Bug 924102 - Add filter processing code for many SVG filters. r=Bas

This commit is contained in:
Markus Stange 2013-11-27 12:22:27 +01:00
parent 410c0c2218
commit 4c57389fc2
7 changed files with 1880 additions and 0 deletions

224
gfx/2d/FilterProcessing.cpp Normal file
View File

@ -0,0 +1,224 @@
/* -*- Mode: C++; tab-width: 20; indent-tabs-mode: nil; c-basic-offset: 2 -*-
* 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 "FilterProcessing.h"
namespace mozilla {
namespace gfx {
TemporaryRef<DataSourceSurface>
FilterProcessing::ExtractAlpha(DataSourceSurface* aSource)
{
IntSize size = aSource->GetSize();
RefPtr<DataSourceSurface> alpha = Factory::CreateDataSourceSurface(size, FORMAT_A8);
uint8_t* sourceData = aSource->GetData();
int32_t sourceStride = aSource->Stride();
uint8_t* alphaData = alpha->GetData();
int32_t alphaStride = alpha->Stride();
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
ExtractAlpha_SSE2(size, sourceData, sourceStride, alphaData, alphaStride);
#endif
} else {
ExtractAlpha_Scalar(size, sourceData, sourceStride, alphaData, alphaStride);
}
return alpha;
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ConvertToB8G8R8A8(SourceSurface* aSurface)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
return ConvertToB8G8R8A8_SSE2(aSurface);
#endif
}
return ConvertToB8G8R8A8_Scalar(aSurface);
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyBlending(DataSourceSurface* aInput1, DataSourceSurface* aInput2,
BlendMode aBlendMode)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
return ApplyBlending_SSE2(aInput1, aInput2, aBlendMode);
#endif
}
return ApplyBlending_Scalar(aInput1, aInput2, aBlendMode);
}
void
FilterProcessing::ApplyMorphologyHorizontal(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOp)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
ApplyMorphologyHorizontal_SSE2(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius, aOp);
#endif
} else {
ApplyMorphologyHorizontal_Scalar(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius, aOp);
}
}
void
FilterProcessing::ApplyMorphologyVertical(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOp)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
ApplyMorphologyVertical_SSE2(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius, aOp);
#endif
} else {
ApplyMorphologyVertical_Scalar(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius, aOp);
}
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyColorMatrix(DataSourceSurface* aInput, const Matrix5x4 &aMatrix)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
return ApplyColorMatrix_SSE2(aInput, aMatrix);
#endif
}
return ApplyColorMatrix_Scalar(aInput, aMatrix);
}
void
FilterProcessing::ApplyComposition(DataSourceSurface* aSource, DataSourceSurface* aDest,
CompositeOperator aOperator)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
ApplyComposition_SSE2(aSource, aDest, aOperator);
#endif
} else {
ApplyComposition_Scalar(aSource, aDest, aOperator);
}
}
void
FilterProcessing::SeparateColorChannels(DataSourceSurface* aSource,
RefPtr<DataSourceSurface>& aChannel0,
RefPtr<DataSourceSurface>& aChannel1,
RefPtr<DataSourceSurface>& aChannel2,
RefPtr<DataSourceSurface>& aChannel3)
{
IntSize size = aSource->GetSize();
aChannel0 = Factory::CreateDataSourceSurface(size, FORMAT_A8);
aChannel1 = Factory::CreateDataSourceSurface(size, FORMAT_A8);
aChannel2 = Factory::CreateDataSourceSurface(size, FORMAT_A8);
aChannel3 = Factory::CreateDataSourceSurface(size, FORMAT_A8);
uint8_t* sourceData = aSource->GetData();
int32_t sourceStride = aSource->Stride();
uint8_t* channel0Data = aChannel0->GetData();
uint8_t* channel1Data = aChannel1->GetData();
uint8_t* channel2Data = aChannel2->GetData();
