intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-25 13:33:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

Correct programming of implicit args on pre-XeHp platforms

Browse Source 
On pre-XeHp platforms implicit args aren't at the beginning of indirect data,
GPU address of implicit args buffer is programmed within cross thread data

Related-To: NEO-5081, IGC-4710
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
This commit is contained in:
Mateusz Jablonski
2022-02-23 13:27:24 +00:00
committed by Compute-Runtime-Automation
parent 06a4d2cc02
commit a2386ad216
12 changed files with 230 additions and 61 deletions
Download Patch File Download Diff File Expand all files Collapse all files

104
shared/test/unit_test/kernel/implicit_args_helper_tests.cpp
View File

@@ -60,10 +60,12 @@ TEST(ImplicitArgsHelperTest, givenNoImplicitArgsWhenGettingSizeForImplicitArgsPr
EXPECT_EQ(0u, ImplicitArgsHelper::getSizeForImplicitArgsPatching(nullptr, kernelDescriptor, hwInfo));
}
TEST(ImplicitArgsHelperTest, givenImplicitArgsWhenGettingSizeForImplicitArgsProgrammingThenCorrectSizeIsReturned) {
TEST(ImplicitArgsHelperTest, givenImplicitArgsWithoutImplicitArgsBufferOffsetInPayloadMappingWhenGettingSizeForImplicitArgsProgrammingThenCorrectSizeIsReturned) {
ImplicitArgs implicitArgs{sizeof(ImplicitArgs)};
KernelDescriptor kernelDescriptor{};
EXPECT_TRUE(isUndefinedOffset<>(kernelDescriptor.payloadMappings.implicitArgs.implcitArgsBuffer));
const auto &hwInfo = *defaultHwInfo;
implicitArgs.simdWidth = 32;
@@ -76,3 +78,103 @@ TEST(ImplicitArgsHelperTest, givenImplicitArgsWhenGettingSizeForImplicitArgsProg
auto localIdsSize = alignUp(PerThreadDataHelper::getPerThreadDataSizeTotal(implicitArgs.simdWidth, hwInfo.capabilityTable.grfSize, 3u, totalWorkgroupSize), MemoryConstants::cacheLineSize);
EXPECT_EQ(localIdsSize + implicitArgs.structSize, ImplicitArgsHelper::getSizeForImplicitArgsPatching(&implicitArgs, kernelDescriptor, hwInfo));
}
TEST(ImplicitArgsHelperTest, givenImplicitArgsWithImplicitArgsBufferOffsetInPayloadMappingWhenGettingSizeForImplicitArgsProgrammingThenCorrectSizeIsReturned) {
ImplicitArgs implicitArgs{sizeof(ImplicitArgs)};
KernelDescriptor kernelDescriptor{};
kernelDescriptor.payloadMappings.implicitArgs.implcitArgsBuffer = 0x10;
EXPECT_TRUE(isValidOffset<>(kernelDescriptor.payloadMappings.implicitArgs.implcitArgsBuffer));
const auto &hwInfo = *defaultHwInfo;
implicitArgs.simdWidth = 32;
implicitArgs.localSizeX = 2;
implicitArgs.localSizeY = 3;
implicitArgs.localSizeZ = 4;
EXPECT_EQ(alignUp(implicitArgs.structSize, MemoryConstants::cacheLineSize), ImplicitArgsHelper::getSizeForImplicitArgsPatching(&implicitArgs, kernelDescriptor, hwInfo));
}
TEST(ImplicitArgsHelperTest, givenImplicitArgsWithoutImplicitArgsBufferOffsetInPayloadMappingWhenPatchingImplicitArgsThenOnlyProperRegionIsPatched) {
ImplicitArgs implicitArgs{sizeof(ImplicitArgs)};
KernelDescriptor kernelDescriptor{};
kernelDescriptor.kernelAttributes.workgroupDimensionsOrder[0] = 0;
kernelDescriptor.kernelAttributes.workgroupDimensionsOrder[1] = 1;
kernelDescriptor.kernelAttributes.workgroupDimensionsOrder[2] = 2;
EXPECT_TRUE(isUndefinedOffset<>(kernelDescriptor.payloadMappings.implicitArgs.implcitArgsBuffer));
const auto &hwInfo = *defaultHwInfo;
implicitArgs.simdWidth = 1;
implicitArgs.localSizeX = 2;
implicitArgs.localSizeY = 3;
implicitArgs.localSizeZ = 4;
auto totalSizeForPatching = ImplicitArgsHelper::getSizeForImplicitArgsPatching(&implicitArgs, kernelDescriptor, hwInfo);
auto totalWorkgroupSize = implicitArgs.localSizeX * implicitArgs.localSizeY * implicitArgs.localSizeZ;
auto localIdsPatchingSize = totalWorkgroupSize * 3 * sizeof(uint16_t);
auto memoryToPatch = std::make_unique<uint8_t[]>(totalSizeForPatching);
uint8_t pattern = 0xcd;
memset(memoryToPatch.get(), pattern, totalSizeForPatching);
auto retVal = ImplicitArgsHelper::patchImplicitArgs(memoryToPatch.get(), implicitArgs, kernelDescriptor, hwInfo, {});
EXPECT_EQ(retVal, ptrOffset(memoryToPatch.get(), totalSizeForPatching));
uint32_t offset = 0;
for (; offset < localIdsPatchingSize; offset++) {
EXPECT_NE(pattern, memoryToPatch.get()[offset]) << offset;
}
for (; offset < totalSizeForPatching - sizeof(implicitArgs); offset++) {
EXPECT_EQ(pattern, memoryToPatch.get()[offset]);
}
for (; offset < totalSizeForPatching; offset++) {
EXPECT_NE(pattern, memoryToPatch.get()[offset]);
}
}
TEST(ImplicitArgsHelperTest, givenImplicitArgsWithImplicitArgsBufferOffsetInPayloadMappingWhenPatchingImplicitArgsThenOnlyProperRegionIsPatched) {
ImplicitArgs implicitArgs{sizeof(ImplicitArgs)};
KernelDescriptor kernelDescriptor{};
kernelDescriptor.payloadMappings.implicitArgs.implcitArgsBuffer = 0x10;
EXPECT_TRUE(isValidOffset<>(kernelDescriptor.payloadMappings.implicitArgs.implcitArgsBuffer));
const auto &hwInfo = *defaultHwInfo;
implicitArgs.simdWidth = 32;
implicitArgs.localSizeX = 2;
implicitArgs.localSizeY = 3;
implicitArgs.localSizeZ = 4;
auto totalSizeForPatching = ImplicitArgsHelper::getSizeForImplicitArgsPatching(&implicitArgs, kernelDescriptor, hwInfo);
EXPECT_EQ(0x80u, totalSizeForPatching);
auto memoryToPatch = std::make_unique<uint8_t[]>(totalSizeForPatching);
uint8_t pattern = 0xcd;
memset(memoryToPatch.get(), pattern, totalSizeForPatching);
auto retVal = ImplicitArgsHelper::patchImplicitArgs(memoryToPatch.get(), implicitArgs, kernelDescriptor, hwInfo, {});
EXPECT_EQ(retVal, ptrOffset(memoryToPatch.get(), totalSizeForPatching));
uint32_t offset = 0;
for (; offset < sizeof(implicitArgs); offset++) {
EXPECT_NE(pattern, memoryToPatch.get()[offset]);
}
for (; offset < totalSizeForPatching; offset++) {
EXPECT_EQ(pattern, memoryToPatch.get()[offset]);
}
}