intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-30 01:35:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

Store ClDevice in DispatchInfo

Browse Source 
Related-To: NEO-5001
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
This commit is contained in:
Mateusz Jablonski
2020-11-16 12:08:30 +01:00
committed by Compute-Runtime-Automation
parent b848f20abc
commit 97119f5f3e
56 changed files with 335 additions and 317 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
opencl/source/command_queue/enqueue_common.h
View File

@@ -75,7 +75,7 @@ void CommandQueueHw<GfxFamily>::enqueueHandler(Surface *(&surfaces)[surfaceCount
}
if (kernel->getKernelInfo().builtinDispatchBuilder == nullptr) {
DispatchInfoBuilder<SplitDispatch::Dim::d3D, SplitDispatch::SplitMode::WalkerSplit> builder;
DispatchInfoBuilder<SplitDispatch::Dim::d3D, SplitDispatch::SplitMode::WalkerSplit> builder(getClDevice());
builder.setDispatchGeometry(workDim, workItems, enqueuedWorkSizes, globalOffsets, Vec3<size_t>{0, 0, 0}, localWorkSizesIn);
builder.setKernel(kernel);
builder.bake(multiDispatchInfo);
@@ -105,11 +105,12 @@ void CommandQueueHw<GfxFamily>::enqueueHandler(Surface *(&surfaces)[surfaceCount
template <typename GfxFamily>
void CommandQueueHw<GfxFamily>::forceDispatchScheduler(NEO::MultiDispatchInfo &multiDispatchInfo) {
SchedulerKernel &scheduler = getContext().getSchedulerKernel();
DispatchInfo dispatchInfo(&scheduler, 1, Vec3<size_t>(scheduler.getGws(), 1, 1), Vec3<size_t>(scheduler.getLws(), 1, 1), Vec3<size_t>(0, 0, 0));
auto devQueue = this->getContext().getDefaultDeviceQueue();
DeviceQueueHw<GfxFamily> *devQueueHw = castToObjectOrAbort<DeviceQueueHw<GfxFamily>>(devQueue);
DispatchInfo dispatchInfo(devQueue->getClDevice(), &scheduler, 1, Vec3<size_t>(scheduler.getGws(), 1, 1), Vec3<size_t>(scheduler.getLws(), 1, 1), Vec3<size_t>(0, 0, 0));
scheduler.createReflectionSurface();
GraphicsAllocation *reflectionSurface = scheduler.getKernelReflectionSurface();

5
opencl/source/command_queue/local_work_size.cpp
View File

@@ -414,9 +414,10 @@ Vec3<size_t> computeWorkgroupSize(const DispatchInfo &dispatchInfo) {
auto kernel = dispatchInfo.getKernel();
if (kernel != nullptr) {
const auto &hwInfo = kernel->getDevice().getHardwareInfo();
auto &device = dispatchInfo.getClDevice();
const auto &hwInfo = device.getHardwareInfo();
auto &hwHelper = HwHelper::get(hwInfo.platform.eRenderCoreFamily);
auto isSimulation = kernel->getDevice().isSimulation();
auto isSimulation = device.isSimulation();
if (kernel->requiresLimitedWorkgroupSize() && hwHelper.isSpecialWorkgroupSizeRequired(hwInfo, isSimulation)) {
setSpecialWorkgroupSize(workGroupSize);
} else if (DebugManager.flags.EnableComputeWorkSizeND.get()) {