intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-30 01:35:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

Handle SIMD-1 scenario when programming local ids for implicit args

Browse Source 
according to implicit args design for SIMD-1 local ids are one-by-one

Resolves: NEO-6692
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
This commit is contained in:
Mateusz Jablonski
2022-02-17 12:41:06 +00:00
committed by Compute-Runtime-Automation
parent 28e89b2c30
commit 4f71aaf595
8 changed files with 184 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
opencl/test/unit_test/command_queue/dispatch_walker_tests.cpp
View File

@@ -1561,6 +1561,7 @@ HWTEST_F(DispatchWalkerTest, WhenKernelRequiresImplicitArgsThenIohRequiresMoreSp
Vec3<size_t> localWorkgroupSize(workGroupSize);
auto blockedCommandsData = createBlockedCommandsData(*pCmdQ);
kernelInfo.kernelDescriptor.kernelAttributes.simdSize = 1u;
kernelInfo.kernelDescriptor.kernelAttributes.flags.requiresImplicitArgs = false;
MockKernel kernelWithoutImplicitArgs(program.get(), kernelInfo, *pClDevice);
ASSERT_EQ(CL_SUCCESS, kernelWithoutImplicitArgs.initialize());
@@ -1609,11 +1610,13 @@ HWTEST_F(DispatchWalkerTest, WhenKernelRequiresImplicitArgsThenIohRequiresMoreSp
{
auto numChannels = kernelInfo.kernelDescriptor.kernelAttributes.numLocalIdChannels;
auto simdSize = kernelInfo.getMaxSimdSize();
uint32_t grfSize = sizeof(typename FamilyType::GRF);
auto grfSizeForImplicitArgs = ImplicitArgsHelper::getGrfSize(simdSize, grfSize);
auto size = kernelWithImplicitArgs.getCrossThreadDataSize() +
HardwareCommandsHelper<FamilyType>::getPerThreadDataSizeTotal(kernelInfo.getMaxSimdSize(), grfSize, numChannels, Math::computeTotalElementsCount(localWorkgroupSize)) +
HardwareCommandsHelper<FamilyType>::getPerThreadDataSizeTotal(simdSize, grfSize, numChannels, Math::computeTotalElementsCount(localWorkgroupSize)) +
sizeof(ImplicitArgs) +
alignUp(HardwareCommandsHelper<FamilyType>::getPerThreadDataSizeTotal(kernelInfo.getMaxSimdSize(), grfSize, 3u, Math::computeTotalElementsCount(localWorkgroupSize)), MemoryConstants::cacheLineSize);
alignUp(HardwareCommandsHelper<FamilyType>::getPerThreadDataSizeTotal(simdSize, grfSizeForImplicitArgs, 3u, Math::computeTotalElementsCount(localWorkgroupSize)), MemoryConstants::cacheLineSize);
size = alignUp(size, MemoryConstants::cacheLineSize);
EXPECT_EQ(size, iohSizeWithImplicitArgs);