intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-11 16:45:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: indirect access in external functions

Browse Source 
Read indirect_stateless_count in module external functions.

If greater than 0, mark all kernels that have the has_stack_calls flag
set from this module as having indirect accesses.

Related-To: NEO-7712

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
This commit is contained in:
Dominik Dabek
2024-01-18 13:23:34 +00:00
committed by Compute-Runtime-Automation
parent af5977766d
commit e0ccf22557
12 changed files with 123 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
opencl/source/kernel/kernel.cpp
View File

@@ -296,6 +296,7 @@ cl_int Kernel::initialize() {
kernelDescriptor.kernelAttributes.hasNonKernelArgStore ||
kernelDescriptor.kernelAttributes.hasNonKernelArgAtomic ||
kernelDescriptor.kernelAttributes.hasIndirectStatelessAccess ||
(program->getFunctionPointerWithIndirectAccessExists() && kernelDescriptor.kernelAttributes.flags.useStackCalls) ||
NEO::KernelHelper::isAnyArgumentPtrByValue(kernelDescriptor);
} else {
this->kernelHasIndirectAccess = true;

1
opencl/source/program/process_device_binary.cpp
View File

@@ -289,6 +289,7 @@ cl_int Program::processProgramInfo(ProgramInfo &src, const ClDevice &clDevice) {
}
indirectDetectionVersion = src.indirectDetectionVersion;
functionPointerWithIndirectAccessExists = src.functionPointerWithIndirectAccessExists;
return linkBinary(&clDevice.getDevice(), src.globalConstants.initData, src.globalConstants.size, src.globalVariables.initData,
src.globalVariables.size, src.globalStrings, src.externalFunctions);

2
opencl/source/program/program.h
View File

@@ -247,6 +247,7 @@ class Program : public BaseObject<_cl_program> {
MOCKABLE_VIRTUAL std::string getInternalOptions() const;
uint32_t getMaxRootDeviceIndex() const { return maxRootDeviceIndex; }
uint32_t getIndirectDetectionVersion() const { return indirectDetectionVersion; }
bool getFunctionPointerWithIndirectAccessExists() const { return functionPointerWithIndirectAccessExists; }
void retainForKernel() {
std::unique_lock<std::mutex> lock{lockMutex};
exposedKernels++;
@@ -372,6 +373,7 @@ class Program : public BaseObject<_cl_program> {
ClDeviceVector clDevicesInProgram;
uint32_t indirectDetectionVersion = 0u;
bool functionPointerWithIndirectAccessExists = false;
bool isBuiltIn = false;
bool isGeneratedByIgc = true;

