intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-09 06:23:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

Queue stall mode for RelaxedOrdering

Browse Source 
Related-To: NEO-7458

Signed-off-by: Dunajski, Bartosz <bartosz.dunajski@intel.com>
This commit is contained in:
Dunajski, Bartosz
2022-11-19 18:25:04 +00:00
committed by Compute-Runtime-Automation
parent bb308c04ed
commit bc619fcbec
28 changed files with 549 additions and 200 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
shared/test/common/helpers/batch_buffer_helper.h
View File

@@ -11,11 +11,11 @@
using namespace NEO;
struct BatchBufferHelper {
static BatchBuffer createDefaultBatchBuffer(GraphicsAllocation *commandBufferAllocation, LinearStream *stream, size_t usedSize) {
static BatchBuffer createDefaultBatchBuffer(GraphicsAllocation *commandBufferAllocation, LinearStream *stream, size_t usedSize, size_t chainedBatchBufferStartOffset) {
return BatchBuffer(
commandBufferAllocation, // commandBufferAllocation
0, // startOffset
0, // chainedBatchBufferStartOffset
chainedBatchBufferStartOffset, // chainedBatchBufferStartOffset
0, // taskStartAddress
nullptr, // chainedBatchBuffer
false, // requiresCoherency
@@ -25,7 +25,12 @@ struct BatchBufferHelper {
usedSize, // usedSize
stream, // stream
nullptr, // endCmdPtr
false // useSingleSubdevice
false, // useSingleSubdevice
false // hasStallingCmds
);
}
static BatchBuffer createDefaultBatchBuffer(GraphicsAllocation *commandBufferAllocation, LinearStream *stream, size_t usedSize) {
return createDefaultBatchBuffer(commandBufferAllocation, stream, usedSize, 0);
}
};

61
shared/test/common/helpers/dispatch_flags_helper.h
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2019-2021 Intel Corporation
* Copyright (C) 2019-2022 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -14,35 +14,36 @@ using namespace NEO;
struct DispatchFlagsHelper {
static DispatchFlags createDefaultDispatchFlags() {
return DispatchFlags(
{}, //csrDependencies
nullptr, //barrierTimestampPacketNodes
{}, //pipelineSelectArgs
nullptr, //flushStampReference
QueueThrottle::MEDIUM, //throttle
PreemptionMode::Disabled, //preemptionMode
GrfConfig::DefaultGrfNumber, //numGrfRequired
L3CachingSettings::l3CacheOn, //l3CacheSettings
ThreadArbitrationPolicy::NotPresent, //threadArbitrationPolicy
AdditionalKernelExecInfo::NotApplicable, //additionalKernelExecInfo
KernelExecutionType::NotApplicable, //kernelExecutionType
MemoryCompressionState::NotApplicable, //memoryCompressionState
QueueSliceCount::defaultSliceCount, //sliceCount
false, //blocking
false, //dcFlush
false, //useSLM
false, //guardCommandBufferWithPipeControl
false, //gsba32BitRequired
false, //requiresCoherency
false, //lowPriority
false, //implicitFlush
false, //outOfOrderExecutionAllowed
false, //epilogueRequired
false, //usePerDssBackedBuffer
false, //useSingleSubdevice
false, //useGlobalAtomics
false, //areMultipleSubDevicesInContext
false, //memoryMigrationRequired
false //textureCacheFlush
{}, // csrDependencies
nullptr, // barrierTimestampPacketNodes
{}, // pipelineSelectArgs
nullptr, // flushStampReference
QueueThrottle::MEDIUM, // throttle
PreemptionMode::Disabled, // preemptionMode
GrfConfig::DefaultGrfNumber, // numGrfRequired
L3CachingSettings::l3CacheOn, // l3CacheSettings
ThreadArbitrationPolicy::NotPresent, // threadArbitrationPolicy
AdditionalKernelExecInfo::NotApplicable, // additionalKernelExecInfo
KernelExecutionType::NotApplicable, // kernelExecutionType
MemoryCompressionState::NotApplicable, // memoryCompressionState
QueueSliceCount::defaultSliceCount, // sliceCount
false, // blocking
false, // dcFlush
false, // useSLM
false, // guardCommandBufferWithPipeControl
false, // gsba32BitRequired
false, // requiresCoherency
false, // lowPriority
false, // implicitFlush
false, // outOfOrderExecutionAllowed
false, // epilogueRequired
false, // usePerDssBackedBuffer
false, // useSingleSubdevice
false, // useGlobalAtomics
false, // areMultipleSubDevicesInContext
false, // memoryMigrationRequired
false, // textureCacheFlush
false // hasStallingCmds
);
}
};

18
shared/test/unit_test/command_stream/aub_command_stream_receiver_2_tests.cpp
View File

