intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-10 07:08:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

refactor: unify getMaxWorkGroupCount logic

Browse Source 
Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
This commit is contained in:
Dunajski, Bartosz
2024-01-19 16:00:58 +00:00
committed by Compute-Runtime-Automation
parent 1002cb9f34
commit 6f4ed10919
6 changed files with 93 additions and 69 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
opencl/source/kernel/kernel.cpp
View File

@@ -1128,29 +1128,18 @@ uint32_t Kernel::getMaxWorkGroupCount(const cl_uint workDim, const size_t *local
auto engineGroupType = helper.getEngineGroupType(commandQueue->getGpgpuEngine().getEngineType(),
commandQueue->getGpgpuEngine().getEngineUsage(), hardwareInfo);
const auto &kernelDescriptor = kernelInfo.kernelDescriptor;
auto dssCount = hardwareInfo.gtSystemInfo.DualSubSliceCount;
if (dssCount == 0) {
dssCount = hardwareInfo.gtSystemInfo.SubSliceCount;
}
auto availableThreadCount = helper.calculateAvailableThreadCount(hardwareInfo, kernelDescriptor.kernelAttributes.numGrfRequired);
auto availableSlmSize = static_cast<uint32_t>(dssCount * MemoryConstants::kiloByte * hardwareInfo.capabilityTable.slmSize);
auto usedSlmSize = helper.alignSlmSize(slmTotalSize);
auto maxBarrierCount = static_cast<uint32_t>(helper.getMaxBarrierRegisterPerSlice());
auto barrierCount = kernelDescriptor.kernelAttributes.barrierCount;
auto maxWorkGroupCount = KernelHelper::getMaxWorkGroupCount(kernelInfo.getMaxSimdSize(),
availableThreadCount,
dssCount,
availableSlmSize,
usedSlmSize,
maxBarrierCount,
barrierCount,
workDim,
localWorkSize);
auto isEngineInstanced = commandQueue->getGpgpuCommandStreamReceiver().getOsContext().isEngineInstanced();
maxWorkGroupCount = helper.adjustMaxWorkGroupCount(maxWorkGroupCount, engineGroupType, rootDeviceEnvironment, isEngineInstanced);
auto usedSlmSize = helper.alignSlmSize(slmTotalSize);
auto maxWorkGroupCount = KernelHelper::getMaxWorkGroupCount(rootDeviceEnvironment,
kernelInfo.kernelDescriptor,
usedSlmSize,
workDim,
localWorkSize,
engineGroupType,
isEngineInstanced);
return maxWorkGroupCount;
}