uint8_t* channel3Data = aChannel3->GetData();
int32_t channelStride = aChannel0->Stride();
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
SeparateColorChannels_SSE2(size, sourceData, sourceStride, channel0Data, channel1Data, channel2Data, channel3Data, channelStride);
#endif
} else {
SeparateColorChannels_Scalar(size, sourceData, sourceStride, channel0Data, channel1Data, channel2Data, channel3Data, channelStride);
}
}
TemporaryRef<DataSourceSurface>
FilterProcessing::CombineColorChannels(DataSourceSurface* aChannel0, DataSourceSurface* aChannel1,
DataSourceSurface* aChannel2, DataSourceSurface* aChannel3)
{
IntSize size = aChannel0->GetSize();
RefPtr<DataSourceSurface> result =
Factory::CreateDataSourceSurface(size, FORMAT_B8G8R8A8);
int32_t resultStride = result->Stride();
uint8_t* resultData = result->GetData();
int32_t channelStride = aChannel0->Stride();
uint8_t* channel0Data = aChannel0->GetData();
uint8_t* channel1Data = aChannel1->GetData();
uint8_t* channel2Data = aChannel2->GetData();
uint8_t* channel3Data = aChannel3->GetData();
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
CombineColorChannels_SSE2(size, resultStride, resultData, channelStride, channel0Data, channel1Data, channel2Data, channel3Data);
#endif
} else {
CombineColorChannels_Scalar(size, resultStride, resultData, channelStride, channel0Data, channel1Data, channel2Data, channel3Data);
}
return result;
}
void
FilterProcessing::DoPremultiplicationCalculation(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
DoPremultiplicationCalculation_SSE2(
aSize, aTargetData, aTargetStride, aSourceData, aSourceStride);
#endif
} else {
DoPremultiplicationCalculation_Scalar(
aSize, aTargetData, aTargetStride, aSourceData, aSourceStride);
}
}
void
FilterProcessing::DoUnpremultiplicationCalculation(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
DoUnpremultiplicationCalculation_SSE2(
aSize, aTargetData, aTargetStride, aSourceData, aSourceStride);
#endif
} else {
DoUnpremultiplicationCalculation_Scalar(
aSize, aTargetData, aTargetStride, aSourceData, aSourceStride);
}
}
TemporaryRef<DataSourceSurface>
FilterProcessing::RenderTurbulence(const IntSize &aSize, const Point &aOffset, const Size &aBaseFrequency,
int32_t aSeed, int aNumOctaves, TurbulenceType aType, bool aStitch, const Rect &aTileRect)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
return RenderTurbulence_SSE2(aSize, aOffset, aBaseFrequency, aSeed, aNumOctaves, aType, aStitch, aTileRect);
#endif
}
return RenderTurbulence_Scalar(aSize, aOffset, aBaseFrequency, aSeed, aNumOctaves, aType, aStitch, aTileRect);
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyArithmeticCombine(DataSourceSurface* aInput1, DataSourceSurface* aInput2, Float aK1, Float aK2, Float aK3, Float aK4)
{
if (Factory::HasSSE2()) {
#ifdef USE_SSE2
return ApplyArithmeticCombine_SSE2(aInput1, aInput2, aK1, aK2, aK3, aK4);
#endif
}
return ApplyArithmeticCombine_Scalar(aInput1, aInput2, aK1, aK2, aK3, aK4);
}
} // namespace gfx
} // namespace mozilla

142
gfx/2d/FilterProcessing.h Normal file
View File

@ -0,0 +1,142 @@
/* -*- Mode: C++; tab-width: 20; indent-tabs-mode: nil; c-basic-offset: 2 -*-
* 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/. */
#ifndef _MOZILLA_GFX_FILTERPROCESSING_H_
#define _MOZILLA_GFX_FILTERPROCESSING_H_
#include "2D.h"
#include "Filters.h"
namespace mozilla {
namespace gfx {
const ptrdiff_t B8G8R8A8_COMPONENT_BYTEOFFSET_B = 0;
const ptrdiff_t B8G8R8A8_COMPONENT_BYTEOFFSET_G = 1;
const ptrdiff_t B8G8R8A8_COMPONENT_BYTEOFFSET_R = 2;
const ptrdiff_t B8G8R8A8_COMPONENT_BYTEOFFSET_A = 3;
class FilterProcessing
{
public:
// Fast approximate division by 255. It has the property that
// for all 0 <= v <= 255*255, FastDivideBy255(v) == v/255.