@@ -59,8 +59,8 @@ HWTEST_F(AubCommandStreamReceiverTests, givenAubCommandStreamReceiverWhenForcedB
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, chainedBatchBuffer, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed(), 128);
batchBuffer.chainedBatchBuffer = chainedBatchBuffer;
size_t sizeBatchBuffer = 0xffffu;
std::unique_ptr<GraphicsAllocation, std::function<void(GraphicsAllocation *)>> flatBatchBuffer(
@@ -84,7 +84,7 @@ HWTEST_F(AubCommandStreamReceiverTests, givenAubCommandStreamReceiverWhenForcedB
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed(), 128);
size_t sizeBatchBuffer = 0xffffu;
@@ -111,7 +111,7 @@ HWTEST_F(AubCommandStreamReceiverTests, givenAubCommandStreamReceiverWhenForcedB
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, chainedBatchBuffer, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed(), 128);
size_t sizeBatchBuffer = 0xffffu;
@@ -133,7 +133,7 @@ HWTEST_F(FlatBatchBufferHelperAubTests, givenAubCommandStreamReceiverWhenRegiste
auto aubCsr = aubExecutionEnvironment->template getCsr<AUBCommandStreamReceiverHw<FamilyType>>();
LinearStream cs(aubExecutionEnvironment->commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed(), 128);
aubCsr->getFlatBatchBufferHelper().registerCommandChunk(batchBuffer, sizeof(MI_BATCH_BUFFER_START));
ASSERT_EQ(1u, aubCsr->getFlatBatchBufferHelper().getCommandChunkList().size());
@@ -227,7 +227,7 @@ HWTEST_F(FlatBatchBufferHelperAubTests, givenAubCommandStreamReceiverWhenForcedB
ASSERT_EQ(3u, aubCsr->getFlatBatchBufferHelper().getPatchInfoCollection().size());
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed(), 128);
size_t sizeBatchBuffer = 0u;
@@ -350,7 +350,7 @@ HWTEST_F(AubCommandStreamReceiverTests, givenAubCommandStreamReceiverWhenForcedF
auto chainedBatchBuffer = aubExecutionEnvironment->executionEnvironment->memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{pDevice->getRootDeviceIndex(), MemoryConstants::pageSize});
ASSERT_NE(nullptr, chainedBatchBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, chainedBatchBuffer, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed(), 128);
aubCsr->makeResident(*chainedBatchBuffer);
@@ -385,7 +385,7 @@ HWTEST_F(AubCommandStreamReceiverTests, givenAubCommandStreamReceiverWhenForcedF
aubCsr->overwriteFlatBatchBufferHelper(mockHelper);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed(), 128);
aubCsr->flush(batchBuffer, allocationsForResidency);
@@ -408,7 +408,7 @@ HWTEST_F(AubCommandStreamReceiverTests, givenAubCommandStreamReceiverWhenForcedF
aubCsr->overwriteFlatBatchBufferHelper(mockHelper);
ResidencyContainer allocationsForResidency;
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed(), 128);
aubCsr->flush(batchBuffer, allocationsForResidency);

20
shared/test/unit_test/command_stream/aub_file_stream_tests.cpp
View File

@@ -592,7 +592,7 @@ HWTEST_F(AubFileStreamTests, givenAubCommandStreamReceiverWhenFlushIsCalledThenI
aubCsr.initializeTagAllocation();
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 1, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 1, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false, false};
ResidencyContainer allocationsForResidency;
aubCsr.flush(batchBuffer, allocationsForResidency);
@@ -761,9 +761,6 @@ HWTEST_F(AddPatchInfoCommentsAubTests, givenAddPatchInfoCommentsCalledWhenNoPatc
auto aubExecutionEnvironment = getEnvironment<AUBCommandStreamReceiverHw<FamilyType>>(false, true, true);
auto aubCsr = aubExecutionEnvironment->template getCsr<AUBCommandStreamReceiverHw<FamilyType>>();
LinearStream cs(aubExecutionEnvironment->commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
aubCsr->stream = mockAubFileStream.get();
bool result = aubCsr->addPatchInfoComments();
@@ -781,9 +778,6 @@ HWTEST_F(AddPatchInfoCommentsAubTests, givenAddPatchInfoCommentsCalledWhenFirstA
auto aubExecutionEnvironment = getEnvironment<AUBCommandStreamReceiverHw<FamilyType>>(false, true, true);
auto aubCsr = aubExecutionEnvironment->template getCsr<AUBCommandStreamReceiverHw<FamilyType>>();
LinearStream cs(aubExecutionEnvironment->commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
aubCsr->stream = mockAubFileStream.get();
mockAubFileStream->addCommentResult = false;
@@ -797,9 +791,6 @@ HWTEST_F(AddPatchInfoCommentsAubTests, givenAddPatchInfoCommentsCalledWhenSecond
auto aubExecutionEnvironment = getEnvironment<AUBCommandStreamReceiverHw<FamilyType>>(false, true, true);
auto aubCsr = aubExecutionEnvironment->template getCsr<AUBCommandStreamReceiverHw<FamilyType>>();
LinearStream cs(aubExecutionEnvironment->commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
aubCsr->stream = mockAubFileStream.get();
mockAubFileStream->addCommentResults = {true, false};
@@ -814,9 +805,6 @@ HWTEST_F(AddPatchInfoCommentsAubTests, givenAddPatchInfoCommentsCalledWhenPatchI
auto aubExecutionEnvironment = getEnvironment<AUBCommandStreamReceiverHw<FamilyType>>(false, true, true);
auto aubCsr = aubExecutionEnvironment->template getCsr<AUBCommandStreamReceiverHw<FamilyType>>();
LinearStream cs(aubExecutionEnvironment->commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
aubCsr->stream = mockAubFileStream.get();
PatchInfoData patchInfoData[2] = {{0xAAAAAAAA, 128u, PatchInfoAllocationType::Default, 0xBBBBBBBB, 256u, PatchInfoAllocationType::Default},
@@ -879,9 +867,6 @@ HWTEST_F(AddPatchInfoCommentsAubTests, givenAddPatchInfoCommentsCalledWhenSource
auto aubExecutionEnvironment = getEnvironment<AUBCommandStreamReceiverHw<FamilyType>>(false, true, true);
auto aubCsr = aubExecutionEnvironment->template getCsr<AUBCommandStreamReceiverHw<FamilyType>>();
LinearStream cs(aubExecutionEnvironment->commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
aubCsr->stream = mockAubFileStream.get();
PatchInfoData patchInfoData = {0x0, 0u, PatchInfoAllocationType::Default, 0xBBBBBBBB, 0u, PatchInfoAllocationType::Default};
@@ -926,9 +911,6 @@ HWTEST_F(AddPatchInfoCommentsAubTests, givenAddPatchInfoCommentsCalledWhenTarget
auto aubExecutionEnvironment = getEnvironment<AUBCommandStreamReceiverHw<FamilyType>>(false, true, true);
auto aubCsr = aubExecutionEnvironment->template getCsr<AUBCommandStreamReceiverHw<FamilyType>>();
LinearStream cs(aubExecutionEnvironment->commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 128u, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
aubCsr->stream = mockAubFileStream.get();
PatchInfoData patchInfoData = {0xAAAAAAAA, 0u, PatchInfoAllocationType::Default, 0x0, 0u, PatchInfoAllocationType::Default};

