intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-21 09:14:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

apply dispatch hint to blocking workload.

Browse Source 
Change-Id: I521d237864db5a6eb0ee78a9b8495552241ba129
Signed-off-by: Lindong Wu <lindong.wu@intel.com>
This commit is contained in:
Lindong Wu
2019-10-28 11:23:14 +08:00
committed by sys_ocldev
parent ec87330fe7
commit 10c3a5eba3
2 changed files with 39 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
runtime/helpers/task_information.cpp
View File

@@ -220,6 +220,11 @@ CompletionStamp &CommandComputeKernel::submit(uint32_t taskLevel, bool terminate
dispatchFlags.l3CacheSettings = L3CachingSettings::l3AndL1On;
}
if (commandQueue.dispatchHints != 0) {
dispatchFlags.engineHints = commandQueue.dispatchHints;
dispatchFlags.epilogueRequired = true;
}
DEBUG_BREAK_IF(taskLevel >= Event::eventNotReady);
gtpinNotifyPreFlushTask(&commandQueue);

34
unit_tests/helpers/task_information_tests.cpp
View File

@@ -287,3 +287,37 @@ HWTEST_F(DispatchFlagsTests, givenCommandWithoutKernelWhenSubmitThenPassCorrectD
EXPECT_EQ(mockCmdQ->isMultiEngineQueue(), mockCsr->passedDispatchFlags.multiEngineQueue);
EXPECT_FALSE(mockCsr->passedDispatchFlags.epilogueRequired);
}
HWTEST_F(DispatchFlagsTests, givenCommandComputeKernelWhenSubmitThenPassCorrectDispatchHints) {
using CsrType = MockCsr1<FamilyType>;
SetUpImpl<CsrType>();
auto mockCmdQ = std::make_unique<MockCommandQueueHw<FamilyType>>(context.get(), device.get(), nullptr);
auto mockCsr = static_cast<CsrType *>(&mockCmdQ->getGpgpuCommandStreamReceiver());
IndirectHeap *ih1 = nullptr, *ih2 = nullptr, *ih3 = nullptr;
mockCmdQ->allocateHeapMemory(IndirectHeap::DYNAMIC_STATE, 1, ih1);
mockCmdQ->allocateHeapMemory(IndirectHeap::INDIRECT_OBJECT, 1, ih2);
mockCmdQ->allocateHeapMemory(IndirectHeap::SURFACE_STATE, 1, ih3);
mockCmdQ->dispatchHints = 1234;
PreemptionMode preemptionMode = device->getPreemptionMode();
auto cmdStream = new LinearStream(device->getMemoryManager()->allocateGraphicsMemoryWithProperties({1, GraphicsAllocation::AllocationType::COMMAND_BUFFER}));
std::vector<Surface *> surfaces;
auto kernelOperation = std::make_unique<KernelOperation>(cmdStream, *mockCmdQ->getGpgpuCommandStreamReceiver().getInternalAllocationStorage());
MockKernelWithInternals kernel(*device);
kernelOperation->setHeaps(ih1, ih2, ih3);
bool flushDC = false;
bool slmUsed = false;
bool ndRangeKernel = false;
bool requiresCoherency = false;
for (auto &surface : surfaces) {
requiresCoherency |= surface->IsCoherent;
}
std::unique_ptr<Command> command(new CommandComputeKernel(*mockCmdQ, kernelOperation, surfaces, flushDC, slmUsed, ndRangeKernel, nullptr, preemptionMode, kernel, 1));
command->submit(20, false);
EXPECT_TRUE(mockCsr->passedDispatchFlags.epilogueRequired);
EXPECT_EQ(1234u, mockCsr->passedDispatchFlags.engineHints);
}