// But it only uses two adds and two shifts instead of an
// integer division (which is expensive on many processors).
template<class B, class A>
static B FastDivideBy255(A v)
{
return ((v << 8) + v + 255) >> 16;
}
static TemporaryRef<DataSourceSurface> ExtractAlpha(DataSourceSurface* aSource);
static TemporaryRef<DataSourceSurface> ConvertToB8G8R8A8(SourceSurface* aSurface);
static TemporaryRef<DataSourceSurface> ApplyBlending(DataSourceSurface* aInput1, DataSourceSurface* aInput2, BlendMode aBlendMode);
static void ApplyMorphologyHorizontal(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOperator);
static void ApplyMorphologyVertical(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOperator);
static TemporaryRef<DataSourceSurface> ApplyColorMatrix(DataSourceSurface* aInput, const Matrix5x4 &aMatrix);
static void ApplyComposition(DataSourceSurface* aSource, DataSourceSurface* aDest, CompositeOperator aOperator);
static void SeparateColorChannels(DataSourceSurface* aSource,
RefPtr<DataSourceSurface>& aChannel0,
RefPtr<DataSourceSurface>& aChannel1,
RefPtr<DataSourceSurface>& aChannel2,
RefPtr<DataSourceSurface>& aChannel3);
static TemporaryRef<DataSourceSurface>
CombineColorChannels(DataSourceSurface* aChannel0, DataSourceSurface* aChannel1,
DataSourceSurface* aChannel2, DataSourceSurface* aChannel3);
static void DoPremultiplicationCalculation(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride);
static void DoUnpremultiplicationCalculation(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride);
static TemporaryRef<DataSourceSurface>
RenderTurbulence(const IntSize &aSize, const Point &aOffset, const Size &aBaseFrequency,
int32_t aSeed, int aNumOctaves, TurbulenceType aType, bool aStitch, const Rect &aTileRect);
static TemporaryRef<DataSourceSurface>
ApplyArithmeticCombine(DataSourceSurface* aInput1, DataSourceSurface* aInput2, Float aK1, Float aK2, Float aK3, Float aK4);
protected:
static void ExtractAlpha_Scalar(const IntSize& size, uint8_t* sourceData, int32_t sourceStride, uint8_t* alphaData, int32_t alphaStride);
static TemporaryRef<DataSourceSurface> ConvertToB8G8R8A8_Scalar(SourceSurface* aSurface);
static TemporaryRef<DataSourceSurface> ApplyBlending_Scalar(DataSourceSurface* aInput1, DataSourceSurface* aInput2, BlendMode aBlendMode);
static void ApplyMorphologyHorizontal_Scalar(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOperator);
static void ApplyMorphologyVertical_Scalar(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOperator);
static TemporaryRef<DataSourceSurface> ApplyColorMatrix_Scalar(DataSourceSurface* aInput, const Matrix5x4 &aMatrix);
static void ApplyComposition_Scalar(DataSourceSurface* aSource, DataSourceSurface* aDest, CompositeOperator aOperator);
static void SeparateColorChannels_Scalar(const IntSize &size, uint8_t* sourceData, int32_t sourceStride, uint8_t* channel0Data, uint8_t* channel1Data, uint8_t* channel2Data, uint8_t* channel3Data, int32_t channelStride);
static void CombineColorChannels_Scalar(const IntSize &size, int32_t resultStride, uint8_t* resultData, int32_t channelStride, uint8_t* channel0Data, uint8_t* channel1Data, uint8_t* channel2Data, uint8_t* channel3Data);
static void DoPremultiplicationCalculation_Scalar(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride);
static void DoUnpremultiplicationCalculation_Scalar(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride);
static TemporaryRef<DataSourceSurface>
RenderTurbulence_Scalar(const IntSize &aSize, const Point &aOffset, const Size &aBaseFrequency,
int32_t aSeed, int aNumOctaves, TurbulenceType aType, bool aStitch, const Rect &aTileRect);
static TemporaryRef<DataSourceSurface>
ApplyArithmeticCombine_Scalar(DataSourceSurface* aInput1, DataSourceSurface* aInput2, Float aK1, Float aK2, Float aK3, Float aK4);
#ifdef USE_SSE2
static void ExtractAlpha_SSE2(const IntSize& size, uint8_t* sourceData, int32_t sourceStride, uint8_t* alphaData, int32_t alphaStride);
