intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-29 17:13:29 +08:00
Code Issues Packages Projects Releases Wiki Activity

refactor: add max front end threads in EncodeWalkerArgs

Browse Source 
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
This commit is contained in:
Kamil Kopryk
2023-12-07 09:35:23 +00:00
committed by Compute-Runtime-Automation
parent f1841527a9
commit 7a14db7f9b
5 changed files with 7 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
opencl/source/command_queue/hardware_interface_bdw_and_later.inl
View File

@@ -102,7 +102,7 @@ inline void HardwareInterface<GfxFamily>::programWalker(
0,
commandQueue.getDevice());
EncodeWalkerArgs encodeWalkerArgs{kernel.getExecutionType(), false, kernel.getKernelInfo().kernelDescriptor, NEO::RequiredDispatchWalkOrder::None, 0};
EncodeWalkerArgs encodeWalkerArgs{kernel.getExecutionType(), false, kernel.getKernelInfo().kernelDescriptor, NEO::RequiredDispatchWalkOrder::None, 0, 0};
EncodeDispatchKernel<GfxFamily>::encodeAdditionalWalkerFields(rootDeviceEnvironment, walkerCmd, encodeWalkerArgs);
*walkerCmdBuf = walkerCmd;
}

3
opencl/source/command_queue/hardware_interface_xehp_and_later.inl
View File

@@ -142,7 +142,8 @@ inline void HardwareInterface<GfxFamily>::programWalker(
kernelSystemAllocation = kernel.isAnyKernelArgumentUsingSystemMemory();
}
bool requiredSystemFence = kernelSystemAllocation && walkerArgs.event != nullptr;
EncodeWalkerArgs encodeWalkerArgs{kernel.getExecutionType(), requiredSystemFence, kernelInfo.kernelDescriptor, NEO::RequiredDispatchWalkOrder::None, 0};
auto maxFrontEndThreads = commandQueue.getDevice().getDeviceInfo().maxFrontEndThreads;
EncodeWalkerArgs encodeWalkerArgs{kernel.getExecutionType(), requiredSystemFence, kernelInfo.kernelDescriptor, NEO::RequiredDispatchWalkOrder::None, 0, maxFrontEndThreads};
EncodeDispatchKernel<GfxFamily>::template encodeAdditionalWalkerFields<WalkerType>(rootDeviceEnvironment, walkerCmd, encodeWalkerArgs);
auto devices = queueCsr.getOsContext().getDeviceBitfield();

2
opencl/test/unit_test/command_queue/dispatch_walker_tests_dg2_and_later.cpp
View File

@@ -84,7 +84,7 @@ HWTEST2_F(WalkerDispatchTestDg2AndLater, givenDebugVariableSetWhenProgramCompute
auto &rootDeviceEnvironment = *mockExecutionEnvironment.rootDeviceEnvironments[0];
KernelDescriptor kernelDescriptor;
EncodeWalkerArgs walkerArgs{KernelExecutionType::defaultType, true, kernelDescriptor, NEO::RequiredDispatchWalkOrder::None, 0};
EncodeWalkerArgs walkerArgs{KernelExecutionType::defaultType, true, kernelDescriptor, NEO::RequiredDispatchWalkOrder::None, 0, 0};
for (auto forceL3PrefetchForComputeWalker : {false, true}) {
debugManager.flags.ForceL3PrefetchForComputeWalker.set(forceL3PrefetchForComputeWalker);
EncodeDispatchKernel<FamilyType>::encodeAdditionalWalkerFields(rootDeviceEnvironment, walkerCmd, walkerArgs);

1
shared/source/command_container/command_encoder.h
View File

@@ -93,6 +93,7 @@ struct EncodeWalkerArgs {
const KernelDescriptor &kernelDescriptor;
NEO::RequiredDispatchWalkOrder requiredDispatchWalkOrder = NEO::RequiredDispatchWalkOrder::None;
uint32_t additionalSizeParam = NEO::additionalKernelLaunchSizeParamNotSet;
uint32_t maxFrontEndThreads = 0;
};
template <typename GfxFamily>

3
shared/source/command_container/command_encoder_xehp_and_later.inl
View File

@@ -395,7 +395,8 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
args.isHostScopeSignalEvent && args.isKernelUsingSystemAllocation,
kernelDescriptor,
args.requiredDispatchWalkOrder,
args.additionalSizeParam};
args.additionalSizeParam,
args.device->getDeviceInfo().maxFrontEndThreads};
EncodeDispatchKernel<Family>::encodeAdditionalWalkerFields(rootDeviceEnvironment, walkerCmd, walkerArgs);
PreemptionHelper::applyPreemptionWaCmdsBegin<Family>(listCmdBufferStream, *args.device);