOSX
// Some drivers require an alignment of 32 bytes for efficient texture upload.
return GetAlignedStride<32>(aWidth, BytesPerPixel(aFormat));
#else
return GetAlignedStride<4>(aWidth, BytesPerPixel(aFormat));
Bug 1479145 - Give RGB textures a 32-byte aligned stride on macOS in order to improve texture upload efficiency on certain drivers. r=mattwoodrow In particular, it looks like this alignment is required by the Intel driver on macOS if you want to avoid CPU copies. It was already known that the efficiency gains from client storage only materialize if you follow certain restrictions: - The textures need to use the TEXTURE_RECTANGLE_ARB texture target. - The textures' format, internalFormat and type need to be chosen from a small list of supported configurations. Unsupported configurations will trigger format conversions on the CPU. - The GL_TEXTURE_STORAGE_HINT_APPLE may need to be set to shared or cached. - glTextureRangeAPPLE may or may not make a difference. It now appears that the stride alignment is another requirement: When uploading textures which otherwise comply with the above requirements, the Intel driver will still make copies using the CPU if the texture's stride is not 32-byte aligned. These CPU copies are reflected in a high CPU usage (as observed in Activity Monitor) and they show up in profiles as time spent inside _platform_memmove under glrUpdateTexture. However, when uploading 32-byte stride aligned textures which comply with the above requirements, this CPU usage goes away. There might still be hardware copies behind the scenes, but they no longer take up CPU time. Differential Revision: https://phabricator.services.mozilla.com/D25316 --HG-- extra : moz-landing-system : lando
2019-03-29 20:11:12 +00:00
#endif
}
int32_t GetRGBStride(const RGBDescriptor& aDescriptor) {
return ComputeRGBStride(aDescriptor.format(), aDescriptor.size().width);
}
uint32_t ComputeRGBBufferSize(IntSize aSize, SurfaceFormat aFormat) {
MOZ_ASSERT(aSize.height >= 0 && aSize.width >= 0);
// This takes care of checking whether there could be overflow
// with enough margin for the metadata.
if (!gfx::Factory::AllowedSurfaceSize(aSize)) {
return 0;
}
// Note we're passing height instad of the bpp parameter, but the end
// result is the same - and the bpp was already taken care of in the
// ComputeRGBStride function.
int32_t bufsize = GetAlignedStride<16>(ComputeRGBStride(aFormat, aSize.width),
aSize.height);
if (bufsize < 0) {
// This should not be possible thanks to Factory::AllowedSurfaceSize
return 0;
}
return bufsize;
}
// Minimum required shmem size in bytes
uint32_t ComputeYCbCrBufferSize(const gfx::IntSize& aYSize, int32_t aYStride,
const gfx::IntSize& aCbCrSize,
int32_t aCbCrStride) {
MOZ_ASSERT(aYSize.height >= 0 && aYSize.width >= 0);
if (aYSize.height < 0 || aYSize.width < 0 || aCbCrSize.height < 0 ||
aCbCrSize.width < 0 ||
!gfx::Factory::AllowedSurfaceSize(IntSize(aYStride, aYSize.height)) ||
!gfx::Factory::AllowedSurfaceSize(
IntSize(aCbCrStride, aCbCrSize.height))) {
return 0;
}
// Overflow checks are performed in AllowedSurfaceSize
return GetAlignedStride<4>(aYSize.height, aYStride) +
2 * GetAlignedStride<4>(aCbCrSize.height, aCbCrStride);
}
2018-05-22 15:25:49 +00:00
