Revert "fix: correctly patch implicit arg buffer in indirect data"

This reverts commit e345d55fe5.

Signed-off-by: Compute-Runtime-Validation <compute-runtime-validation@intel.com>
This commit is contained in:
Compute-Runtime-Validation
2025-05-06 03:46:46 +02:00
committed by Compute-Runtime-Automation
parent 3fa9229483
commit 26d3c7527c
10 changed files with 33 additions and 44 deletions

View File

@@ -924,11 +924,11 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, CmdlistAppendLaunchKernelWithImplicitArgsTests, giv
auto grfSize = ImplicitArgsHelper::getGrfSize(expectedImplicitArgs.simdWidth);
auto numGrf = GrfConfig::defaultGrfNumber;
auto expectedLocalIds = alignedMalloc(implicitArgsProgrammingSize - ImplicitArgsV0::getAlignedSize(), MemoryConstants::cacheLineSize);
auto expectedLocalIds = alignedMalloc(implicitArgsProgrammingSize - ImplicitArgsV0::getSize(), MemoryConstants::cacheLineSize);
const auto &rootDeviceEnvironment = device->getNEODevice()->getRootDeviceEnvironment();
generateLocalIDs(expectedLocalIds, expectedImplicitArgs.simdWidth, localSize, workgroupDimOrder, false, grfSize, numGrf, rootDeviceEnvironment);
auto localIdsProgrammingSize = implicitArgsProgrammingSize - ImplicitArgsV0::getAlignedSize();
auto localIdsProgrammingSize = implicitArgsProgrammingSize - ImplicitArgsV0::getSize();
size_t sizeForLocalIds = NEO::PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, numGrf, 3u, totalLocalSize, !kernelRequiresGenerationOfLocalIdsByRuntime, rootDeviceEnvironment);
EXPECT_EQ(0, memcmp(expectedLocalIds, indirectHeapAllocation->getUnderlyingBuffer(), sizeForLocalIds));
@@ -944,7 +944,7 @@ HWCMDTEST_F(IGFX_GEN12LP_CORE, CmdlistAppendLaunchKernelWithImplicitArgsTests, g
auto implicitArgsInIndirectData = indirectHeapAllocation->getUnderlyingBuffer();
EXPECT_EQ(0, memcmp(implicitArgsInIndirectData, &expectedImplicitArgs, ImplicitArgsV0::getSize()));
auto crossThreadDataInIndirectData = ptrOffset(indirectHeapAllocation->getUnderlyingBuffer(), ImplicitArgsV0::getAlignedSize());
auto crossThreadDataInIndirectData = ptrOffset(indirectHeapAllocation->getUnderlyingBuffer(), alignUp(ImplicitArgsV0::getSize(), 64));
auto programmedImplicitArgsGpuVA = reinterpret_cast<uint64_t *>(crossThreadDataInIndirectData)[0];
EXPECT_EQ(indirectHeapAllocation->getGpuAddress(), programmedImplicitArgsGpuVA);
@@ -975,7 +975,7 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, CmdlistAppendLaunchKernelWithImplicitArgsTests, giv
const auto &rootDeviceEnvironment = device->getNEODevice()->getRootDeviceEnvironment();
generateLocalIDs(expectedLocalIds, expectedImplicitArgs.simdWidth, localSize, expectedDimOrder, false, grfSize, numGrf, rootDeviceEnvironment);
auto localIdsProgrammingSize = implicitArgsProgrammingSize - ImplicitArgsV0::getAlignedSize();
auto localIdsProgrammingSize = implicitArgsProgrammingSize - ImplicitArgsV0::getSize();
size_t sizeForLocalIds = NEO::PerThreadDataHelper::getPerThreadDataSizeTotal(expectedImplicitArgs.simdWidth, grfSize, numGrf, 3u, totalLocalSize, !kernelRequiresGenerationOfLocalIdsByRuntime, rootDeviceEnvironment);
EXPECT_EQ(0, memcmp(expectedLocalIds, indirectHeapAllocation->getUnderlyingBuffer(), sizeForLocalIds));
@@ -1004,7 +1004,7 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, CmdlistAppendLaunchKernelWithImplicitArgsTests, giv
EXPECT_EQ(0, memcmp(expectedLocalIds, indirectHeapAllocation->getUnderlyingBuffer(), sizeof(expectedLocalIds)));
auto localIdsProgrammingSize = implicitArgsProgrammingSize - ImplicitArgsV0::getAlignedSize();
auto localIdsProgrammingSize = implicitArgsProgrammingSize - ImplicitArgsV0::getSize();
EXPECT_EQ(alignUp(sizeof(expectedLocalIds), MemoryConstants::cacheLineSize), localIdsProgrammingSize);

View File

@@ -942,7 +942,7 @@ struct CommandListAppendLaunchKernelWithImplicitArgs : CommandListAppendLaunchKe
if (FamilyType::supportsCmdSet(IGFX_XE_HP_CORE)) {
const auto &rootDeviceEnvironment = device->getNEODevice()->getRootDeviceEnvironment();
auto implicitArgsProgrammingSize = ImplicitArgsHelper::getSizeForImplicitArgsPatching(kernel.pImplicitArgs.get(), kernel.getKernelDescriptor(), !kernel.kernelRequiresGenerationOfLocalIdsByRuntime, rootDeviceEnvironment);
return implicitArgsProgrammingSize - kernel.pImplicitArgs->getAlignedSize();
return implicitArgsProgrammingSize - kernel.pImplicitArgs->v0.header.structSize;
} else {
return 0u;
}