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

refactor: cleanup getBindlessMode() usage

Browse Source 
- getGlobalBindlessHeapConfiguration() should be used to choose global
alloctor for SSH
- remove not needed and incorrect unit tests
- remove not needed branches
- bindless mode controls bindless compilation only

Related-To: NEO-7063

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
This commit is contained in:
Mateusz Hoppe
2023-06-05 16:53:29 +00:00
committed by Compute-Runtime-Automation
parent dadd19e17b
commit 0844ca0ac8
25 changed files with 187 additions and 532 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
shared/source/command_container/command_encoder_bdw_and_later.inl
View File

@@ -124,14 +124,10 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
if (kernelDescriptor.payloadMappings.samplerTable.numSamplers > 0) {
auto dsHeap = args.dynamicStateHeap;
if (dsHeap == nullptr) {
if (!ApiSpecificConfig::getBindlessMode()) {
dsHeap = container.getIndirectHeap(HeapType::DYNAMIC_STATE);
auto dshSizeRequired = NEO::EncodeDispatchKernel<Family>::getSizeRequiredDsh(kernelDescriptor, container.getNumIddPerBlock());
if (dsHeap->getAvailableSpace() <= dshSizeRequired) {
dsHeap = container.getHeapWithRequiredSizeAndAlignment(HeapType::DYNAMIC_STATE, dsHeap->getMaxAvailableSpace(), NEO::EncodeDispatchKernel<Family>::getDefaultDshAlignment());
}
} else {
dsHeap = args.device->getBindlessHeapsHelper()->getHeap(BindlessHeapsHelper::GLOBAL_DSH);
dsHeap = container.getIndirectHeap(HeapType::DYNAMIC_STATE);
auto dshSizeRequired = NEO::EncodeDispatchKernel<Family>::getSizeRequiredDsh(kernelDescriptor, container.getNumIddPerBlock());
if (dsHeap->getAvailableSpace() <= dshSizeRequired) {
dsHeap = container.getHeapWithRequiredSizeAndAlignment(HeapType::DYNAMIC_STATE, dsHeap->getMaxAvailableSpace(), NEO::EncodeDispatchKernel<Family>::getDefaultDshAlignment());
}
}
UNRECOVERABLE_IF(!dsHeap);
@@ -259,10 +255,6 @@ void EncodeDispatchKernel<Family>::encode(CommandContainer &container, EncodeDis
cmd.setPredicateEnable(args.isPredicate);
if (ApiSpecificConfig::getBindlessMode()) {
container.getResidencyContainer().push_back(args.device->getBindlessHeapsHelper()->getHeap(NEO::BindlessHeapsHelper::BindlesHeapType::GLOBAL_DSH)->getGraphicsAllocation());
}
auto threadGroupCount = cmd.getThreadGroupIdXDimension() * cmd.getThreadGroupIdYDimension() * cmd.getThreadGroupIdZDimension();
EncodeDispatchKernel<Family>::adjustInterfaceDescriptorData(idd, *args.device, hwInfo, threadGroupCount, kernelDescriptor.kernelAttributes.numGrfRequired, cmd);
@@ -304,11 +296,7 @@ void EncodeMediaInterfaceDescriptorLoad<Family>::encode(CommandContainer &contai
if (childDsh != nullptr) {
heapBase = childDsh->getCpuBase();
} else {
if (ApiSpecificConfig::getBindlessMode()) {
heapBase = container.getDevice()->getBindlessHeapsHelper()->getHeap(BindlessHeapsHelper::GLOBAL_DSH)->getGraphicsAllocation()->getUnderlyingBuffer();
} else {
heapBase = container.getIndirectHeap(HeapType::DYNAMIC_STATE)->getCpuBase();
}
heapBase = container.getIndirectHeap(HeapType::DYNAMIC_STATE)->getCpuBase();
}
auto mediaStateFlush = container.getCommandStream()->getSpaceForCmd<MEDIA_STATE_FLUSH>();
@@ -316,10 +304,6 @@ void EncodeMediaInterfaceDescriptorLoad<Family>::encode(CommandContainer &contai
auto iddOffset = static_cast<uint32_t>(ptrDiff(container.getIddBlock(), heapBase));
iddOffset += ApiSpecificConfig::getBindlessMode() ? static_cast<uint32_t>(container.getDevice()->getBindlessHeapsHelper()->getHeap(BindlessHeapsHelper::GLOBAL_DSH)->getGraphicsAllocation()->getGpuAddress() -
container.getDevice()->getBindlessHeapsHelper()->getHeap(BindlessHeapsHelper::GLOBAL_DSH)->getGraphicsAllocation()->getGpuBaseAddress())
: 0;
MEDIA_INTERFACE_DESCRIPTOR_LOAD cmd = Family::cmdInitMediaInterfaceDescriptorLoad;
cmd.setInterfaceDescriptorDataStartAddress(iddOffset);
cmd.setInterfaceDescriptorTotalLength(sizeof(INTERFACE_DESCRIPTOR_DATA) * container.getNumIddPerBlock());