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

fix: align total slm size for preferred size 

Related-To: NEO-12639

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
This commit is contained in:
Zbigniew Zdanowicz 2024-10-10 18:22:42 +00:00 committed by Compute-Runtime-Automation
parent f4746a5b07
commit 227ce579fe
4 changed files with 13 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -133,6 +133,7 @@ struct EncodeDispatchKernel {
const uint32_t threadsPerThreadGroup, uint32_t slmTotalSize, SlmPolicy slmPolicy);
static uint32_t getThreadCountPerSubslice(const HardwareInfo &hwInfo);
static uint32_t alignPreferredSlmSize(uint32_t slmSize);
template <typename InterfaceDescriptorType>
static void encodeEuSchedulingPolicy(InterfaceDescriptorType *pInterfaceDescriptor, const KernelDescriptor &kernelDesc, int32_t defaultPipelinedThreadArbitrationPolicy);

5
shared/source/command_container/command_encoder_pre_xe2_hpg_core.inl
View File

@ -16,4 +16,9 @@ uint32_t EncodeDispatchKernel<Family>::getThreadCountPerSubslice(const HardwareI
return hwInfo.gtSystemInfo.ThreadCount / hwInfo.gtSystemInfo.DualSubSliceCount;
}
template <typename Family>
uint32_t EncodeDispatchKernel<Family>::alignPreferredSlmSize(uint32_t slmSize) {
return EncodeDispatchKernel<Family>::alignSlmSize(slmSize);
}
} // namespace NEO

5
shared/source/command_container/command_encoder_xe2_hpg_core_and_later.inl
View File

@ -20,4 +20,9 @@ uint32_t EncodeDispatchKernel<Family>::getThreadCountPerSubslice(const HardwareI
return hwInfo.gtSystemInfo.ThreadCount / hwInfo.gtSystemInfo.SubSliceCount;
}
template <typename Family>
uint32_t EncodeDispatchKernel<Family>::alignPreferredSlmSize(uint32_t slmSize) {
return slmSize;
}
} // namespace NEO

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

@ -1070,6 +1070,8 @@ void EncodeDispatchKernel<Family>::setupPreferredSlmSize(InterfaceDescriptorType
const uint32_t threadsPerDssCount = EncodeDispatchKernel<Family>::getThreadCountPerSubslice(hwInfo);
const uint32_t workGroupCountPerDss = static_cast<uint32_t>(Math::divideAndRoundUp(threadsPerDssCount, threadsPerThreadGroup));
slmTotalSize = EncodeDispatchKernel<Family>::alignPreferredSlmSize(slmTotalSize);
uint32_t slmSize = 0u;
switch (slmPolicy) {