inline void FELighting::platformApply(LightingData& data, LightSource::PaintingData& paintingData)
{
// The selection here eventually should happen dynamically on some platforms.
#if CPU(ARM_NEON) && CPU(ARM_TRADITIONAL) && COMPILER(GCC)
platformApplyNeon(data, paintingData);
#else
platformApplyGeneric(data, paintingData);
#endif
}
inline void FEGaussianBlur::platformApply(Uint8ClampedArray* srcPixelArray, Uint8ClampedArray* tmpPixelArray, unsigned kernelSizeX, unsigned kernelSizeY, IntSize& paintSize)
{
int scanline = 4 * paintSize.width();
int extraHeight = 3 * kernelSizeY * 0.5f;
int optimalThreadNumber = (paintSize.width() * paintSize.height()) / (s_minimalRectDimension + extraHeight * paintSize.width());
if (optimalThreadNumber > 1) {
WTF::ParallelJobs<PlatformApplyParameters> parallelJobs(&platformApplyWorker, optimalThreadNumber);
int jobs = parallelJobs.numberOfJobs();
if (jobs > 1) {
int blockHeight = paintSize.height() / jobs;
--jobs;
for (int job = jobs; job >= 0; --job) {
PlatformApplyParameters& params = parallelJobs.parameter(job);
params.filter = this;
int startY;
int endY;
if (!job) {
startY = 0;
endY = blockHeight + extraHeight;
params.srcPixelArray = srcPixelArray;
params.dstPixelArray = tmpPixelArray;
} else {
if (job == jobs) {
startY = job * blockHeight - extraHeight;
endY = paintSize.height();
} else {
startY = job * blockHeight - extraHeight;
endY = (job + 1) * blockHeight + extraHeight;
}
int blockSize = (endY - startY) * scanline;
params.srcPixelArray = Uint8ClampedArray::createUninitialized(blockSize);
params.dstPixelArray = Uint8ClampedArray::createUninitialized(blockSize);
memcpy(params.srcPixelArray->data(), srcPixelArray->data() + startY * scanline, blockSize);
}
params.width = paintSize.width();
params.height = endY - startY;
params.kernelSizeX = kernelSizeX;
params.kernelSizeY = kernelSizeY;
}
parallelJobs.execute();
// Copy together the parts of the image.
for (int job = jobs; job >= 1; --job) {
PlatformApplyParameters& params = parallelJobs.parameter(job);
int sourceOffset;
int destinationOffset;
int size;
if (job == jobs) {
sourceOffset = extraHeight * scanline;
destinationOffset = job * blockHeight * scanline;
size = (paintSize.height() - job * blockHeight) * scanline;
} else {
sourceOffset = extraHeight * scanline;
destinationOffset = job * blockHeight * scanline;
size = blockHeight * scanline;
}
memcpy(srcPixelArray->data() + destinationOffset, params.srcPixelArray->data() + sourceOffset, size);
}
return;
}
// Fallback to single threaded mode.
}
// The selection here eventually should happen dynamically on some platforms.
#if CPU(ARM_NEON) && COMPILER(GCC)
platformApplyNeon(srcPixelArray, tmpPixelArray, kernelSizeX, kernelSizeY, paintSize);
#else
platformApplyGeneric(srcPixelArray, tmpPixelArray, kernelSizeX, kernelSizeY, paintSize);
#endif
}
