intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-05 18:06:32 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: correct suggested number of work groups for concurrent kernels on PVC

Browse Source 
value depends on CCS count:
- single CCS mode (default) - no limitations
- two CCS mode - 25% available
- four CCS mode - 12.5% available

Related-To: NEO-8377
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
This commit is contained in:
Mateusz Jablonski
2023-09-14 13:53:58 +00:00
committed by Compute-Runtime-Automation
parent 513045bd90
commit 6fc673b0fe
8 changed files with 31 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
level_zero/core/test/unit_tests/xe_hpc_core/pvc/test_kernel_pvc.cpp
View File

@@ -22,19 +22,19 @@ using KernelImpSuggestMaxCooperativeGroupCountTestsPvc = KernelImpSuggestMaxCoop
PVCTEST_F(KernelImpSuggestMaxCooperativeGroupCountTestsPvc, GivenNoBarriersOrSlmUsedWhenCalculatingMaxCooperativeGroupCountThenResultIsCalculatedWithSimd) {
auto workGroupSize = lws[0] * lws[1] * lws[2];
auto expected = (availableThreadCount / Math::divideAndRoundUp(workGroupSize, simd)) / PVC::numberOfpartsInTileForConcurrentKernels;
auto expected = availableThreadCount / Math::divideAndRoundUp(workGroupSize, simd);
EXPECT_EQ(expected, getMaxWorkGroupCount());
}
PVCTEST_F(KernelImpSuggestMaxCooperativeGroupCountTestsPvc, GivenBarriersWhenCalculatingMaxCooperativeGroupCountThenResultIsCalculatedWithRegardToBarriersCount) {
usesBarriers = 1;
auto expected = (dssCount * (maxBarrierCount / usesBarriers)) / PVC::numberOfpartsInTileForConcurrentKernels;
auto expected = dssCount * (maxBarrierCount / usesBarriers);
EXPECT_EQ(expected, getMaxWorkGroupCount());
}
PVCTEST_F(KernelImpSuggestMaxCooperativeGroupCountTestsPvc, GivenUsedSlmSizeWhenCalculatingMaxCooperativeGroupCountThenResultIsCalculatedWithRegardToUsedSlmSize) {
usedSlm = 64 * KB;
auto expected = (availableSlm / usedSlm) / PVC::numberOfpartsInTileForConcurrentKernels;
auto expected = availableSlm / usedSlm;
EXPECT_EQ(expected, getMaxWorkGroupCount());
}