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 new EngineUsage

Browse Source 
Simplify verifying EngineUsage in tryGetEngine function.
Remove unused getGpgpuEnginesCount function.

Related-To: NEO-4940

Signed-off-by: Filip Hazubski <filip.hazubski@intel.com>
This commit is contained in:
Filip Hazubski
2021-08-16 18:24:13 +00:00
committed by Compute-Runtime-Automation
parent fccd22e3c7
commit 63f8c9d98b
29 changed files with 89 additions and 97 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
opencl/source/api/api.cpp
View File

@ -5932,13 +5932,6 @@ cl_int CL_API_CALL clEnqueueNDCountKernelINTEL(cl_command_queue commandQueue,
auto &device = pCommandQueue->getClDevice();
auto rootDeviceIndex = device.getRootDeviceIndex();
auto &hardwareInfo = device.getHardwareInfo();
auto &hwHelper = HwHelper::get(hardwareInfo.platform.eRenderCoreFamily);
auto engineGroupType = hwHelper.getEngineGroupType(pCommandQueue->getGpgpuEngine().getEngineType(), hardwareInfo);
if (!hwHelper.isCooperativeDispatchSupported(engineGroupType, hardwareInfo.platform.eProductFamily)) {
retVal = CL_INVALID_COMMAND_QUEUE;
return retVal;
}
pKernel = pMultiDeviceKernel->getKernel(rootDeviceIndex);
size_t globalWorkSize[3];
@ -5946,6 +5939,22 @@ cl_int CL_API_CALL clEnqueueNDCountKernelINTEL(cl_command_queue commandQueue,
globalWorkSize[i] = workgroupCount[i] * localWorkSize[i];
}
if (pKernel->usesSyncBuffer()) {
if (pKernel->getExecutionType() != KernelExecutionType::Concurrent) {
retVal = CL_INVALID_KERNEL;
return retVal;
}
auto &hardwareInfo = device.getHardwareInfo();
auto &hwHelper = HwHelper::get(hardwareInfo.platform.eRenderCoreFamily);
auto engineGroupType = hwHelper.getEngineGroupType(pCommandQueue->getGpgpuEngine().getEngineType(),
pCommandQueue->getGpgpuEngine().getEngineUsage(), hardwareInfo);
if (!hwHelper.isCooperativeDispatchSupported(engineGroupType)) {
retVal = CL_INVALID_COMMAND_QUEUE;
return retVal;
}
}
if (pKernel->getExecutionType() == KernelExecutionType::Concurrent) {
size_t requestedNumberOfWorkgroups = 1;
for (size_t i = 0; i < workDim; i++) {
@ -5958,20 +5967,15 @@ cl_int CL_API_CALL clEnqueueNDCountKernelINTEL(cl_command_queue commandQueue,
}
}
if (pKernel->usesSyncBuffer()) {
if (pKernel->getExecutionType() != KernelExecutionType::Concurrent) {
retVal = CL_INVALID_KERNEL;
return retVal;
}
device.getDevice().allocateSyncBufferHandler();
}
if (!pCommandQueue->validateCapabilityForOperation(CL_QUEUE_CAPABILITY_KERNEL_INTEL, numEventsInWaitList, eventWaitList, event)) {
retVal = CL_INVALID_OPERATION;
return retVal;
}
if (pKernel->usesSyncBuffer()) {
device.getDevice().allocateSyncBufferHandler();
}
TakeOwnershipWrapper<MultiDeviceKernel> kernelOwnership(*pMultiDeviceKernel, gtpinIsGTPinInitialized());
if (gtpinIsGTPinInitialized()) {
gtpinNotifyKernelSubmit(kernel, pCommandQueue);