intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-05 09:09:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

Cleanup MediaInterfaceDescriptorLoad logic in command encoder

Browse Source 
Add a patch to command encoder for samplers when DSH is dirty.

Signed-off-by: Tratnack, Geoffrey <geoffrey.tratnack@intel.com>
Related-To: LOCI-3365
This commit is contained in:
Tratnack, Geoffrey
2022-10-13 00:38:44 +00:00
committed by Compute-Runtime-Automation
parent 5cfb59c27e
commit 349af0bd5e
8 changed files with 137 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
shared/source/command_container/command_encoder_bdw_and_later.inl
View File

@@ -104,6 +104,15 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
PreemptionHelper::programInterfaceDescriptorDataPreemption<Family>(&idd, args.preemptionMode);
if (!ApiSpecificConfig::getBindlessConfiguration()) {
auto heap = container.getIndirectHeap(HeapType::DYNAMIC_STATE);
auto dshSizeRequired = NEO::EncodeDispatchKernel<Family>::getSizeRequiredDsh(kernelDescriptor, container.getNumIddPerBlock());
if (heap->getAvailableSpace() <= dshSizeRequired) {
heap = container.getHeapWithRequiredSizeAndAlignment(HeapType::DYNAMIC_STATE, heap->getUsed() + heap->getAvailableSpace(), 0);
UNRECOVERABLE_IF(!heap);
}
}
uint32_t samplerStateOffset = 0;
uint32_t samplerCount = 0;
@@ -166,6 +175,9 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
args.dispatchInterface->getPerThreadData(), sizePerThreadDataForWholeGroup);
}
uint32_t numIDD = 0u;
void *iddPtr = getInterfaceDescriptor(container, numIDD);
auto slmSizeNew = args.dispatchInterface->getSlmTotalSize();
bool dirtyHeaps = container.isAnyHeapDirty();
bool flush = container.slmSize != slmSizeNew || dirtyHeaps || args.requiresUncachedMocs;
@@ -199,15 +211,12 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
if (container.slmSize != slmSizeNew) {
EncodeL3State<Family>::encode(container, slmSizeNew != 0u);
container.slmSize = slmSizeNew;
if (container.nextIddInBlock != container.getNumIddPerBlock()) {
EncodeMediaInterfaceDescriptorLoad<Family>::encode(container);
}
}
}
uint32_t numIDD = 0u;
void *iddPtr = getInterfaceDescriptor(container, numIDD, hwInfo);
if (numIDD == 0 || flush) {
EncodeMediaInterfaceDescriptorLoad<Family>::encode(container);
}
cmd.setIndirectDataStartAddress(static_cast<uint32_t>(offsetThreadData));
cmd.setIndirectDataLength(sizeThreadData);
@@ -545,8 +554,8 @@ template <typename Family>
void EncodeDispatchKernel<Family>::adjustWalkOrder(WALKER_TYPE &walkerCmd, uint32_t requiredWorkGroupOrder, const HardwareInfo &hwInfo) {}
template <typename Family>
uint32_t EncodeDispatchKernel<Family>::additionalSizeRequiredDsh() {
return sizeof(typename Family::INTERFACE_DESCRIPTOR_DATA);
uint32_t EncodeDispatchKernel<Family>::additionalSizeRequiredDsh(uint32_t numIddsPerBlock) {
return sizeof(typename Family::INTERFACE_DESCRIPTOR_DATA) * numIddsPerBlock;
}
} // namespace NEO