10
shared/test/unit_test/command_stream/command_stream_receiver_tests.cpp
View File

@@ -2425,4 +2425,14 @@ HWTEST_F(CommandStreamReceiverHwTest, givenOutOfMemoryFailureOnFlushWhenInitiali
commandStreamReceiver.flushReturnValue = SubmissionStatus::OUT_OF_HOST_MEMORY;
EXPECT_EQ(SubmissionStatus::OUT_OF_HOST_MEMORY, commandStreamReceiver.initializeDeviceWithFirstSubmission());
}
HWTEST_F(CommandStreamReceiverHwTest, whenFlushTagUpdateThenSetStallingCmdsFlag) {
auto &ultCsr = pDevice->getUltCommandStreamReceiver<FamilyType>();
ultCsr.recordFlusheBatchBuffer = true;
EXPECT_EQ(SubmissionStatus::SUCCESS, ultCsr.flushTagUpdate());
EXPECT_TRUE(ultCsr.latestFlushedBatchBuffer.hasStallingCmds);
}

2
shared/test/unit_test/command_stream/compute_mode_tests.h
View File

@@ -95,6 +95,6 @@ struct ComputeModeRequirements : public ::testing::Test {
CommandStreamReceiver *csr = nullptr;
std::unique_ptr<MockDevice> device;
DispatchFlags flags{{}, nullptr, {}, nullptr, QueueThrottle::MEDIUM, PreemptionMode::Disabled, GrfConfig::DefaultGrfNumber, L3CachingSettings::l3CacheOn, ThreadArbitrationPolicy::NotPresent, AdditionalKernelExecInfo::NotApplicable, KernelExecutionType::NotApplicable, MemoryCompressionState::NotApplicable, QueueSliceCount::defaultSliceCount, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false};
DispatchFlags flags{{}, nullptr, {}, nullptr, QueueThrottle::MEDIUM, PreemptionMode::Disabled, GrfConfig::DefaultGrfNumber, L3CachingSettings::l3CacheOn, ThreadArbitrationPolicy::NotPresent, AdditionalKernelExecInfo::NotApplicable, KernelExecutionType::NotApplicable, MemoryCompressionState::NotApplicable, QueueSliceCount::defaultSliceCount, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false, false};
GraphicsAllocation *alloc = nullptr;
};

4
shared/test/unit_test/command_stream/tbx_command_stream_tests.cpp
View File

@@ -467,7 +467,7 @@ HWTEST_F(TbxCommandStreamTests, givenTbxCommandStreamReceiverWhenFlushIsCalledTh
auto commandBuffer = pDevice->executionEnvironment->memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{pDevice->getRootDeviceIndex(), MemoryConstants::pageSize, pDevice->getDeviceBitfield()});
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 1, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 1, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false, false};
MockGraphicsAllocation allocation(reinterpret_cast<void *>(0x1000), 0x1000);
ResidencyContainer allocationsForResidency = {&allocation};
@@ -493,7 +493,7 @@ HWTEST_F(TbxCommandStreamTests, givenTbxCommandStreamReceiverInBatchedModeWhenFl
auto commandBuffer = pDevice->executionEnvironment->memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{pDevice->getRootDeviceIndex(), MemoryConstants::pageSize, pDevice->getDeviceBitfield()});
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 1, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 1, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false, false};
ResidencyContainer allocationsForResidency;
tbxCsr.flush(batchBuffer, allocationsForResidency);

206
shared/test/unit_test/direct_submission/direct_submission_tests_2.cpp
View File