uint32_t ComputeYCbCrBufferSize(const gfx::IntSize& aYSize, int32_t aYStride,
const gfx::IntSize& aCbCrSize,
int32_t aCbCrStride, uint32_t aYOffset,
uint32_t aCbOffset, uint32_t aCrOffset) {
MOZ_ASSERT(aYSize.height >= 0 && aYSize.width >= 0);
if (aYSize.height < 0 || aYSize.width < 0 || aCbCrSize.height < 0 ||
aCbCrSize.width < 0 ||
!gfx::Factory::AllowedSurfaceSize(IntSize(aYStride, aYSize.height)) ||
!gfx::Factory::AllowedSurfaceSize(
IntSize(aCbCrStride, aCbCrSize.height))) {
return 0;
}
uint32_t yLength = GetAlignedStride<4>(aYStride, aYSize.height);
uint32_t cbCrLength = GetAlignedStride<4>(aCbCrStride, aCbCrSize.height);
if (yLength == 0 || cbCrLength == 0) {
return 0;
}
CheckedInt<uint32_t> yEnd = aYOffset;
yEnd += yLength;
CheckedInt<uint32_t> cbEnd = aCbOffset;
cbEnd += cbCrLength;
CheckedInt<uint32_t> crEnd = aCrOffset;
crEnd += cbCrLength;
if (!yEnd.isValid() || !cbEnd.isValid() || !crEnd.isValid() ||
yEnd.value() > aCbOffset || cbEnd.value() > aCrOffset) {
return 0;
}
return crEnd.value();
}
uint32_t ComputeYCbCrBufferSize(uint32_t aBufferSize) {
return GetAlignedStride<4>(aBufferSize, 1);
}
void ComputeYCbCrOffsets(int32_t yStride, int32_t yHeight, int32_t cbCrStride,
int32_t cbCrHeight, uint32_t& outYOffset,
uint32_t& outCbOffset, uint32_t& outCrOffset) {
outYOffset = 0;
outCbOffset = outYOffset + GetAlignedStride<4>(yStride, yHeight);
outCrOffset = outCbOffset + GetAlignedStride<4>(cbCrStride, cbCrHeight);
}
gfx::SurfaceFormat FormatFromBufferDescriptor(
const BufferDescriptor& aDescriptor) {
switch (aDescriptor.type()) {
case BufferDescriptor::TRGBDescriptor:
return aDescriptor.get_RGBDescriptor().format();
case BufferDescriptor::TYCbCrDescriptor:
return gfx::SurfaceFormat::YUV;
default:
MOZ_CRASH("GFX: FormatFromBufferDescriptor");
}
}
gfx::IntSize SizeFromBufferDescriptor(const BufferDescriptor& aDescriptor) {
switch (aDescriptor.type()) {
case BufferDescriptor::TRGBDescriptor:
return aDescriptor.get_RGBDescriptor().size();
case BufferDescriptor::TYCbCrDescriptor:
return aDescriptor.get_YCbCrDescriptor().ySize();
default:
MOZ_CRASH("GFX: SizeFromBufferDescriptor");
}
}
Maybe<gfx::IntSize> CbCrSizeFromBufferDescriptor(
const BufferDescriptor& aDescriptor) {
switch (aDescriptor.type()) {
case BufferDescriptor::TRGBDescriptor:
return Nothing();
case BufferDescriptor::TYCbCrDescriptor:
return Some(aDescriptor.get_YCbCrDescriptor().cbCrSize());
default:
MOZ_CRASH("GFX: CbCrSizeFromBufferDescriptor");
}
}
Maybe<gfx::YUVColorSpace> YUVColorSpaceFromBufferDescriptor(
const BufferDescriptor& aDescriptor) {
switch (aDescriptor.type()) {
case BufferDescriptor::TRGBDescriptor:
return Nothing();
case BufferDescriptor::TYCbCrDescriptor:
return Some(aDescriptor.get_YCbCrDescriptor().yUVColorSpace());
default:
MOZ_CRASH("GFX: YUVColorSpaceFromBufferDescriptor");
}
}
Maybe<gfx::ColorDepth> ColorDepthFromBufferDescriptor(
const BufferDescriptor& aDescriptor) {
switch (aDescriptor.type()) {
case BufferDescriptor::TRGBDescriptor:
return Nothing();
case BufferDescriptor::TYCbCrDescriptor:
return Some(aDescriptor.get_YCbCrDescriptor().colorDepth());
default:
MOZ_CRASH("GFX: ColorDepthFromBufferDescriptor");
}
}
Maybe<gfx::ColorRange> ColorRangeFromBufferDescriptor(
const BufferDescriptor& aDescriptor) {
switch (aDescriptor.type()) {