static TemporaryRef<DataSourceSurface> ConvertToB8G8R8A8_SSE2(SourceSurface* aSurface);
static TemporaryRef<DataSourceSurface> ApplyBlending_SSE2(DataSourceSurface* aInput1, DataSourceSurface* aInput2, BlendMode aBlendMode);
static void ApplyMorphologyHorizontal_SSE2(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOperator);
static void ApplyMorphologyVertical_SSE2(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOperator);
static TemporaryRef<DataSourceSurface> ApplyColorMatrix_SSE2(DataSourceSurface* aInput, const Matrix5x4 &aMatrix);
static void ApplyComposition_SSE2(DataSourceSurface* aSource, DataSourceSurface* aDest, CompositeOperator aOperator);
static void SeparateColorChannels_SSE2(const IntSize &size, uint8_t* sourceData, int32_t sourceStride, uint8_t* channel0Data, uint8_t* channel1Data, uint8_t* channel2Data, uint8_t* channel3Data, int32_t channelStride);
static void CombineColorChannels_SSE2(const IntSize &size, int32_t resultStride, uint8_t* resultData, int32_t channelStride, uint8_t* channel0Data, uint8_t* channel1Data, uint8_t* channel2Data, uint8_t* channel3Data);
static void DoPremultiplicationCalculation_SSE2(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride);
static void DoUnpremultiplicationCalculation_SSE2(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride);
static TemporaryRef<DataSourceSurface>
RenderTurbulence_SSE2(const IntSize &aSize, const Point &aOffset, const Size &aBaseFrequency,
int32_t aSeed, int aNumOctaves, TurbulenceType aType, bool aStitch, const Rect &aTileRect);
static TemporaryRef<DataSourceSurface>
ApplyArithmeticCombine_SSE2(DataSourceSurface* aInput1, DataSourceSurface* aInput2, Float aK1, Float aK2, Float aK3, Float aK4);
#endif
};
// Constant-time max and min functions for unsigned arguments
static inline unsigned
umax(unsigned a, unsigned b)
{
return a - ((a - b) & -(a < b));
}
static inline unsigned
umin(unsigned a, unsigned b)
{
return a - ((a - b) & -(a > b));
}
} // namespace gfx
} // namespace mozilla
#endif // _MOZILLA_GFX_FILTERPROCESSING_H_

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,112 @@
/* -*- Mode: C++; tab-width: 20; indent-tabs-mode: nil; c-basic-offset: 2 -*-
* 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/. */
#define SIMD_COMPILE_SSE2
#include "FilterProcessingSIMD-inl.h"
#ifndef USE_SSE2
static_assert(false, "If this file is built, FilterProcessing.h should know about it!");
#endif
namespace mozilla {
namespace gfx {
void
FilterProcessing::ExtractAlpha_SSE2(const IntSize& size, uint8_t* sourceData, int32_t sourceStride, uint8_t* alphaData, int32_t alphaStride)
{
ExtractAlpha_SIMD<__m128i>(size, sourceData, sourceStride, alphaData, alphaStride);
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ConvertToB8G8R8A8_SSE2(SourceSurface* aSurface)
{
return ConvertToB8G8R8A8_SIMD<__m128i>(aSurface);
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyBlending_SSE2(DataSourceSurface* aInput1, DataSourceSurface* aInput2,
BlendMode aBlendMode)
{
return ApplyBlending_SIMD<__m128i,__m128i,__m128i>(aInput1, aInput2, aBlendMode);
}
void
FilterProcessing::ApplyMorphologyHorizontal_SSE2(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOp)
{
ApplyMorphologyHorizontal_SIMD<__m128i,__m128i>(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius, aOp);
}
void
FilterProcessing::ApplyMorphologyVertical_SSE2(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOp)
{
ApplyMorphologyVertical_SIMD<__m128i,__m128i>(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius, aOp);
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyColorMatrix_SSE2(DataSourceSurface* aInput, const Matrix5x4 &aMatrix)
{
return ApplyColorMatrix_SIMD<__m128i,__m128i,__m128i>(aInput, aMatrix);
}
void
FilterProcessing::ApplyComposition_SSE2(DataSourceSurface* aSource, DataSourceSurface* aDest,
CompositeOperator aOperator)
{
return ApplyComposition_SIMD<__m128i,__m128i,__m128i>(aSource, aDest, aOperator);