@@ -913,7 +913,7 @@ struct DirectSubmissionRelaxedOrderingTests : public DirectSubmissionDispatchBuf
}
template <typename FamilyType>
bool verifySchedulerProgramming(LinearStream &cs, uint64_t deferredTaskListVa, uint64_t semaphoreGpuVa, uint32_t semaphoreValue, size_t offset);
bool verifySchedulerProgramming(LinearStream &cs, uint64_t deferredTaskListVa, uint64_t semaphoreGpuVa, uint32_t semaphoreValue, size_t offset, size_t &endOffset);
template <typename FamilyType>
bool verifyMiPredicate(void *miPredicateCmd, MiPredicateType predicateType);
@@ -1187,7 +1187,7 @@ bool DirectSubmissionRelaxedOrderingTests::verifyConditionalDataRegBbStart(void
}
template <typename FamilyType>
bool DirectSubmissionRelaxedOrderingTests::verifySchedulerProgramming(LinearStream &cs, uint64_t deferredTaskListVa, uint64_t semaphoreGpuVa, uint32_t semaphoreValue, size_t offset) {
bool DirectSubmissionRelaxedOrderingTests::verifySchedulerProgramming(LinearStream &cs, uint64_t deferredTaskListVa, uint64_t semaphoreGpuVa, uint32_t semaphoreValue, size_t offset, size_t &endOffset) {
using MI_BATCH_BUFFER_START = typename FamilyType::MI_BATCH_BUFFER_START;
using MI_SEMAPHORE_WAIT = typename FamilyType::MI_SEMAPHORE_WAIT;
using MI_SET_PREDICATE = typename FamilyType::MI_SET_PREDICATE;
@@ -1501,6 +1501,8 @@ bool DirectSubmissionRelaxedOrderingTests::verifySchedulerProgramming(LinearStre
if (!verifyLri<FamilyType>(lriCmd, CS_GPR_R5, 0)) {
continue;
}
lriCmd++;
endOffset = ptrDiff(lriCmd, cs.getCpuBase());
success = true;
break;
@@ -1704,9 +1706,6 @@ HWTEST_F(DirectSubmissionRelaxedOrderingTests, givenNotEnoughSpaceForTaskStoreSe
}
HWTEST2_F(DirectSubmissionRelaxedOrderingTests, whenDispatchingWorkThenDispatchScheduler, IsAtLeastXeHpcCore) {
using MI_LOAD_REGISTER_IMM = typename FamilyType::MI_LOAD_REGISTER_IMM;
using MI_MATH_ALU_INST_INLINE = typename FamilyType::MI_MATH_ALU_INST_INLINE;
using MI_MATH = typename FamilyType::MI_MATH;
using Dispatcher = RenderDispatcher<FamilyType>;
MockDirectSubmissionHw<FamilyType, Dispatcher> directSubmission(*pDevice->getDefaultEngine().commandStreamReceiver);
@@ -1717,18 +1716,209 @@ HWTEST2_F(DirectSubmissionRelaxedOrderingTests, whenDispatchingWorkThenDispatchS
uint64_t deferredTasksListVa = directSubmission.deferredTasksListAllocation->getGpuAddress();
uint64_t semaphoreGpuVa = directSubmission.semaphoreGpuVa;
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, 0));
size_t endOffset = 0;
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, 0, endOffset));
FlushStampTracker flushStamp(true);
directSubmission.dispatchCommandBuffer(batchBuffer, flushStamp);
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, offset));
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, offset, endOffset));
offset = directSubmission.ringCommandStream.getUsed();
directSubmission.dispatchCommandBuffer(batchBuffer, flushStamp);
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, offset));
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, offset, endOffset));
}
HWTEST2_F(DirectSubmissionRelaxedOrderingTests, givenBbWithStallingCmdsWhenDispatchingThenProgramSchedulerWithR5, IsAtLeastXeHpcCore) {
using MI_LOAD_REGISTER_IMM = typename FamilyType::MI_LOAD_REGISTER_IMM;
using Dispatcher = RenderDispatcher<FamilyType>;
MockDirectSubmissionHw<FamilyType, Dispatcher> directSubmission(*pDevice->getDefaultEngine().commandStreamReceiver);
directSubmission.initialize(true, false);
size_t offset = directSubmission.ringCommandStream.getUsed();
uint64_t deferredTasksListVa = directSubmission.deferredTasksListAllocation->getGpuAddress();
uint64_t semaphoreGpuVa = directSubmission.semaphoreGpuVa;
size_t endOffset = 0;
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, 0, endOffset));
FlushStampTracker flushStamp(true);
batchBuffer.hasStallingCmds = false;
directSubmission.dispatchCommandBuffer(batchBuffer, flushStamp);
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, offset, endOffset));
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, endOffset, endOffset));
offset = directSubmission.ringCommandStream.getUsed();
batchBuffer.hasStallingCmds = true;
directSubmission.dispatchCommandBuffer(batchBuffer, flushStamp);
HardwareParse hwParse;
hwParse.parseCommands<FamilyType>(directSubmission.ringCommandStream, offset);
hwParse.findHardwareCommands<FamilyType>();
bool success = false;
MI_LOAD_REGISTER_IMM *lriCmd = nullptr;
for (auto &it : hwParse.cmdList) {
lriCmd = genCmdCast<MI_LOAD_REGISTER_IMM *>(it);
if (lriCmd) {
if (CS_GPR_R5 == lriCmd->getRegisterOffset() && lriCmd->getDataDword() == 1) {
success = true;
break;
}
}
}
ASSERT_TRUE(success);
offset = ptrDiff(++lriCmd, directSubmission.ringCommandStream.getCpuBase());
