intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-03 14:55:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

feature: improve reporting max cooperative group count

Browse Source 
Related-To: NEO-8210

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
This commit is contained in:
Dunajski, Bartosz
2024-01-22 11:57:53 +00:00
committed by Compute-Runtime-Automation
parent 8d56f8fb6b
commit b77e1a6a71
7 changed files with 97 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
opencl/source/kernel/kernel.cpp
View File

@@ -8,6 +8,7 @@
#include "opencl/source/kernel/kernel.h"
#include "shared/source/built_ins/built_ins.h"
#include "shared/source/command_container/implicit_scaling.h"
#include "shared/source/command_stream/command_stream_receiver.h"
#include "shared/source/debug_settings/debug_settings_manager.h"
#include "shared/source/execution_environment/execution_environment.h"
@@ -1132,8 +1133,18 @@ uint32_t Kernel::getMaxWorkGroupCount(const cl_uint workDim, const size_t *local
auto usedSlmSize = helper.alignSlmSize(slmTotalSize);
uint32_t numSubDevicesForExecution = 1;
bool platformImplicitScaling = helper.platformSupportsImplicitScaling(rootDeviceEnvironment);
auto deviceBitfield = commandQueue->getClDevice().getDeviceBitfield();
if (NEO::ImplicitScalingHelper::isImplicitScalingEnabled(deviceBitfield, platformImplicitScaling)) {
numSubDevicesForExecution = static_cast<uint32_t>(deviceBitfield.count());
}
auto maxWorkGroupCount = KernelHelper::getMaxWorkGroupCount(rootDeviceEnvironment,
kernelInfo.kernelDescriptor,
numSubDevicesForExecution,
usedSlmSize,
workDim,
localWorkSize,