intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-08 14:02:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

Use DualSubSliceCount to calculate workgroup size

Browse Source 
Related-To: NEO-5719

Signed-off-by: Sebastian Luzynski <sebastian.jozef.luzynski@intel.com>
This commit is contained in:
Sebastian Luzynski
2022-06-15 13:29:47 +00:00
committed by Compute-Runtime-Automation
parent 2c853adac3
commit 8ea5bbd3b5
8 changed files with 49 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
level_zero/core/test/unit_tests/main.cpp
View File

@@ -299,6 +299,7 @@ int main(int argc, char **argv) {
// clang-format off
gtSystemInfo.SliceCount = sliceCount;
gtSystemInfo.SubSliceCount = gtSystemInfo.SliceCount * subSlicePerSliceCount;
gtSystemInfo.DualSubSliceCount = gtSystemInfo.SubSliceCount;
gtSystemInfo.EUCount = gtSystemInfo.SubSliceCount * euPerSubSlice - dieRecovery;
gtSystemInfo.ThreadCount = gtSystemInfo.EUCount * threadsPerEu;
gtSystemInfo.MaxEuPerSubSlice = std::max(gtSystemInfo.MaxEuPerSubSlice, euPerSubSlice);

9
level_zero/core/test/unit_tests/sources/kernel/test_function.cpp
View File

@@ -301,6 +301,15 @@ TEST_F(KernelImpSuggestMaxCooperativeGroupCountTests, GivenNoBarriersOrSlmUsedWh
EXPECT_EQ(expected, getMaxWorkGroupCount());
}
TEST_F(KernelImpSuggestMaxCooperativeGroupCountTests, GivenNoBarriersOrSlmUsedAndDSSCountEqualZeroWhenCalculatingMaxCooperativeGroupCountThenResultIsCalculatedWithSimd) {
auto workGroupSize = lws[0] * lws[1] * lws[2];
auto expected = availableThreadCount / Math::divideAndRoundUp(workGroupSize, simd);
auto mutableHwInfo = neoDevice->getRootDeviceEnvironment().getMutableHardwareInfo();
mutableHwInfo->gtSystemInfo.DualSubSliceCount = 0;
EXPECT_EQ(expected, getMaxWorkGroupCount());
}
TEST_F(KernelImpSuggestMaxCooperativeGroupCountTests, GivenBarriersWhenCalculatingMaxCooperativeGroupCountThenResultIsCalculatedWithRegardToBarriersCount) {
usesBarriers = 1;
auto expected = dssCount * (maxBarrierCount / usesBarriers);