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount - 1, offset, endOffset));
EXPECT_TRUE(endOffset > offset);
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, endOffset, endOffset));
}
HWTEST2_F(DirectSubmissionRelaxedOrderingTests, givenFirstBbWithStallingCmdsWhenDispatchingThenDontProgramSchedulerWithR5, IsAtLeastXeHpcCore) {
using MI_LOAD_REGISTER_IMM = typename FamilyType::MI_LOAD_REGISTER_IMM;
using Dispatcher = RenderDispatcher<FamilyType>;
MockDirectSubmissionHw<FamilyType, Dispatcher> directSubmission(*pDevice->getDefaultEngine().commandStreamReceiver);
directSubmission.initialize(true, false);
size_t offset = directSubmission.ringCommandStream.getUsed();
uint64_t deferredTasksListVa = directSubmission.deferredTasksListAllocation->getGpuAddress();
uint64_t semaphoreGpuVa = directSubmission.semaphoreGpuVa;
size_t endOffset = 0;
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, 0, endOffset));
FlushStampTracker flushStamp(true);
batchBuffer.hasStallingCmds = true;
directSubmission.dispatchCommandBuffer(batchBuffer, flushStamp);
HardwareParse hwParse;
hwParse.parseCommands<FamilyType>(directSubmission.ringCommandStream, offset);
hwParse.findHardwareCommands<FamilyType>();
bool success = false;
MI_LOAD_REGISTER_IMM *lriCmd = nullptr;
for (auto &it : hwParse.cmdList) {
lriCmd = genCmdCast<MI_LOAD_REGISTER_IMM *>(it);
if (lriCmd) {
if (CS_GPR_R5 == lriCmd->getRegisterOffset() && lriCmd->getDataDword() == 1) {
success = true;
break;
}
}
}
EXPECT_FALSE(success);
}
HWTEST2_F(DirectSubmissionRelaxedOrderingTests, whenStoppingRingThenProgramSchedulerWithR5, IsAtLeastXeHpcCore) {
using MI_LOAD_REGISTER_IMM = typename FamilyType::MI_LOAD_REGISTER_IMM;
using Dispatcher = RenderDispatcher<FamilyType>;
MockDirectSubmissionHw<FamilyType, Dispatcher> directSubmission(*pDevice->getDefaultEngine().commandStreamReceiver);
directSubmission.initialize(true, false);
size_t offset = directSubmission.ringCommandStream.getUsed();
uint64_t deferredTasksListVa = directSubmission.deferredTasksListAllocation->getGpuAddress();
uint64_t semaphoreGpuVa = directSubmission.semaphoreGpuVa;
size_t endOffset = 0;
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, 0, endOffset));
FlushStampTracker flushStamp(true);
batchBuffer.hasStallingCmds = false;
directSubmission.dispatchCommandBuffer(batchBuffer, flushStamp);
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, offset, endOffset));
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, endOffset, endOffset));
offset = directSubmission.ringCommandStream.getUsed();
directSubmission.stopRingBuffer();
HardwareParse hwParse;
hwParse.parseCommands<FamilyType>(directSubmission.ringCommandStream, offset);
hwParse.findHardwareCommands<FamilyType>();
bool success = false;
MI_LOAD_REGISTER_IMM *lriCmd = nullptr;
for (auto &it : hwParse.cmdList) {
lriCmd = genCmdCast<MI_LOAD_REGISTER_IMM *>(it);
if (lriCmd) {
if (CS_GPR_R5 == lriCmd->getRegisterOffset() && lriCmd->getDataDword() == 1) {
success = true;
break;
}
}
}
ASSERT_TRUE(success);
offset = ptrDiff(lriCmd, directSubmission.ringCommandStream.getCpuBase());
EXPECT_TRUE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, offset, endOffset));
EXPECT_TRUE(endOffset > offset);
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, endOffset, endOffset));
}
HWTEST2_F(DirectSubmissionRelaxedOrderingTests, WhenStoppingRingWithoutSubmissionThenDontProgramSchedulerWithR5, IsAtLeastXeHpcCore) {
using MI_LOAD_REGISTER_IMM = typename FamilyType::MI_LOAD_REGISTER_IMM;
using Dispatcher = RenderDispatcher<FamilyType>;
MockDirectSubmissionHw<FamilyType, Dispatcher> directSubmission(*pDevice->getDefaultEngine().commandStreamReceiver);
directSubmission.initialize(true, false);
size_t offset = directSubmission.ringCommandStream.getUsed();
uint64_t deferredTasksListVa = directSubmission.deferredTasksListAllocation->getGpuAddress();
uint64_t semaphoreGpuVa = directSubmission.semaphoreGpuVa;
size_t endOffset = 0;
EXPECT_FALSE(verifySchedulerProgramming<FamilyType>(directSubmission.ringCommandStream, deferredTasksListVa, semaphoreGpuVa, directSubmission.currentQueueWorkCount, 0, endOffset));
directSubmission.stopRingBuffer();
HardwareParse hwParse;
hwParse.parseCommands<FamilyType>(directSubmission.ringCommandStream, offset);
hwParse.findHardwareCommands<FamilyType>();
bool success = false;
MI_LOAD_REGISTER_IMM *lriCmd = nullptr;
for (auto &it : hwParse.cmdList) {
lriCmd = genCmdCast<MI_LOAD_REGISTER_IMM *>(it);
if (lriCmd) {
if (CS_GPR_R5 == lriCmd->getRegisterOffset() && lriCmd->getDataDword() == 1) {
success = true;
break;
}
}
}
EXPECT_FALSE(success);
}
HWTEST2_F(DirectSubmissionRelaxedOrderingTests, whenProgrammingEndingCmdsThenSetReturnRegisters, IsAtLeastXeHpcCore) {

