Media/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-09-15 13:01:45 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add warning when SLM is not enough for kernel

Browse Source 
Signed-off-by: Jaime Arteaga <jaime.a.arteaga.molina@intel.com>
This commit is contained in:
Jaime Arteaga
2022-08-23 19:16:50 +00:00
committed by Compute-Runtime-Automation
parent 02cf62902b
commit f7b3b20f81
2 changed files with 77 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

74
opencl/test/unit_test/command_queue/local_work_size_tests.cpp
View File

@ -41,6 +41,80 @@ TEST(localWorkSizeTest, givenDisableEUFusionWhenCreatingWorkSizeInfoThenCorrectM
EXPECT_EQ(expectedMinWGS, wsInfo.minWorkGroupSize);
}
TEST(localWorkSizeTest, GivenSlmLargerThanLocalThenWarningIsReturned) {
DebugManagerStateRestore dbgRestorer;
DebugManager.flags.PrintDebugMessages.set(true);
::testing::internal::CaptureStderr();
EXPECT_THROW(WorkSizeInfo wsInfo(256, // maxWorkGroupSize
1u, // hasBariers
8, // simdSize
128u, // slmTotalSize
defaultHwInfo.get(), // hardwareInfo
32u, // numThreadsPerSubSlice
64u, // localMemorySize
false, // imgUsed
false, // yTiledSurface
false // disableEUFusion
),
std::exception);
std::string output = testing::internal::GetCapturedStderr();
EXPECT_EQ(std::string("Size of SLM (128) larger than available (64)\n"), output);
}
TEST(localWorkSizeTest, GivenSlmSmallerThanLocalThenWarningIsNotReturned) {
DebugManagerStateRestore dbgRestorer;
DebugManager.flags.PrintDebugMessages.set(true);
::testing::internal::CaptureStderr();
WorkSizeInfo wsInfo(256, // maxWorkGroupSize
1u, // hasBariers
8, // simdSize
64u, // slmTotalSize
defaultHwInfo.get(), // hardwareInfo
32u, // numThreadsPerSubSlice
128u, // localMemorySize
false, // imgUsed
false, // yTiledSurface
false // disableEUFusion
);
std::string output = testing::internal::GetCapturedStderr();
EXPECT_EQ(std::string(""), output);
}
TEST(localWorkSizeTest, whenSettingHasBarriersWithNoFusedDispatchThenMinWorkGroupSizeIsSetCorrectly) {
DebugManagerStateRestore dbgRestorer;
DebugManager.flags.CFEFusedEUDispatch.set(0);
WorkSizeInfo wsInfo0(256, // maxWorkGroupSize
0u, // hasBariers
8, // simdSize
0u, // slmTotalSize
defaultHwInfo.get(), // hardwareInfo
32u, // numThreadsPerSubSlice
128u, // localMemorySize
false, // imgUsed
false, // yTiledSurface
false // disableEUFusion
);
EXPECT_EQ(0u, wsInfo0.minWorkGroupSize);
WorkSizeInfo wsInfo1(256, // maxWorkGroupSize
1u, // hasBariers
8, // simdSize
0u, // slmTotalSize
defaultHwInfo.get(), // hardwareInfo
32u, // numThreadsPerSubSlice
128u, // localMemorySize
false, // imgUsed
false, // yTiledSurface
false // disableEUFusion
);
EXPECT_NE(0u, wsInfo1.minWorkGroupSize);
}
TEST(localWorkSizeTest, given3DimWorkGroupAndSimdEqual8AndBarriersWhenComputeCalledThenLocalGroupComputedCorrectly) {
WorkSizeInfo wsInfo(256, // maxWorkGroupSize
1u, // hasBariers

3
shared/source/program/kernel_info.cpp
View File

@ -61,6 +61,9 @@ void WorkSizeInfo::setMinWorkGroupSize(const HardwareInfo *hwInfo, bool disableE
minWorkGroupSize = numThreadsPerSubSlice * simdSize / maxBarriersPerHSlice;
}
if (slmTotalSize > 0) {
if (localMemSize < slmTotalSize) {
PRINT_DEBUG_STRING(NEO::DebugManager.flags.PrintDebugMessages.get(), stderr, "Size of SLM (%d) larger than available (%d)\n", slmTotalSize, localMemSize);
}
UNRECOVERABLE_IF(localMemSize < slmTotalSize);
minWorkGroupSize = std::max(maxWorkGroupSize / ((localMemSize / slmTotalSize)), minWorkGroupSize);
}