case BufferDescriptor::TRGBDescriptor:
return Nothing();
case BufferDescriptor::TYCbCrDescriptor:
return Some(aDescriptor.get_YCbCrDescriptor().colorRange());
default:
MOZ_CRASH("GFX: YUVFullRangeFromBufferDescriptor");
}
}
Maybe<StereoMode> StereoModeFromBufferDescriptor(
const BufferDescriptor& aDescriptor) {
switch (aDescriptor.type()) {
case BufferDescriptor::TRGBDescriptor:
return Nothing();
case BufferDescriptor::TYCbCrDescriptor:
return Some(aDescriptor.get_YCbCrDescriptor().stereoMode());
default:
MOZ_CRASH("GFX: StereoModeFromBufferDescriptor");
}
}
uint8_t* GetYChannel(uint8_t* aBuffer, const YCbCrDescriptor& aDescriptor) {
return aBuffer + aDescriptor.yOffset();
}
uint8_t* GetCbChannel(uint8_t* aBuffer, const YCbCrDescriptor& aDescriptor) {
return aBuffer + aDescriptor.cbOffset();
}
uint8_t* GetCrChannel(uint8_t* aBuffer, const YCbCrDescriptor& aDescriptor) {
return aBuffer + aDescriptor.crOffset();
}
already_AddRefed<DataSourceSurface> DataSourceSurfaceFromYCbCrDescriptor(
uint8_t* aBuffer, const YCbCrDescriptor& aDescriptor,
gfx::DataSourceSurface* aSurface) {
gfx::IntSize ySize = aDescriptor.ySize();
RefPtr<DataSourceSurface> result;
if (aSurface) {
MOZ_ASSERT(aSurface->GetSize() == ySize);
MOZ_ASSERT(aSurface->GetFormat() == gfx::SurfaceFormat::B8G8R8X8);
if (aSurface->GetSize() == ySize &&
aSurface->GetFormat() == gfx::SurfaceFormat::B8G8R8X8) {
result = aSurface;
}
}
if (!result) {
result =
Factory::CreateDataSourceSurface(ySize, gfx::SurfaceFormat::B8G8R8X8);
}
if (NS_WARN_IF(!result)) {
return nullptr;
}
DataSourceSurface::MappedSurface map;
if (NS_WARN_IF(!result->Map(DataSourceSurface::MapType::WRITE, &map))) {
return nullptr;
}
layers::PlanarYCbCrData ycbcrData;
ycbcrData.mYChannel = GetYChannel(aBuffer, aDescriptor);
ycbcrData.mYStride = aDescriptor.yStride();
ycbcrData.mYSize = ySize;
ycbcrData.mCbChannel = GetCbChannel(aBuffer, aDescriptor);
ycbcrData.mCrChannel = GetCrChannel(aBuffer, aDescriptor);
ycbcrData.mCbCrStride = aDescriptor.cbCrStride();
ycbcrData.mCbCrSize = aDescriptor.cbCrSize();
ycbcrData.mPicSize = ySize;
ycbcrData.mYUVColorSpace = aDescriptor.yUVColorSpace();
ycbcrData.mColorDepth = aDescriptor.colorDepth();
gfx::ConvertYCbCrToRGB(ycbcrData, gfx::SurfaceFormat::B8G8R8X8, ySize,
map.mData, map.mStride);
result->Unmap();
return result.forget();
}
void ConvertAndScaleFromYCbCrDescriptor(uint8_t* aBuffer,
const YCbCrDescriptor& aDescriptor,
const gfx::SurfaceFormat& aDestFormat,
const gfx::IntSize& aDestSize,
unsigned char* aDestBuffer,
int32_t aStride) {
MOZ_ASSERT(aBuffer);
layers::PlanarYCbCrData ycbcrData;
ycbcrData.mYChannel = GetYChannel(aBuffer, aDescriptor);
ycbcrData.mYStride = aDescriptor.yStride();
;
ycbcrData.mYSize = aDescriptor.ySize();
ycbcrData.mCbChannel = GetCbChannel(aBuffer, aDescriptor);
ycbcrData.mCrChannel = GetCrChannel(aBuffer, aDescriptor);
ycbcrData.mCbCrStride = aDescriptor.cbCrStride();
ycbcrData.mCbCrSize = aDescriptor.cbCrSize();
ycbcrData.mPicSize = aDescriptor.ySize();
ycbcrData.mYUVColorSpace = aDescriptor.yUVColorSpace();
ycbcrData.mColorDepth = aDescriptor.colorDepth();
gfx::ConvertYCbCrToRGB(ycbcrData, aDestFormat, aDestSize, aDestBuffer,
aStride);
}
} // namespace ImageDataSerializer
} // namespace layers
} // namespace mozilla