46
shared/test/unit_test/os_interface/linux/drm_command_stream_tests_1.cpp
View File

@@ -212,7 +212,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenLowPriorityContextWhenFlushingThen
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0, nullptr, false, true, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
csr->flush(batchBuffer, csr->getResidencyAllocations());
EXPECT_NE(cs.getCpuBase(), nullptr);
@@ -228,7 +228,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenLowPriorityContextWhenFlushingThen
}
HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenInvalidAddressWhenFlushingThenSucceeds) {
//allocate command buffer manually
// allocate command buffer manually
char *commandBuffer = new (std::nothrow) char[1024];
ASSERT_NE(nullptr, commandBuffer); // NOLINT(clang-analyzer-cplusplus.NewDeleteLeaks)
@@ -300,7 +300,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenNotAlignedWhenFlushingThenSucceeds
auto &cs = csr->getCS();
auto commandBuffer = static_cast<DrmAllocation *>(cs.getGraphicsAllocation());
//make sure command buffer with offset is not page aligned
// make sure command buffer with offset is not page aligned
ASSERT_NE(0u, (reinterpret_cast<uintptr_t>(commandBuffer->getUnderlyingBuffer()) + 4) & (this->alignment - 1));
ASSERT_EQ(4u, (reinterpret_cast<uintptr_t>(commandBuffer->getUnderlyingBuffer()) + 4) & 0x7F);
@@ -309,7 +309,8 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenNotAlignedWhenFlushingThenSucceeds
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 4, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.startOffset = 4;
csr->flush(batchBuffer, csr->getResidencyAllocations());
EXPECT_EQ(1, mock->ioctlCount.gemUserptr);
@@ -344,7 +345,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenCheckDrmFreeWhenFlushingThenSuccee
auto &cs = csr->getCS();
auto commandBuffer = static_cast<DrmAllocation *>(cs.getGraphicsAllocation());
//make sure command buffer with offset is not page aligned
// make sure command buffer with offset is not page aligned
ASSERT_NE(0u, (reinterpret_cast<uintptr_t>(commandBuffer->getUnderlyingBuffer()) + 4) & (this->alignment - 1));
ASSERT_EQ(4u, (reinterpret_cast<uintptr_t>(commandBuffer->getUnderlyingBuffer()) + 4) & 0x7F);
@@ -356,7 +357,8 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenCheckDrmFreeWhenFlushingThenSuccee
csr->makeResident(allocation);
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 4, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.startOffset = 4;
csr->flush(batchBuffer, csr->getResidencyAllocations());
EXPECT_EQ(1, mock->ioctlCount.gemUserptr);
@@ -384,7 +386,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenCheckDrmFreeCloseFailedWhenFlushin
auto &cs = csr->getCS();
auto commandBuffer = static_cast<DrmAllocation *>(cs.getGraphicsAllocation());
//make sure command buffer with offset is not page aligned
// make sure command buffer with offset is not page aligned
ASSERT_NE(0u, (reinterpret_cast<uintptr_t>(commandBuffer->getUnderlyingBuffer()) + 4) & (this->alignment - 1));
ASSERT_EQ(4u, (reinterpret_cast<uintptr_t>(commandBuffer->getUnderlyingBuffer()) + 4) & 0x7F);
@@ -398,7 +400,8 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, GivenCheckDrmFreeCloseFailedWhenFlushin
csr->makeResident(allocation);
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 4, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.startOffset = 4;
csr->flush(batchBuffer, csr->getResidencyAllocations());
EXPECT_EQ(1, mock->ioctlCount.gemUserptr);
@@ -486,12 +489,12 @@ HWTEST_TEMPLATED_F(DrmCommandStreamBatchingTests, givenCsrWhenDispatchPolicyIsSe
csr->flushTask(cs, 0u, &cs, &cs, &cs, 0u, dispatchFlags, *device);
//make sure command buffer is recorded
// make sure command buffer is recorded
auto &cmdBuffers = mockedSubmissionsAggregator->peekCommandBuffers();
EXPECT_FALSE(cmdBuffers.peekIsEmpty());
EXPECT_NE(nullptr, cmdBuffers.peekHead());
//preemption allocation
// preemption allocation
size_t csrSurfaceCount = (device->getPreemptionMode() == PreemptionMode::MidThread) ? 2 : 0;
csrSurfaceCount += testedCsr->globalFenceAllocation ? 1 : 0;
csrSurfaceCount += testedCsr->clearColorAllocation ? 1 : 0;
@@ -500,7 +503,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamBatchingTests, givenCsrWhenDispatchPolicyIsSe
auto recordedCmdBuffer = cmdBuffers.peekHead();
EXPECT_EQ(3u + csrSurfaceCount, recordedCmdBuffer->surfaces.size());
//try to find all allocations
// try to find all allocations
auto elementInVector = std::find(recordedCmdBuffer->surfaces.begin(), recordedCmdBuffer->surfaces.end(), dummyAllocation);
EXPECT_NE(elementInVector, recordedCmdBuffer->surfaces.end());
@@ -544,7 +547,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamBatchingTests, givenRecordedCommandBufferWhen
csr->setTagAllocation(static_cast<DrmAllocation *>(allocations->getGraphicsAllocation(csr->getRootDeviceIndex())));
auto &submittedCommandBuffer = csr->getCS(1024);
//use some bytes
// use some bytes
submittedCommandBuffer.getSpace(4);
DispatchFlags dispatchFlags = DispatchFlagsHelper::createDefaultDispatchFlags();
@@ -566,13 +569,13 @@ HWTEST_TEMPLATED_F(DrmCommandStreamBatchingTests, givenRecordedCommandBufferWhen
auto commandBufferGraphicsAllocation = submittedCommandBuffer.getGraphicsAllocation();
EXPECT_TRUE(commandBufferGraphicsAllocation->isResident(csr->getOsContext().getContextId()));
//preemption allocation
// preemption allocation
size_t csrSurfaceCount = (device->getPreemptionMode() == PreemptionMode::MidThread) ? 2 : 0;
csrSurfaceCount += testedCsr->globalFenceAllocation ? 1 : 0;
csrSurfaceCount += testedCsr->clearColorAllocation ? 1 : 0;
csrSurfaceCount += testedCsr->getKernelArgsBufferAllocation() ? 1 : 0;
//validate that submited command buffer has what we want
// validate that submited command buffer has what we want
EXPECT_EQ(3u + csrSurfaceCount, this->mock->execBuffer.getBufferCount());
EXPECT_EQ(4u, this->mock->execBuffer.getBatchStartOffset());
EXPECT_EQ(submittedCommandBuffer.getUsed(), this->mock->execBuffer.getBatchLen());
@@ -813,7 +816,8 @@ HWTEST_TEMPLATED_F(DrmCommandStreamDirectSubmissionTest, givenDirectSubmissionFa
auto &cs = csr->getCS();
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 4, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.startOffset = 4;
uint8_t bbStart[64];
batchBuffer.endCmdPtr = &bbStart[0];
@@ -832,7 +836,8 @@ HWTEST_TEMPLATED_F(DrmCommandStreamBlitterDirectSubmissionTest, givenBlitterDire
auto &cs = csr->getCS();
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 4, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.startOffset = 4;
uint8_t bbStart[64];
batchBuffer.endCmdPtr = &bbStart[0];
@@ -855,7 +860,8 @@ HWTEST_TEMPLATED_F(DrmCommandStreamDirectSubmissionTest, givenEnabledDirectSubmi
auto &cs = csr->getCS();
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 4, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.startOffset = 4;
uint8_t bbStart[64];
batchBuffer.endCmdPtr = &bbStart[0];
@@ -874,7 +880,8 @@ HWTEST_TEMPLATED_F(DrmCommandStreamDirectSubmissionTest, givenEnabledDirectSubmi
auto &cs = csr->getCS();
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 4, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.startOffset = 4;
uint8_t bbStart[64];
batchBuffer.endCmdPtr = &bbStart[0];
@@ -892,7 +899,8 @@ HWTEST_TEMPLATED_F(DrmCommandStreamBlitterDirectSubmissionTest, givenEnabledDire
auto &cs = csr->getCS();
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 4, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.startOffset = 4;
uint8_t bbStart[64];
batchBuffer.endCmdPtr = &bbStart[0];

