Media
/
compute-runtime
mirror of https://github.com/intel/compute-runtime.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Include dynamic SLM in clGetKernelWorkGroupInfo 

Current implementation only takes static slmInlineSize into account.
With this change we also include dynamic SLM passed as a kernel arguments.

Related-To: NEO-5761
Signed-off-by: Fabian Zwolinski <fabian.zwolinski@intel.com>
This commit is contained in:
Fabian Zwolinski 2022-10-10 14:31:29 +00:00 committed by Compute-Runtime-Automation
parent d3f30b4046
commit 320b020dd6
2 changed files with 41 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
opencl/source/kernel/kernel.cpp
View File

@ -530,7 +530,7 @@ cl_int Kernel::getWorkGroupInfo(cl_kernel_work_group_info paramName,
struct size_t3 {
size_t val[3];
} requiredWorkGroupSize;
cl_ulong localMemorySize;
size_t localMemorySize;
const auto &kernelDescriptor = kernelInfo.kernelDescriptor;
size_t preferredWorkGroupSizeMultiple = 0;
cl_ulong scratchSize;
@ -561,7 +561,7 @@ cl_int Kernel::getWorkGroupInfo(cl_kernel_work_group_info paramName,
break;
case CL_KERNEL_LOCAL_MEM_SIZE:
localMemorySize = kernelInfo.kernelDescriptor.kernelAttributes.slmInlineSize;
localMemorySize = static_cast<size_t>(this->getSlmTotalSize());
srcSize = sizeof(localMemorySize);
pSrc = &localMemorySize;
break;

41
opencl/test/unit_test/kernel/kernel_tests.cpp
View File

@ -312,7 +312,7 @@ TEST_F(KernelTests, GivenKernelCompileWorkGroupSizeWhenGettingWorkGroupInfoThenC
EXPECT_EQ(paramValueSize, paramValueSizeRet);
}
TEST_F(KernelTests, GivenRequiredDisabledEUFusionFlagWhenGettingPrefferedWorkGroupSizeMultipleThenCorectValueIsReturned) {
TEST_F(KernelTests, GivenRequiredDisabledEUFusionFlagWhenGettingPrefferedWorkGroupSizeMultipleThenCorrectValueIsReturned) {
KernelInfo kernelInfo = {};
kernelInfo.kernelDescriptor.kernelAttributes.flags.requiresDisabledEUFusion = true;
MockKernel kernel(pProgram, kernelInfo, *pClDevice);
@ -337,7 +337,44 @@ TEST_F(KernelTests, GivenRequiredDisabledEUFusionFlagWhenGettingPrefferedWorkGro
EXPECT_EQ(expectedValue, paramValue);
}
TEST_F(KernelTests, GivenCFEFusedEUDispatchEnabledAndRequiredDisabledUEFusionWhenGettingPrefferedWorkGroupSizeMultipleThenCorectValueIsReturned) {
TEST_F(KernelTests, GivenSlmInlineSizeAndSlmOffsetWhenGettingWorkGroupInfoThenCorrectValueIsReturned) {
MockKernelInfo kernelInfo = {};
kernelInfo.kernelDescriptor.kernelAttributes.slmInlineSize = 100u;
kernelInfo.addArgLocal(0, 0x10, 0x1);
kernelInfo.addArgBuffer(1, 0x20, sizeof(void *));
kernelInfo.addArgBuffer(2, 0x20, sizeof(void *));
kernelInfo.addArgLocal(3, 0x30, 0x10);
MockKernel kernel(pProgram, kernelInfo, *pClDevice);
kernel.kernelArguments.resize(4);
kernel.slmSizes.resize(4);
uint32_t crossThreadData[0x40]{};
crossThreadData[0x20 / sizeof(uint32_t)] = 0x12344321;
kernel.setCrossThreadData(crossThreadData, sizeof(crossThreadData));
kernel.setArgLocal(0, 4096, nullptr);
kernel.setArgLocal(3, 0, nullptr);
cl_kernel_info paramName = CL_KERNEL_LOCAL_MEM_SIZE;
size_t paramValue;
size_t paramValueSize = sizeof(paramValue);
size_t paramValueSizeRet = 0;
size_t expectedValue = 4096 + 0 + 100;
retVal = kernel.getWorkGroupInfo(
paramName,
paramValueSize,
&paramValue,
&paramValueSizeRet);
EXPECT_EQ(CL_SUCCESS, retVal);
EXPECT_EQ(paramValueSize, paramValueSizeRet);
EXPECT_EQ(expectedValue, paramValue);
}
TEST_F(KernelTests, GivenCFEFusedEUDispatchEnabledAndRequiredDisabledUEFusionWhenGettingPrefferedWorkGroupSizeMultipleThenCorrectValueIsReturned) {
DebugManagerStateRestore dbgRestorer;
DebugManager.flags.CFEFusedEUDispatch.set(0);