Media/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-09-10 12:53:42 +08:00
Code Issues Packages Projects Releases Wiki Activity

Require LWS to be in ascending format X >= Y >= Z

Browse Source 
This change fixes problem with memory locality.
When calculating work group size do not take into account
work group sizes where there's bigger number of elements in
higher dimensions namely: Y>X or Z>Y.

Related-To: NEO-5719

Signed-off-by: Krystian Chmielewski <krystian.chmielewski@intel.com>
This commit is contained in:
Krystian Chmielewski
2022-10-17 14:30:21 +00:00
committed by Compute-Runtime-Automation
parent 0fdb308207
commit c224fd0666
2 changed files with 35 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
opencl/test/unit_test/command_queue/local_work_size_tests.cpp
View File

@ -165,7 +165,7 @@ TEST(localWorkSizeTest, given2DimWorkGroupAndSimdEqual8AndNoBarriersWhenComputeC
DebugManagerStateRestore dbgRestore;
DebugManager.flags.EnableComputeWorkSizeSquared.set(true);
//wsInfo maxWorkGroupSize, hasBariers, simdSize, slmTotalSize, hardwareInfo, numThreadsPerSubSlice, localMemorySize, imgUsed, yTiledSurface, disableEUFusion
// wsInfo maxWorkGroupSize, hasBariers, simdSize, slmTotalSize, hardwareInfo, numThreadsPerSubSlice, localMemorySize, imgUsed, yTiledSurface, disableEUFusion
WorkSizeInfo wsInfo(256, 0u, 8, 0u, defaultHwInfo.get(), 32u, 0u, false, false, false);
uint32_t workDim = 2;
size_t workGroup[3] = {10003, 10003, 1};
@ -185,7 +185,7 @@ TEST(localWorkSizeTest, given2DimWorkGroupAndSimdEqual8AndNoBarriersWhenComputeC
}
TEST(localWorkSizeTest, given1DimWorkGroupAndSimdEqual8WhenComputeCalledThenLocalGroupComputed) {
//wsInfo maxWorkGroupSize, hasBariers, simdSize, slmTotalSize, hardwareInfo, numThreadsPerSubSlice, localMemorySize, imgUsed, yTiledSurface, disableEUFusion
// wsInfo maxWorkGroupSize, hasBariers, simdSize, slmTotalSize, hardwareInfo, numThreadsPerSubSlice, localMemorySize, imgUsed, yTiledSurface, disableEUFusion
WorkSizeInfo wsInfo(256, 0u, 8, 0u, defaultHwInfo.get(), 32u, 0u, false, false, false);
uint32_t workDim = 1;
size_t workGroup[3] = {6144, 1, 1};
@ -656,8 +656,8 @@ TEST(localWorkSizeTest, GivenUseStrictRatioWhenLwsIsBeingComputedThenWgsIsCalcul
workGroup[0] = 194;
workGroup[1] = 234;
NEO::computeWorkgroupSizeND(wsInfo, workGroupSize, workGroup, workDim);
EXPECT_EQ(workGroupSize[0], 2u);
EXPECT_EQ(workGroupSize[1], 117u);
EXPECT_EQ(workGroupSize[0], 97u);
EXPECT_EQ(workGroupSize[1], 2u);
EXPECT_EQ(workGroupSize[2], 1u);
workGroup[0] = 100;
@ -694,15 +694,17 @@ TEST(localWorkSizeTest, GivenUseBarriersWhenLwsIsBeingComputedThenWgsIsCalculate
wsInfo.yTiledSurfaces = false;
wsInfo.imgUsed = false;
NEO::computeWorkgroupSizeND(wsInfo, workGroupSize, workGroup, workDim);
EXPECT_EQ(workGroupSize[0], 2u);
EXPECT_EQ(workGroupSize[1], 78u);
EXPECT_EQ(workGroupSize[0], 194u);
EXPECT_EQ(workGroupSize[1], 1u);
EXPECT_EQ(workGroupSize[2], 1u);
wsInfo.useRatio = false;
wsInfo.useStrictRatio = false;
workDim = 3;
workGroup[2] = 4;
NEO::computeWorkgroupSizeND(wsInfo, workGroupSize, workGroup, workDim);
EXPECT_EQ(workGroupSize[0], 2u);
EXPECT_EQ(workGroupSize[1], 78u);
EXPECT_EQ(workGroupSize[0], 194u);
EXPECT_EQ(workGroupSize[1], 1u);
EXPECT_EQ(workGroupSize[2], 1u);
}