}
void
FilterProcessing::SeparateColorChannels_SSE2(const IntSize &size, uint8_t* sourceData, int32_t sourceStride, uint8_t* channel0Data, uint8_t* channel1Data, uint8_t* channel2Data, uint8_t* channel3Data, int32_t channelStride)
{
SeparateColorChannels_SIMD<__m128i>(size, sourceData, sourceStride, channel0Data, channel1Data, channel2Data, channel3Data, channelStride);
}
void
FilterProcessing::CombineColorChannels_SSE2(const IntSize &size, int32_t resultStride, uint8_t* resultData, int32_t channelStride, uint8_t* channel0Data, uint8_t* channel1Data, uint8_t* channel2Data, uint8_t* channel3Data)
{
CombineColorChannels_SIMD<__m128i>(size, resultStride, resultData, channelStride, channel0Data, channel1Data, channel2Data, channel3Data);
}
void
FilterProcessing::DoPremultiplicationCalculation_SSE2(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride)
{
DoPremultiplicationCalculation_SIMD<__m128i,__m128i,__m128i>(aSize, aTargetData, aTargetStride, aSourceData, aSourceStride);
}
void
FilterProcessing::DoUnpremultiplicationCalculation_SSE2(
const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride)
{
DoUnpremultiplicationCalculation_SIMD<__m128i,__m128i>(aSize, aTargetData, aTargetStride, aSourceData, aSourceStride);
}
TemporaryRef<DataSourceSurface>
FilterProcessing::RenderTurbulence_SSE2(const IntSize &aSize, const Point &aOffset, const Size &aBaseFrequency,
int32_t aSeed, int aNumOctaves, TurbulenceType aType, bool aStitch, const Rect &aTileRect)
{
return RenderTurbulence_SIMD<__m128,__m128i,__m128i>(aSize, aOffset, aBaseFrequency, aSeed, aNumOctaves, aType, aStitch, aTileRect);
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyArithmeticCombine_SSE2(DataSourceSurface* aInput1, DataSourceSurface* aInput2, Float aK1, Float aK2, Float aK3, Float aK4)
{
return ApplyArithmeticCombine_SIMD<__m128i,__m128i,__m128i>(aInput1, aInput2, aK1, aK2, aK3, aK4);
}
} // namespace mozilla
} // namespace gfx

View File

@ -0,0 +1,321 @@
/* -*- Mode: C++; tab-width: 20; indent-tabs-mode: nil; c-basic-offset: 2 -*-
* 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/. */
#define FILTER_PROCESSING_SCALAR
#include "FilterProcessingSIMD-inl.h"
namespace mozilla {
namespace gfx {
void
FilterProcessing::ExtractAlpha_Scalar(const IntSize& size, uint8_t* sourceData, int32_t sourceStride, uint8_t* alphaData, int32_t alphaStride)
{
for (int32_t y = 0; y < size.height; y++) {
for (int32_t x = 0; x < size.width; x++) {
int32_t sourceIndex = y * sourceStride + 4 * x;
int32_t targetIndex = y * alphaStride + x;
alphaData[targetIndex] = sourceData[sourceIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_A];
}
}
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ConvertToB8G8R8A8_Scalar(SourceSurface* aSurface)
{
return ConvertToB8G8R8A8_SIMD<simd::Scalaru8x16_t>(aSurface);
}
template<BlendMode aBlendMode>
static TemporaryRef<DataSourceSurface>
ApplyBlending_Scalar(DataSourceSurface* aInput1, DataSourceSurface* aInput2)
{
IntSize size = aInput1->GetSize();
RefPtr<DataSourceSurface> target =
Factory::CreateDataSourceSurface(size, FORMAT_B8G8R8A8);
if (!target) {
return nullptr;
}
uint8_t* source1Data = aInput1->GetData();
uint8_t* source2Data = aInput2->GetData();
uint8_t* targetData = target->GetData();
uint32_t targetStride = target->Stride();
uint32_t source1Stride = aInput1->Stride();
uint32_t source2Stride = aInput2->Stride();
for (int32_t y = 0; y < size.height; y++) {
for (int32_t x = 0; x < size.width; x++) {
uint32_t targetIndex = y * targetStride + 4 * x;
uint32_t source1Index = y * source1Stride + 4 * x;
uint32_t source2Index = y * source2Stride + 4 * x;
uint32_t qa = source1Data[source1Index + B8G8R8A8_COMPONENT_BYTEOFFSET_A];
uint32_t qb = source2Data[source2Index + B8G8R8A8_COMPONENT_BYTEOFFSET_A];
for (int32_t i = std::min(B8G8R8A8_COMPONENT_BYTEOFFSET_B, B8G8R8A8_COMPONENT_BYTEOFFSET_R);
