intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-24 21:18:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

feature(internal): logging kernel dispatch params

Browse Source 
Use debug flag PrintKernelDispatchParameters to print params used in
thread group dispatch size heuristic when encoding kernel dispatch.

Related-To: NEO-6989

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
This commit is contained in:
Dominik Dabek
2023-10-17 12:30:47 +02:00
committed by Compute-Runtime-Automation
parent ef6ea0b30c
commit 75c4844987
4 changed files with 40 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
shared/source/command_container/command_encoder_xehp_and_later.inl
View File

@@ -328,6 +328,20 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
auto threadGroupCount = walkerCmd.getThreadGroupIdXDimension() * walkerCmd.getThreadGroupIdYDimension() * walkerCmd.getThreadGroupIdZDimension();
EncodeDispatchKernel<Family>::adjustInterfaceDescriptorData(idd, *args.device, hwInfo, threadGroupCount, kernelDescriptor.kernelAttributes.numGrfRequired, walkerCmd);
if (DebugManager.flags.PrintKernelDispatchParameters.get()) {
fprintf(stdout, "kernel, %s, numGrf, %d, simdSize, %d, tilesCount, %d, implicitScaling, %s, threadGroupCount, %d, numberOfThreadsInGpgpuThreadGroup, %d, threadGroupDimensions, %d, %d, %d, threadGroupDispatchSize enum, %d\n",
kernelDescriptor.kernelMetadata.kernelName.c_str(),
kernelDescriptor.kernelAttributes.numGrfRequired,
kernelDescriptor.kernelAttributes.simdSize,
args.device->getNumSubDevices(),
ImplicitScalingHelper::isImplicitScalingEnabled(args.device->getDeviceBitfield(), true) ? "Yes" : "no",
threadGroupCount,
idd.getNumberOfThreadsInGpgpuThreadGroup(),
walkerCmd.getThreadGroupIdXDimension(),
walkerCmd.getThreadGroupIdYDimension(),
walkerCmd.getThreadGroupIdZDimension(),
idd.getThreadGroupDispatchSize());
}
EncodeDispatchKernel<Family>::appendAdditionalIDDFields(&idd, rootDeviceEnvironment, threadsPerThreadGroup,
args.dispatchInterface->getSlmTotalSize(),