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

fix: Unify logic calculating threads per work group part 3

Browse Source 
Related-To: NEO-8087
Signed-off-by: Cencelewska, Katarzyna <katarzyna.cencelewska@intel.com>
This commit is contained in:
Cencelewska, Katarzyna
2023-06-30 14:17:13 +00:00
committed by Compute-Runtime-Automation
parent d2f1cf98d7
commit 61f701aba5
17 changed files with 93 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
level_zero/core/source/kernel/kernel_imp.cpp
View File

@@ -345,7 +345,7 @@ ze_result_t KernelImp::setGroupSize(uint32_t groupSizeX, uint32_t groupSizeY,
auto grfSize = this->module->getDevice()->getHwInfo().capabilityTable.grfSize;
uint32_t perThreadDataSizeForWholeThreadGroupNeeded =
static_cast<uint32_t>(NEO::PerThreadDataHelper::getPerThreadDataSizeTotal(
simdSize, grfSize, numChannels, itemsInGroup));
simdSize, grfSize, numChannels, itemsInGroup, !kernelRequiresGenerationOfLocalIdsByRuntime, gfxCoreHelper));
if (perThreadDataSizeForWholeThreadGroupNeeded >
perThreadDataSizeForWholeThreadGroupAllocated) {
alignedFree(perThreadDataForWholeThreadGroup);

7
level_zero/core/test/unit_tests/sources/cmdlist/test_cmdlist_append_launch_kernel_2.cpp
View File

@@ -991,7 +991,8 @@ struct CmdlistAppendLaunchKernelWithImplicitArgsTests : CmdlistAppendLaunchKerne
result = commandList->appendLaunchKernel(kernel->toHandle(), &groupCount, nullptr, 0, nullptr, launchParams, false);
EXPECT_EQ(ZE_RESULT_SUCCESS, result);
implicitArgsProgrammingSize = ImplicitArgsHelper::getSizeForImplicitArgsPatching(&expectedImplicitArgs, *kernelDescriptor);
const auto &gfxCoreHelper = device->getGfxCoreHelper();
implicitArgsProgrammingSize = ImplicitArgsHelper::getSizeForImplicitArgsPatching(&expectedImplicitArgs, *kernelDescriptor, !kernelRequiresGenerationOfLocalIdsByRuntime, gfxCoreHelper);
auto sizeCrossThreadData = kernel->getCrossThreadDataSize();
auto sizePerThreadDataForWholeGroup = kernel->getPerThreadDataSizeForWholeThreadGroup();
EXPECT_EQ(indirectHeap->getUsed(), sizeCrossThreadData + sizePerThreadDataForWholeGroup + implicitArgsProgrammingSize);
@@ -1027,7 +1028,7 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, CmdlistAppendLaunchKernelWithImplicitArgsTests, giv
generateLocalIDs(expectedLocalIds, expectedImplicitArgs.simdWidth, localSize, workgroupDimOrder, false, grfSize, gfxCoreHelper);
auto localIdsProgrammingSize = implicitArgsProgrammingSize - sizeof(ImplicitArgs);
size_t sizeForLocalIds = NEO::PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, 3u, totalLocalSize);
size_t sizeForLocalIds = NEO::PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, 3u, totalLocalSize, !kernelRequiresGenerationOfLocalIdsByRuntime, gfxCoreHelper);
EXPECT_EQ(0, memcmp(expectedLocalIds, indirectHeapAllocation->getUnderlyingBuffer(), sizeForLocalIds));
alignedFree(expectedLocalIds);
@@ -1073,7 +1074,7 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, CmdlistAppendLaunchKernelWithImplicitArgsTests, giv
generateLocalIDs(expectedLocalIds, expectedImplicitArgs.simdWidth, localSize, expectedDimOrder, false, grfSize, gfxCoreHelper);
auto localIdsProgrammingSize = implicitArgsProgrammingSize - sizeof(ImplicitArgs);
size_t sizeForLocalIds = NEO::PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, 3u, totalLocalSize);
size_t sizeForLocalIds = NEO::PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, 3u, totalLocalSize, !kernelRequiresGenerationOfLocalIdsByRuntime, gfxCoreHelper);
EXPECT_EQ(0, memcmp(expectedLocalIds, indirectHeapAllocation->getUnderlyingBuffer(), sizeForLocalIds));
alignedFree(expectedLocalIds);

3
level_zero/core/test/unit_tests/sources/cmdlist/test_cmdlist_append_launch_kernel_3.cpp
View File

@@ -2062,7 +2062,8 @@ struct CommandListAppendLaunchKernelWithImplicitArgs : CommandListAppendLaunchKe
template <typename FamilyType>
uint64_t getIndirectHeapOffsetForImplicitArgsBuffer(const Mock<::L0::KernelImp> &kernel) {
if (FamilyType::supportsCmdSet(IGFX_XE_HP_CORE)) {
auto implicitArgsProgrammingSize = ImplicitArgsHelper::getSizeForImplicitArgsPatching(kernel.pImplicitArgs.get(), kernel.getKernelDescriptor());
const auto &gfxCoreHelper = device->getGfxCoreHelper();
auto implicitArgsProgrammingSize = ImplicitArgsHelper::getSizeForImplicitArgsPatching(kernel.pImplicitArgs.get(), kernel.getKernelDescriptor(), !kernel.kernelRequiresGenerationOfLocalIdsByRuntime, gfxCoreHelper);
return implicitArgsProgrammingSize - sizeof(ImplicitArgs);
} else {
return 0u;