i <= std::max(B8G8R8A8_COMPONENT_BYTEOFFSET_B, B8G8R8A8_COMPONENT_BYTEOFFSET_R); i++) {
uint32_t ca = source1Data[source1Index + i];
uint32_t cb = source2Data[source2Index + i];
uint32_t val;
switch (aBlendMode) {
case BLEND_MODE_MULTIPLY:
val = ((255 - qa) * cb + (255 - qb + cb) * ca);
break;
case BLEND_MODE_SCREEN:
val = 255 * (cb + ca) - ca * cb;
break;
case BLEND_MODE_DARKEN:
val = umin((255 - qa) * cb + 255 * ca,
(255 - qb) * ca + 255 * cb);
break;
case BLEND_MODE_LIGHTEN:
val = umax((255 - qa) * cb + 255 * ca,
(255 - qb) * ca + 255 * cb);
break;
default:
MOZ_CRASH();
}
val = umin(FilterProcessing::FastDivideBy255<unsigned>(val), 255U);
targetData[targetIndex + i] = static_cast<uint8_t>(val);
}
uint32_t alpha = 255 * 255 - (255 - qa) * (255 - qb);
targetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_A] =
FilterProcessing::FastDivideBy255<uint8_t>(alpha);
}
}
return target;
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyBlending_Scalar(DataSourceSurface* aInput1, DataSourceSurface* aInput2,
BlendMode aBlendMode)
{
switch (aBlendMode) {
case BLEND_MODE_MULTIPLY:
return gfx::ApplyBlending_Scalar<BLEND_MODE_MULTIPLY>(aInput1, aInput2);
case BLEND_MODE_SCREEN:
return gfx::ApplyBlending_Scalar<BLEND_MODE_SCREEN>(aInput1, aInput2);
case BLEND_MODE_DARKEN:
return gfx::ApplyBlending_Scalar<BLEND_MODE_DARKEN>(aInput1, aInput2);
case BLEND_MODE_LIGHTEN:
return gfx::ApplyBlending_Scalar<BLEND_MODE_LIGHTEN>(aInput1, aInput2);
default:
return nullptr;
}
}
template<MorphologyOperator Operator>
static void
ApplyMorphologyHorizontal_Scalar(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius)
{
static_assert(Operator == MORPHOLOGY_OPERATOR_ERODE ||
Operator == MORPHOLOGY_OPERATOR_DILATE,
"unexpected morphology operator");
for (int32_t y = aDestRect.y; y < aDestRect.YMost(); y++) {
int32_t startX = aDestRect.x - aRadius;
int32_t endX = aDestRect.x + aRadius;
for (int32_t x = aDestRect.x; x < aDestRect.XMost(); x++, startX++, endX++) {
int32_t sourceIndex = y * aSourceStride + 4 * startX;
uint8_t u[4];
for (size_t i = 0; i < 4; i++) {
u[i] = aSourceData[sourceIndex + i];
}
sourceIndex += 4;
for (int32_t ix = startX + 1; ix <= endX; ix++, sourceIndex += 4) {
for (size_t i = 0; i < 4; i++) {
if (Operator == MORPHOLOGY_OPERATOR_ERODE) {
u[i] = umin(u[i], aSourceData[sourceIndex + i]);
} else {
u[i] = umax(u[i], aSourceData[sourceIndex + i]);
}
}
}
int32_t destIndex = y * aDestStride + 4 * x;
for (size_t i = 0; i < 4; i++) {
aDestData[destIndex+i] = u[i];
}
}
}
}
void
FilterProcessing::ApplyMorphologyHorizontal_Scalar(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOp)
{
if (aOp == MORPHOLOGY_OPERATOR_ERODE) {
gfx::ApplyMorphologyHorizontal_Scalar<MORPHOLOGY_OPERATOR_ERODE>(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius);
} else {
gfx::ApplyMorphologyHorizontal_Scalar<MORPHOLOGY_OPERATOR_DILATE>(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius);
}
}
template<MorphologyOperator Operator>
static void ApplyMorphologyVertical_Scalar(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius)
{
static_assert(Operator == MORPHOLOGY_OPERATOR_ERODE ||
Operator == MORPHOLOGY_OPERATOR_DILATE,
"unexpected morphology operator");
int32_t startY = aDestRect.y - aRadius;
int32_t endY = aDestRect.y + aRadius;
for (int32_t y = aDestRect.y; y < aDestRect.YMost(); y++, startY++, endY++) {
for (int32_t x = aDestRect.x; x < aDestRect.XMost(); x++) {
int32_t sourceIndex = startY * aSourceStride + 4 * x;
uint8_t u[4];
for (size_t i = 0; i < 4; i++) {
u[i] = aSourceData[sourceIndex + i];
}
sourceIndex += aSourceStride;
for (int32_t iy = startY + 1; iy <= endY; iy++, sourceIndex += aSourceStride) {
for (size_t i = 0; i < 4; i++) {
if (Operator == MORPHOLOGY_OPERATOR_ERODE) {
u[i] = umin(u[i], aSourceData[sourceIndex + i]);
} else {
u[i] = umax(u[i], aSourceData[sourceIndex + i]);
}
}
}
int32_t destIndex = y * aDestStride + 4 * x;
for (size_t i = 0; i < 4; i++) {
aDestData[destIndex+i] = u[i];
}
}
}
}
void
