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

fix: scratch controller residency

Browse Source 
Related-To: HSD-18039519400

Signed-off-by: Bartosz Dunajski <bartosz.dunajski@intel.com>
This commit is contained in:
Bartosz Dunajski
2024-08-09 13:13:51 +00:00
committed by Compute-Runtime-Automation
parent bc385fa1b9
commit d76ac1d1de
21 changed files with 47 additions and 73 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
shared/source/command_container/command_encoder.h
View File

@@ -204,7 +204,7 @@ struct EncodeDispatchKernel {
static size_t getDefaultIOHAlignment();
template <bool isHeapless>
static void setScratchAddress(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &csr);
static void setScratchAddress(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &submissionCsr);
template <bool isHeapless>
static uint64_t getScratchAddressForImmediatePatching(CommandContainer &container, EncodeDispatchKernelArgs &args);
template <bool isHeapless>

2
shared/source/command_container/command_encoder.inl
View File

@@ -913,7 +913,7 @@ size_t EncodeDispatchKernel<GfxFamily>::getScratchPtrOffsetOfImplicitArgs() {
template <typename Family>
template <bool isHeapless>
void EncodeDispatchKernel<Family>::setScratchAddress(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &csr) {
void EncodeDispatchKernel<Family>::setScratchAddress(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &submissionCsr) {
}
template <typename Family>

4
shared/source/command_container/command_encoder_enablers.inl
View File

@@ -22,8 +22,8 @@ template void NEO::EncodeDispatchKernel<Family>::encode<Family::DefaultWalkerTyp
template void NEO::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 NEO::EncodeDispatchKernel<Family>::adjustWalkOrder<Family::DefaultWalkerType>(Family::DefaultWalkerType &walkerCmd, uint32_t requiredWorkGroupOrder, const RootDeviceEnvironment &rootDeviceEnvironment);
template void NEO::EncodeDispatchKernel<Family>::programBarrierEnable<Family::INTERFACE_DESCRIPTOR_DATA>(Family::INTERFACE_DESCRIPTOR_DATA &interfaceDescriptor, uint32_t value, const HardwareInfo &hwInfo);
template void NEO::EncodeDispatchKernel<Family>::setScratchAddress<false>(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &csr);
template void NEO::EncodeDispatchKernel<Family>::setScratchAddress<true>(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &csr);
template void NEO::EncodeDispatchKernel<Family>::setScratchAddress<false>(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &submissionCsr);
template void NEO::EncodeDispatchKernel<Family>::setScratchAddress<true>(uint64_t &scratchAddress, uint32_t requiredScratchSlot0Size, uint32_t requiredScratchSlot1Size, IndirectHeap *ssh, CommandStreamReceiver &submissionCsr);
template void NEO::EncodeDispatchKernel<Family>::programInlineDataHeapless<false>(uint8_t *inlineDataPtr, EncodeDispatchKernelArgs &args, CommandContainer &container, uint64_t offsetThreadData, uint64_t scratchPtr);
template void NEO::EncodeDispatchKernel<Family>::encodeEuSchedulingPolicy<Family::INTERFACE_DESCRIPTOR_DATA>(Family::INTERFACE_DESCRIPTOR_DATA *pInterfaceDescriptor, const KernelDescriptor &kernelDesc, int32_t defaultPipelinedThreadArbitrationPolicy);
template uint64_t NEO::EncodeDispatchKernel<Family>::getScratchAddressForImmediatePatching<false>(CommandContainer &container, EncodeDispatchKernelArgs &args);