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

Do not switch to batched mode when update task count form wait is enabled.

Browse Source 
There is not need to enable batching in this case as we do not have
synchronization points between enqueues.

Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>
This commit is contained in:
Michal Mrozek
2022-05-19 12:31:11 +00:00
committed by Compute-Runtime-Automation
parent 9cf1fbd627
commit cbd73d2d9c
2 changed files with 26 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
opencl/source/command_queue/command_queue.cpp
View File

@@ -155,14 +155,6 @@ void CommandQueue::initializeGpgpuInternals() const {
auto &hwHelper = NEO::HwHelper::get(hwInfo.platform.eRenderCoreFamily);
const auto &hwInfoConfig = *HwInfoConfig::get(hwInfo.platform.eProductFamily);
if (getCmdQueueProperties<cl_queue_properties>(propertiesVector.data(), CL_QUEUE_PROPERTIES) & static_cast<cl_queue_properties>(CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE)) {
this->gpgpuEngine->commandStreamReceiver->overrideDispatchPolicy(DispatchMode::BatchedDispatch);
if (DebugManager.flags.CsrDispatchMode.get() != 0) {
this->gpgpuEngine->commandStreamReceiver->overrideDispatchPolicy(static_cast<DispatchMode>(DebugManager.flags.CsrDispatchMode.get()));
}
this->gpgpuEngine->commandStreamReceiver->enableNTo1SubmissionModel();
}
if (device->getDevice().getDebugger() && !this->gpgpuEngine->commandStreamReceiver->getDebugSurfaceAllocation()) {
auto maxDbgSurfaceSize = hwHelper.getSipKernelMaxDbgSurfaceSize(hwInfo);
auto debugSurface = this->gpgpuEngine->commandStreamReceiver->allocateDebugSurface(maxDbgSurfaceSize);
@@ -178,6 +170,14 @@ void CommandQueue::initializeGpgpuInternals() const {
gpgpuEngine->osContext->ensureContextInitialized();
gpgpuEngine->commandStreamReceiver->initDirectSubmission();
if (getCmdQueueProperties<cl_queue_properties>(propertiesVector.data(), CL_QUEUE_PROPERTIES) & static_cast<cl_queue_properties>(CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE) && !this->gpgpuEngine->commandStreamReceiver->isUpdateTagFromWaitEnabled()) {
this->gpgpuEngine->commandStreamReceiver->overrideDispatchPolicy(DispatchMode::BatchedDispatch);
if (DebugManager.flags.CsrDispatchMode.get() != 0) {
this->gpgpuEngine->commandStreamReceiver->overrideDispatchPolicy(static_cast<DispatchMode>(DebugManager.flags.CsrDispatchMode.get()));
}
this->gpgpuEngine->commandStreamReceiver->enableNTo1SubmissionModel();
}
}
CommandStreamReceiver &CommandQueue::getGpgpuCommandStreamReceiver() const {