9
shared/test/unit_test/os_interface/linux/drm_command_stream_xehp_and_later_prelim_tests.cpp
View File

@@ -351,11 +351,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamForceTileTest, givenForceExecutionTileThenCor
auto &cs = csr->getCS();
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(),
0, 0, 0, nullptr, false, false,
QueueThrottle::MEDIUM,
QueueSliceCount::defaultSliceCount,
cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
csr->flush(batchBuffer, csr->getResidencyAllocations());
}
@@ -609,7 +605,8 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, DrmImplicitScalingCommandStreamTest, givenUseSingle
csr->CommandStreamReceiver::makeResident(*allocation);
auto &cs = csr->getCS();
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, true};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.useSingleSubdevice = true;
csr->flush(batchBuffer, csr->getResidencyAllocations());

3
shared/test/unit_test/os_interface/linux/drm_memory_manager_localmem_prelim_tests.cpp
View File

@@ -13,6 +13,7 @@
#include "shared/source/memory_manager/memory_banks.h"
#include "shared/source/memory_manager/unified_memory_manager.h"
#include "shared/source/os_interface/linux/allocator_helper.h"
#include "shared/test/common/helpers/batch_buffer_helper.h"
#include "shared/test/common/libult/linux/drm_mock_helper.h"
#include "shared/test/common/libult/linux/drm_mock_prelim_context.h"
#include "shared/test/common/libult/linux/drm_query_mock.h"
@@ -2270,7 +2271,7 @@ struct DrmCommandStreamEnhancedPrelimTest : public DrmCommandStreamEnhancedTempl
LinearStream cs(commandBuffer);
CommandStreamReceiverHw<FamilyType>::addBatchBufferEnd(cs, nullptr);
EncodeNoop<FamilyType>::alignToCacheLine(cs);
this->batchBuffer = BatchBuffer{cs.getGraphicsAllocation(), 0, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
this->batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
this->allocation = this->mm->allocateGraphicsMemoryWithProperties(MockAllocationProperties{csr->getRootDeviceIndex(), MemoryConstants::pageSize});
this->csr->makeResident(*this->allocation);
}

34
shared/test/unit_test/os_interface/windows/device_command_stream_tests.cpp
View File

