Media
/
compute-runtime
mirror of https://github.com/intel/compute-runtime.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

refactor: improve InOrder Walker patching 

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
This commit is contained in:
Dunajski, Bartosz 2023-12-21 13:16:43 +00:00 committed by Compute-Runtime-Automation
parent fa181937a4
commit 0f275ab893
9 changed files with 26 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
shared/source/command_container/command_encoder_bdw_and_later.inl
View File

@ -16,6 +16,7 @@
#include "shared/source/helpers/api_specific_config.h"
#include "shared/source/helpers/cache_policy.h"
#include "shared/source/helpers/gfx_core_helper.h"
#include "shared/source/helpers/in_order_cmd_helpers.h"
#include "shared/source/helpers/pause_on_gpu_properties.h"
#include "shared/source/helpers/pipe_control_args.h"
#include "shared/source/helpers/pipeline_select_args.h"
@ -640,4 +641,9 @@ void EncodeBatchBufferStartOrEnd<Family>::appendBatchBufferStart(MI_BATCH_BUFFER
cmd.setPredicationEnable(predicate);
}
template <typename Family>
void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue) {
UNRECOVERABLE_IF(true);
}
} // namespace NEO

7
shared/source/command_container/command_encoder_xehp_and_later.inl
View File

@ -918,4 +918,11 @@ size_t EncodeStates<Family>::getSshHeapSize() {
return 2 * MemoryConstants::megaByte;
}
template <typename Family>
void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue) {
auto walkerCmd = reinterpret_cast<typename Family::COMPUTE_WALKER *>(cmd1);
auto &postSync = walkerCmd->getPostSync();
postSync.setImmediateData(baseCounterValue + appendCounterValue);
}
} // namespace NEO

2
shared/source/gen11/command_encoder_gen11.cpp
View File

@ -85,6 +85,8 @@ template void EncodeDispatchKernel<Family>::encode<Family::DefaultWalkerType>(Co
template void EncodeDispatchKernel<Family>::encodeThreadData<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, const uint32_t *startWorkGroup, const uint32_t *numWorkGroups, const uint32_t *workGroupSizes, uint32_t simd, uint32_t localIdDimensions, uint32_t threadsPerThreadGroup, uint32_t threadExecutionMask, bool localIdsGenerationByRuntime, bool inlineDataProgrammingRequired, bool isIndirect, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void EncodeDispatchKernel<Family>::adjustWalkOrder<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue);
template struct EncodeStates<Family>;
template struct EncodeMath<Family>;
template struct EncodeMathMMIO<Family>;

2
shared/source/gen12lp/command_encoder_gen12lp.cpp
View File

@ -125,6 +125,8 @@ template void EncodeDispatchKernel<Family>::encode<Family::DefaultWalkerType>(Co
template void EncodeDispatchKernel<Family>::encodeThreadData<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, const uint32_t *startWorkGroup, const uint32_t *numWorkGroups, const uint32_t *workGroupSizes, uint32_t simd, uint32_t localIdDimensions, uint32_t threadsPerThreadGroup, uint32_t threadExecutionMask, bool localIdsGenerationByRuntime, bool inlineDataProgrammingRequired, bool isIndirect, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void EncodeDispatchKernel<Family>::adjustWalkOrder<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue);
template struct EncodeStates<Family>;
template struct EncodeMath<Family>;
template struct EncodeMathMMIO<Family>;

2
shared/source/gen8/command_encoder_gen8.cpp
View File

@ -69,6 +69,8 @@ template void EncodeDispatchKernel<Family>::encode<Family::DefaultWalkerType>(Co
template void EncodeDispatchKernel<Family>::encodeThreadData<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, const uint32_t *startWorkGroup, const uint32_t *numWorkGroups, const uint32_t *workGroupSizes, uint32_t simd, uint32_t localIdDimensions, uint32_t threadsPerThreadGroup, uint32_t threadExecutionMask, bool localIdsGenerationByRuntime, bool inlineDataProgrammingRequired, bool isIndirect, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void EncodeDispatchKernel<Family>::adjustWalkOrder<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue);
template struct EncodeStates<Family>;
template struct EncodeMath<Family>;
template struct EncodeMathMMIO<Family>;

2
shared/source/gen9/command_encoder_gen9.cpp
View File

@ -70,6 +70,8 @@ template void EncodeDispatchKernel<Family>::encode<Family::DefaultWalkerType>(Co
template void EncodeDispatchKernel<Family>::encodeThreadData<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, const uint32_t *startWorkGroup, const uint32_t *numWorkGroups, const uint32_t *workGroupSizes, uint32_t simd, uint32_t localIdDimensions, uint32_t threadsPerThreadGroup, uint32_t threadExecutionMask, bool localIdsGenerationByRuntime, bool inlineDataProgrammingRequired, bool isIndirect, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void EncodeDispatchKernel<Family>::adjustWalkOrder<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue);
template struct EncodeStates<Family>;
template struct EncodeMath<Family>;
template struct EncodeMathMMIO<Family>;

10
shared/source/helpers/in_order_cmd_helpers.h
View File

@ -137,15 +137,7 @@ struct PatchCmd {
semaphoreCmd->setSemaphoreDataDword(static_cast<uint32_t>(baseCounterValue + appendCounterValue));
}
void patchComputeWalker(uint64_t appendCounterValue) {
if constexpr (GfxFamily::walkerPostSyncSupport) {
auto walkerCmd = reinterpret_cast<typename GfxFamily::COMPUTE_WALKER *>(cmd1);
auto &postSync = walkerCmd->getPostSync();
postSync.setImmediateData(baseCounterValue + appendCounterValue);
} else {
UNRECOVERABLE_IF(true);
}
}
void patchComputeWalker(uint64_t appendCounterValue);
void patchLri64b(uint64_t appendCounterValue) {
if (isExternalDependency()) {

2
shared/source/xe_hpc_core/command_encoder_xe_hpc_core.cpp
View File

@ -379,6 +379,8 @@ template void EncodeDispatchKernel<Family>::encode<Family::DefaultWalkerType>(Co
template void EncodeDispatchKernel<Family>::encodeThreadData<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, const uint32_t *startWorkGroup, const uint32_t *numWorkGroups, const uint32_t *workGroupSizes, uint32_t simd, uint32_t localIdDimensions, uint32_t threadsPerThreadGroup, uint32_t threadExecutionMask, bool localIdsGenerationByRuntime, bool inlineDataProgrammingRequired, bool isIndirect, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void EncodeDispatchKernel<Family>::adjustWalkOrder<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue);
template struct EncodeStates<Family>;
template struct EncodeMath<Family>;
template struct EncodeMathMMIO<Family>;

2
shared/source/xe_hpg_core/command_encoder_xe_hpg_core.cpp
View File

@ -232,6 +232,8 @@ template void EncodeDispatchKernel<Family>::encode<Family::DefaultWalkerType>(Co
template void EncodeDispatchKernel<Family>::encodeThreadData<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, const uint32_t *startWorkGroup, const uint32_t *numWorkGroups, const uint32_t *workGroupSizes, uint32_t simd, uint32_t localIdDimensions, uint32_t threadsPerThreadGroup, uint32_t threadExecutionMask, bool localIdsGenerationByRuntime, bool inlineDataProgrammingRequired, bool isIndirect, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void EncodeDispatchKernel<Family>::adjustWalkOrder<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue);
template struct EncodeStates<Family>;
template struct EncodeMath<Family>;
template struct EncodeMathMMIO<Family>;