intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-25 13:33:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

Flush state caches after command list is destroyed

Browse Source 
When state base address tracking is enabled and command list use private heaps
then command list at destroy time must calls all compute CSRs that were using
that heap to invalidate state caches.
This allows new command list to reuse the same heap allocation for different
surface states, so before new use cached states are invalidated.

Related-To: NEO-5055

Signed-off-by: Zbigniew Zdanowicz <zbigniew.zdanowicz@intel.com>
This commit is contained in:
Zbigniew Zdanowicz
2023-03-27 21:37:18 +00:00
committed by Compute-Runtime-Automation
parent 820a189c52
commit 6437c1a91e
16 changed files with 185 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
shared/source/command_stream/command_stream_receiver.h
View File

@@ -93,6 +93,7 @@ class CommandStreamReceiver {
const IndirectHeap *dsh, const IndirectHeap *ioh, const IndirectHeap *ssh,
TaskCountType taskLevel, DispatchFlags &dispatchFlags, Device &device) = 0;
virtual CompletionStamp flushBcsTask(LinearStream &commandStream, size_t commandStreamStart, const DispatchBcsFlags &dispatchBcsFlags, const HardwareInfo &hwInfo) = 0;
virtual SubmissionStatus sendRenderStateCacheFlush() = 0;
virtual bool flushBatchedSubmissions() = 0;
MOCKABLE_VIRTUAL SubmissionStatus submitBatchBuffer(BatchBuffer &batchBuffer, ResidencyContainer &allocationsForResidency);

3
shared/source/command_stream/command_stream_receiver_hw.h
View File

@@ -100,10 +100,11 @@ class CommandStreamReceiverHw : public CommandStreamReceiver {
SubmissionStatus flushTagUpdate() override;
SubmissionStatus flushMiFlushDW();
SubmissionStatus flushPipeControl();
SubmissionStatus flushPipeControl(bool stateCacheFlush);
SubmissionStatus flushSmallTask(LinearStream &commandStreamTask,
size_t commandStreamStartTask);
SubmissionStatus flushHandler(BatchBuffer &batchBuffer, ResidencyContainer &allocationsForResidency);
SubmissionStatus sendRenderStateCacheFlush() override;
bool isUpdateTagFromWaitEnabled() override;
void updateTagFromWait() override;

15
shared/source/command_stream/command_stream_receiver_hw_base.inl
View File

@@ -1365,7 +1365,7 @@ inline SubmissionStatus CommandStreamReceiverHw<GfxFamily>::flushTagUpdate() {
if (EngineHelpers::isBcs(this->osContext->getEngineType())) {
return this->flushMiFlushDW();
} else {
return this->flushPipeControl();
return this->flushPipeControl(false);
}
}
return SubmissionStatus::DEVICE_UNINITIALIZED;
@@ -1393,7 +1393,7 @@ inline SubmissionStatus CommandStreamReceiverHw<GfxFamily>::flushMiFlushDW() {
}
template <typename GfxFamily>
SubmissionStatus CommandStreamReceiverHw<GfxFamily>::flushPipeControl() {
SubmissionStatus CommandStreamReceiverHw<GfxFamily>::flushPipeControl(bool stateCacheFlush) {
auto lock = obtainUniqueOwnership();
PipeControlArgs args;
@@ -1401,6 +1401,12 @@ SubmissionStatus CommandStreamReceiverHw<GfxFamily>::flushPipeControl() {
args.notifyEnable = isUsedNotifyEnableForPostSync();
args.workloadPartitionOffset = isMultiTileOperationEnabled();
if (stateCacheFlush) {
args.textureCacheInvalidationEnable = true;
args.renderTargetCacheFlushEnable = true;
args.stateCacheInvalidationEnable = true;
}
auto dispatchSize = MemorySynchronizationCommands<GfxFamily>::getSizeForBarrierWithPostSyncOperation(peekRootDeviceEnvironment(), args.tlbInvalidation) + this->getCmdSizeForPrologue();
auto &commandStream = getCS(dispatchSize);
@@ -1454,6 +1460,11 @@ SubmissionStatus CommandStreamReceiverHw<GfxFamily>::flushSmallTask(LinearStream
return submissionStatus;
}
template <typename GfxFamily>
SubmissionStatus CommandStreamReceiverHw<GfxFamily>::sendRenderStateCacheFlush() {
return this->flushPipeControl(true);
}
template <typename GfxFamily>
inline SubmissionStatus CommandStreamReceiverHw<GfxFamily>::flushHandler(BatchBuffer &batchBuffer, ResidencyContainer &allocationsForResidency) {
auto status = flush(batchBuffer, allocationsForResidency);