@@ -288,7 +288,7 @@ TEST_F(WddmCommandStreamTest, GivenOffsetWhenFlushingThenFlushIsSubmittedCorrect
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), offset, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), offset, 0, 0, nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false, false};
csr->flush(batchBuffer, csr->getResidencyAllocations());
EXPECT_EQ(1u, wddm->submitResult.called);
EXPECT_TRUE(wddm->submitResult.success);
@@ -428,7 +428,8 @@ TEST_F(WddmCommandStreamTest, givenWdmmWhenSubmitIsCalledWhenEUCountWouldBeOddTh
wddm->getGtSysInfo()->EUCount = 9;
wddm->getGtSysInfo()->SubSliceCount = 3;
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0, nullptr, false, false, QueueThrottle::LOW, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.throttle = QueueThrottle::LOW;
csr->flush(batchBuffer, csr->getResidencyAllocations());
auto commandHeader = wddm->submitResult.commandHeaderSubmitted;
@@ -446,7 +447,8 @@ TEST_F(WddmCommandStreamTest, givenWdmmWhenSubmitIsCalledAndThrottleIsToLowThenS
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0, nullptr, false, false, QueueThrottle::LOW, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.throttle = QueueThrottle::LOW;
csr->flush(batchBuffer, csr->getResidencyAllocations());
auto commandHeader = wddm->submitResult.commandHeaderSubmitted;
@@ -482,7 +484,8 @@ TEST_F(WddmCommandStreamTest, givenWdmmWhenSubmitIsCalledAndThrottleIsToHighThen
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0, nullptr, false, false, QueueThrottle::HIGH, QueueSliceCount::defaultSliceCount, cs.getUsed(), &cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.throttle = QueueThrottle::HIGH;
csr->flush(batchBuffer, csr->getResidencyAllocations());
auto commandHeader = wddm->submitResult.commandHeaderSubmitted;
@@ -881,7 +884,7 @@ HWTEST_TEMPLATED_F(WddmCommandStreamMockGdiTest, WhenMakingResidentThenResidency
HWTEST_TEMPLATED_F(WddmCommandStreamMockGdiTest, givenRecordedCommandBufferWhenItIsSubmittedThenFlushTaskIsProperlyCalled) {
auto mockCsr = static_cast<MockWddmCsr<FamilyType> *>(csr);
//preemption allocation + sip allocation
// preemption allocation + sip allocation
size_t csrSurfaceCount = 0;
if (device->getPreemptionMode() == PreemptionMode::MidThread) {
csrSurfaceCount = 2;
@@ -1120,9 +1123,7 @@ HWTEST_TEMPLATED_F(WddmCommandStreamMockGdiTest, givenDirectSubmissionFailsThenF
GraphicsAllocation *commandBuffer = memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{csr->getRootDeviceIndex(), MemoryConstants::pageSize});
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0,
nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(),
&cs, commandBuffer->getUnderlyingBuffer(), false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
mockCsr->directSubmission = std::make_unique<MockSubmission>(*device->getDefaultEngine().commandStreamReceiver);
auto res = csr->flush(batchBuffer, csr->getResidencyAllocations());
@@ -1155,9 +1156,9 @@ HWTEST_TEMPLATED_F(WddmCommandStreamMockGdiTest, givenDirectSubmissionEnabledOnR
GraphicsAllocation *commandBuffer = memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{csr->getRootDeviceIndex(), MemoryConstants::pageSize});
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0,
nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(),
&cs, commandBuffer->getUnderlyingBuffer(), false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.endCmdPtr = commandBuffer->getUnderlyingBuffer();
csr->flush(batchBuffer, csr->getResidencyAllocations());
auto directSubmission = reinterpret_cast<MockSubmission *>(mockCsr->directSubmission.get());
EXPECT_TRUE(directSubmission->ringStart);
@@ -1196,9 +1197,9 @@ HWTEST_TEMPLATED_F(WddmCommandStreamMockGdiTest, givenDirectSubmissionEnabledOnB
GraphicsAllocation *commandBuffer = memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{csr->getRootDeviceIndex(), MemoryConstants::pageSize});
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0,
nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(),
&cs, commandBuffer->getUnderlyingBuffer(), false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
batchBuffer.endCmdPtr = commandBuffer->getUnderlyingBuffer();
csr->flush(batchBuffer, csr->getResidencyAllocations());
auto directSubmission = reinterpret_cast<MockSubmission *>(mockCsr->blitterDirectSubmission.get());
EXPECT_TRUE(directSubmission->ringStart);
@@ -1223,9 +1224,8 @@ TEST_F(WddmCommandStreamTest, givenResidencyLoggingAvailableWhenFlushingCommandB
GraphicsAllocation *commandBuffer = memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{csr->getRootDeviceIndex(), MemoryConstants::pageSize});
ASSERT_NE(nullptr, commandBuffer);
LinearStream cs(commandBuffer);
BatchBuffer batchBuffer{cs.getGraphicsAllocation(), 0, 0, 0,
nullptr, false, false, QueueThrottle::MEDIUM, QueueSliceCount::defaultSliceCount, cs.getUsed(),
&cs, nullptr, false};
BatchBuffer batchBuffer = BatchBufferHelper::createDefaultBatchBuffer(cs.getGraphicsAllocation(), &cs, cs.getUsed());
DebugManagerStateRestore restorer;
DebugManager.flags.WddmResidencyLogger.set(1);