FilterProcessing::ApplyMorphologyVertical_Scalar(uint8_t* aSourceData, int32_t aSourceStride,
uint8_t* aDestData, int32_t aDestStride,
const IntRect& aDestRect, int32_t aRadius,
MorphologyOperator aOp)
{
if (aOp == MORPHOLOGY_OPERATOR_ERODE) {
gfx::ApplyMorphologyVertical_Scalar<MORPHOLOGY_OPERATOR_ERODE>(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius);
} else {
gfx::ApplyMorphologyVertical_Scalar<MORPHOLOGY_OPERATOR_DILATE>(
aSourceData, aSourceStride, aDestData, aDestStride, aDestRect, aRadius);
}
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyColorMatrix_Scalar(DataSourceSurface* aInput, const Matrix5x4 &aMatrix)
{
return ApplyColorMatrix_SIMD<simd::Scalari32x4_t,simd::Scalari16x8_t,simd::Scalaru8x16_t>(aInput, aMatrix);
}
void
FilterProcessing::ApplyComposition_Scalar(DataSourceSurface* aSource, DataSourceSurface* aDest,
CompositeOperator aOperator)
{
return ApplyComposition_SIMD<simd::Scalari32x4_t,simd::Scalaru16x8_t,simd::Scalaru8x16_t>(aSource, aDest, aOperator);
}
void
FilterProcessing::SeparateColorChannels_Scalar(const IntSize &size, uint8_t* sourceData, int32_t sourceStride, uint8_t* channel0Data, uint8_t* channel1Data, uint8_t* channel2Data, uint8_t* channel3Data, int32_t channelStride)
{
for (int32_t y = 0; y < size.height; y++) {
for (int32_t x = 0; x < size.width; x++) {
int32_t sourceIndex = y * sourceStride + 4 * x;
int32_t targetIndex = y * channelStride + x;
channel0Data[targetIndex] = sourceData[sourceIndex];
channel1Data[targetIndex] = sourceData[sourceIndex+1];
channel2Data[targetIndex] = sourceData[sourceIndex+2];
channel3Data[targetIndex] = sourceData[sourceIndex+3];
}
}
}
void
FilterProcessing::CombineColorChannels_Scalar(const IntSize &size, int32_t resultStride, uint8_t* resultData, int32_t channelStride, uint8_t* channel0Data, uint8_t* channel1Data, uint8_t* channel2Data, uint8_t* channel3Data)
{
for (int32_t y = 0; y < size.height; y++) {
for (int32_t x = 0; x < size.width; x++) {
int32_t resultIndex = y * resultStride + 4 * x;
int32_t channelIndex = y * channelStride + x;
resultData[resultIndex] = channel0Data[channelIndex];
resultData[resultIndex+1] = channel1Data[channelIndex];
resultData[resultIndex+2] = channel2Data[channelIndex];
resultData[resultIndex+3] = channel3Data[channelIndex];
}
}
}
void
FilterProcessing::DoPremultiplicationCalculation_Scalar(const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride)
{
for (int32_t y = 0; y < aSize.height; y++) {
for (int32_t x = 0; x < aSize.width; x++) {
int32_t inputIndex = y * aSourceStride + 4 * x;
int32_t targetIndex = y * aTargetStride + 4 * x;
uint8_t alpha = aSourceData[inputIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_A];
aTargetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_R] =
FastDivideBy255<uint8_t>(aSourceData[inputIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_R] * alpha);
aTargetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_G] =
FastDivideBy255<uint8_t>(aSourceData[inputIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_G] * alpha);
aTargetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_B] =
FastDivideBy255<uint8_t>(aSourceData[inputIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_B] * alpha);
aTargetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_A] = alpha;
}
}
}
void
FilterProcessing::DoUnpremultiplicationCalculation_Scalar(
const IntSize& aSize,
uint8_t* aTargetData, int32_t aTargetStride,
uint8_t* aSourceData, int32_t aSourceStride)
{
for (int32_t y = 0; y < aSize.height; y++) {
for (int32_t x = 0; x < aSize.width; x++) {
int32_t inputIndex = y * aSourceStride + 4 * x;
int32_t targetIndex = y * aTargetStride + 4 * x;
uint8_t alpha = aSourceData[inputIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_A];
uint16_t alphaFactor = sAlphaFactors[alpha];
// inputColor * alphaFactor + 128 is guaranteed to fit into uint16_t
// because the input is premultiplied and thus inputColor <= inputAlpha.
