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

fix: correct limitation for num threads per thread group

Browse Source 
taking into account the max work group limit

Resolves: NEO-14922
Related-To: NEO-11881
Signed-off-by: Katarzyna Cencelewska <katarzyna.cencelewska@intel.com>
This commit is contained in:
Katarzyna Cencelewska
2025-05-20 14:41:49 +00:00
committed by Compute-Runtime-Automation
parent 5f80490385
commit 6ad4ad41b1
26 changed files with 171 additions and 228 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
opencl/test/unit_test/helpers/hardware_commands_helper_tests.cpp
View File

@@ -569,7 +569,7 @@ HWCMDTEST_F(IGFX_GEN12LP_CORE, HardwareCommandsTest, whenSendingIndirectStateThe
auto numChannels = modifiedKernelInfo.kernelDescriptor.kernelAttributes.numLocalIdChannels;
auto numGrf = GrfConfig::defaultGrfNumber;
const auto &rootDeviceEnvironment = pDevice->getRootDeviceEnvironment();
size_t expectedIohSize = PerThreadDataHelper::getPerThreadDataSizeTotal(modifiedKernelInfo.getMaxSimdSize(), grfSize, numGrf, numChannels, localWorkSize, !kernelUsesLocalIds, rootDeviceEnvironment);
size_t expectedIohSize = PerThreadDataHelper::getPerThreadDataSizeTotal(modifiedKernelInfo.getMaxSimdSize(), grfSize, numGrf, numChannels, localWorkSize, rootDeviceEnvironment);
ASSERT_LE(expectedIohSize, ioh.getUsed());
auto expectedLocalIds = alignedMalloc(expectedIohSize, 64);
@@ -1295,7 +1295,7 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, HardwareCommandsImplicitArgsTests, givenKernelWithI
generateLocalIDs(expectedLocalIds, expectedImplicitArgs.simdWidth, localSize, workgroupDimOrder, false, grfSize, numGrf, rootDeviceEnvironment);
auto localIdsProgrammingSize = implicitArgsProgrammingSize - ImplicitArgsV0::getAlignedSize();
size_t sizeForLocalIds = PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, numGrf, 3u, totalLocalSize, false, rootDeviceEnvironment);
size_t sizeForLocalIds = PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, numGrf, 3u, totalLocalSize, rootDeviceEnvironment);
EXPECT_EQ(0, memcmp(expectedLocalIds, indirectHeapAllocation->getUnderlyingBuffer(), sizeForLocalIds));
alignedFree(expectedLocalIds);
@@ -1330,7 +1330,7 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, HardwareCommandsImplicitArgsTests, givenKernelWithI
generateLocalIDs(expectedLocalIds, expectedImplicitArgs.simdWidth, localSize, expectedDimOrder, false, grfSize, numGrf, rootDeviceEnvironment);
auto localIdsProgrammingSize = implicitArgsProgrammingSize - ImplicitArgsV0::getAlignedSize();
size_t sizeForLocalIds = PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, numGrf, 3u, totalLocalSize, false, rootDeviceEnvironment);
size_t sizeForLocalIds = PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, numGrf, 3u, totalLocalSize, rootDeviceEnvironment);
EXPECT_EQ(0, memcmp(expectedLocalIds, indirectHeapAllocation->getUnderlyingBuffer(), sizeForLocalIds));
alignedFree(expectedLocalIds);