Media/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-09-15 13:01:45 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add simple kernel tunning variant

Browse Source 
Related-To: NEO-5327

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
This commit is contained in:
Lukasz Jobczyk
2021-01-11 15:11:14 +00:00
committed by Compute-Runtime-Automation
parent a2eeeff880
commit 40390f7775
4 changed files with 41 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
opencl/source/kernel/kernel.cpp
View File

@ -1164,13 +1164,16 @@ inline void Kernel::makeArgsResident(CommandStreamReceiver &commandStreamReceive
}
void Kernel::performKernelTunning(CommandStreamReceiver &commandStreamReceiver, const Vec3<size_t> &lws, const Vec3<size_t> &gws, const Vec3<size_t> &offsets, TimestampPacketContainer *timestampContainer) {
bool performTunning = false;
auto performTunning = TunningType::DISABLED;
if (DebugManager.flags.EnableKernelTunning.get() != -1) {
performTunning = DebugManager.flags.EnableKernelTunning.get();
performTunning = static_cast<TunningType>(DebugManager.flags.EnableKernelTunning.get());
}
if (performTunning) {
if (performTunning == TunningType::SIMPLE) {
this->singleSubdevicePreferedInCurrentEnqueue = !this->getKernelInfo(commandStreamReceiver.getRootDeviceIndex()).kernelDescriptor.kernelAttributes.flags.useGlobalAtomics;
} else if (performTunning == TunningType::FULL) {
KernelConfig config{gws, lws, offsets};
auto submissionDataIt = this->kernelSubmissionMap.find(config);