// The maximum value this can attain is 65520 (which is less than 65535)
// for color == alpha == 244:
// 244 * sAlphaFactors[244] + 128 == 244 * 268 + 128 == 65520
aTargetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_R] =
(aSourceData[inputIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_R] * alphaFactor + 128) >> 8;
aTargetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_G] =
(aSourceData[inputIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_G] * alphaFactor + 128) >> 8;
aTargetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_B] =
(aSourceData[inputIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_B] * alphaFactor + 128) >> 8;
aTargetData[targetIndex + B8G8R8A8_COMPONENT_BYTEOFFSET_A] = alpha;
}
}
}
TemporaryRef<DataSourceSurface>
FilterProcessing::RenderTurbulence_Scalar(const IntSize &aSize, const Point &aOffset, const Size &aBaseFrequency,
int32_t aSeed, int aNumOctaves, TurbulenceType aType, bool aStitch, const Rect &aTileRect)
{
return RenderTurbulence_SIMD<simd::Scalarf32x4_t,simd::Scalari32x4_t,simd::Scalaru8x16_t>(
aSize, aOffset, aBaseFrequency, aSeed, aNumOctaves, aType, aStitch, aTileRect);
}
TemporaryRef<DataSourceSurface>
FilterProcessing::ApplyArithmeticCombine_Scalar(DataSourceSurface* aInput1, DataSourceSurface* aInput2, Float aK1, Float aK2, Float aK3, Float aK4)
{
return ApplyArithmeticCombine_SIMD<simd::Scalari32x4_t,simd::Scalari16x8_t,simd::Scalaru8x16_t>(aInput1, aInput2, aK1, aK2, aK3, aK4);
}
} // namespace mozilla
} // namespace gfx

View File

@ -52,11 +52,13 @@ ifneq (,$(INTEL_ARCHITECTURE))
ifdef GNU_CC
ImageScalingSSE2.$(OBJ_SUFFIX): CXXFLAGS+=-msse2
BlurSSE2.$(OBJ_SUFFIX): CXXFLAGS+=-msse2
FilterProcessingSSE2.$(OBJ_SUFFIX): CXXFLAGS+=-msse2
endif
ifdef SOLARIS_SUNPRO_CXX
ImageScalingSSE2.$(OBJ_SUFFIX): OS_CXXFLAGS += -xarch=sse2 -xO4
BlurSSE2.$(OBJ_SUFFIX): OS_CXXFLAGS += -xarch=sse2 -xO4
FilterProcessingSSE2.$(OBJ_SUFFIX): OS_CXXFLAGS += -xarch=sse2 -xO4
endif
endif

View File

@ -79,6 +79,7 @@ if CONFIG['INTEL_ARCHITECTURE']:
if CONFIG['_MSC_VER'] != '1400':
SOURCES += [
'BlurSSE2.cpp',
'FilterProcessingSSE2.cpp',
'ImageScalingSSE2.cpp',
]
@ -90,6 +91,8 @@ UNIFIED_SOURCES += [
'DrawTargetDual.cpp',
'DrawTargetRecording.cpp',
'Factory.cpp',
'FilterProcessing.cpp',
'FilterProcessingScalar.cpp',
'ImageScaling.cpp',
'Matrix.cpp',
'Path.cpp',