intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-07 21:27:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: to always use grfs count in calculateNumThreadsPerThreadGroup

Browse Source 
grf size != grf count

Related-To: GSD-8437
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
This commit is contained in:
Katarzyna Cencelewska
2024-03-22 09:39:15 +00:00
committed by Compute-Runtime-Automation
parent df54d67f40
commit da7b03dd15
27 changed files with 86 additions and 68 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
opencl/source/helpers/hardware_commands_helper_base.inl
View File

@@ -55,6 +55,7 @@ size_t HardwareCommandsHelper<GfxFamily>::getSizeRequiredIOH(const Kernel &kerne
const auto &hwInfo = kernel.getHardwareInfo();
auto numChannels = kernelDescriptor.kernelAttributes.numLocalIdChannels;
auto grfCount = kernelDescriptor.kernelAttributes.numGrfRequired;
uint32_t grfSize = hwInfo.capabilityTable.grfSize;
auto simdSize = kernelDescriptor.kernelAttributes.simdSize;
uint32_t requiredWalkOrder = 0u;
@@ -69,7 +70,7 @@ size_t HardwareCommandsHelper<GfxFamily>::getSizeRequiredIOH(const Kernel &kerne
requiredWalkOrder,
simdSize);
auto size = kernel.getCrossThreadDataSize() +
getPerThreadDataSizeTotal(simdSize, grfSize, numChannels, localWorkSize, isHwLocalIdGeneration, rootDeviceEnvironment);
getPerThreadDataSizeTotal(simdSize, grfSize, grfCount, numChannels, localWorkSize, isHwLocalIdGeneration, rootDeviceEnvironment);
auto pImplicitArgs = kernel.getImplicitArgs();
if (pImplicitArgs) {
@@ -268,9 +269,9 @@ size_t HardwareCommandsHelper<GfxFamily>::sendIndirectState(
}
auto &gfxCoreHelper = device.getGfxCoreHelper();
auto grfSize = kernel.getDescriptor().kernelAttributes.numGrfRequired;
auto grfCount = kernel.getDescriptor().kernelAttributes.numGrfRequired;
auto localWorkItems = localWorkSize[0] * localWorkSize[1] * localWorkSize[2];
auto threadsPerThreadGroup = gfxCoreHelper.calculateNumThreadsPerThreadGroup(simd, static_cast<uint32_t>(localWorkItems), grfSize, !localIdsGenerationByRuntime, device.getRootDeviceEnvironment());
auto threadsPerThreadGroup = gfxCoreHelper.calculateNumThreadsPerThreadGroup(simd, static_cast<uint32_t>(localWorkItems), grfCount, !localIdsGenerationByRuntime, device.getRootDeviceEnvironment());
uint32_t sizeCrossThreadData = kernel.getCrossThreadDataSize();