62
opencl/test/unit_test/kernel/kernel_tests.cpp
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2018-2023 Intel Corporation
* Copyright (C) 2018-2024 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -1776,6 +1776,66 @@ HWTEST_F(KernelResidencyTest, givenKernelWithNoKernelArgLoadNorKernelArgStoreNor
memoryManager->freeGraphicsMemory(pKernelInfo->kernelAllocation);
}
HWTEST_F(KernelResidencyTest, givenKernelWithExternalFunctionWithIndirectAccessAndDetectIndirectAccessInKernelEnabledThenKernelHasIndirectAccessIsSetToTrue) {
DebugManagerStateRestore restorer;
debugManager.flags.DetectIndirectAccessInKernel.set(1);
auto pKernelInfo = std::make_unique<KernelInfo>();
pKernelInfo->kernelDescriptor.kernelAttributes.simdSize = 1;
pKernelInfo->kernelDescriptor.kernelAttributes.hasNonKernelArgLoad = false;
pKernelInfo->kernelDescriptor.kernelAttributes.hasNonKernelArgStore = false;
pKernelInfo->kernelDescriptor.kernelAttributes.hasNonKernelArgAtomic = false;
pKernelInfo->kernelDescriptor.kernelAttributes.hasIndirectStatelessAccess = false;
pKernelInfo->kernelDescriptor.kernelAttributes.flags.useStackCalls = true;
auto &commandStreamReceiver = pDevice->getUltCommandStreamReceiver<FamilyType>();
commandStreamReceiver.storeMakeResidentAllocations = true;
auto memoryManager = commandStreamReceiver.getMemoryManager();
pKernelInfo->kernelAllocation = memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{pDevice->getRootDeviceIndex(), MemoryConstants::pageSize});
MockProgram program(toClDeviceVector(*pClDevice));
MockContext ctx;
program.setContext(&ctx);
program.buildInfos[pDevice->getRootDeviceIndex()].globalSurface = new MockGraphicsAllocation();
program.functionPointerWithIndirectAccessExists = true;
std::unique_ptr<MockKernel> kernel(new MockKernel(&program, *pKernelInfo, *pClDevice));
ASSERT_EQ(CL_SUCCESS, kernel->initialize());
EXPECT_TRUE(kernel->getHasIndirectAccess());
memoryManager->freeGraphicsMemory(pKernelInfo->kernelAllocation);
}
HWTEST_F(KernelResidencyTest, givenKernelWithExternalFunctionWithIndirectAccessButNoUseStackCallsAndDetectIndirectAccessInKernelEnabledThenKernelHasIndirectAccessIsSetToFalse) {
DebugManagerStateRestore restorer;
debugManager.flags.DetectIndirectAccessInKernel.set(1);
auto pKernelInfo = std::make_unique<KernelInfo>();
pKernelInfo->kernelDescriptor.kernelAttributes.simdSize = 1;
pKernelInfo->kernelDescriptor.kernelAttributes.hasNonKernelArgLoad = false;
pKernelInfo->kernelDescriptor.kernelAttributes.hasNonKernelArgStore = false;
pKernelInfo->kernelDescriptor.kernelAttributes.hasNonKernelArgAtomic = false;
pKernelInfo->kernelDescriptor.kernelAttributes.hasIndirectStatelessAccess = false;
pKernelInfo->kernelDescriptor.kernelAttributes.flags.useStackCalls = false;
auto &commandStreamReceiver = pDevice->getUltCommandStreamReceiver<FamilyType>();
commandStreamReceiver.storeMakeResidentAllocations = true;
auto memoryManager = commandStreamReceiver.getMemoryManager();
pKernelInfo->kernelAllocation = memoryManager->allocateGraphicsMemoryWithProperties(MockAllocationProperties{pDevice->getRootDeviceIndex(), MemoryConstants::pageSize});
MockProgram program(toClDeviceVector(*pClDevice));
MockContext ctx;
program.setContext(&ctx);
program.buildInfos[pDevice->getRootDeviceIndex()].globalSurface = new MockGraphicsAllocation();
program.functionPointerWithIndirectAccessExists = true;
std::unique_ptr<MockKernel> kernel(new MockKernel(&program, *pKernelInfo, *pClDevice));
ASSERT_EQ(CL_SUCCESS, kernel->initialize());
EXPECT_FALSE(kernel->getHasIndirectAccess());
memoryManager->freeGraphicsMemory(pKernelInfo->kernelAllocation);
}
HWTEST_F(KernelResidencyTest, givenKernelWithPtrByValueArgumentAndDetectIndirectAccessInKernelEnabledThenKernelHasIndirectAccessIsSetToTrue) {
DebugManagerStateRestore restorer;
debugManager.flags.DetectIndirectAccessInKernel.set(1);

3
opencl/test/unit_test/mocks/mock_program.h
View File

@@ -1,5 +1,5 @@
/*
* Copyright (C) 2018-2023 Intel Corporation
* Copyright (C) 2018-2024 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
@@ -55,6 +55,7 @@ class MockProgram : public Program {
using Program::deviceBuildInfos;
using Program::disableZebinIfVmeEnabled;
using Program::extractInternalOptions;
using Program::functionPointerWithIndirectAccessExists;
using Program::getKernelInfo;
using Program::getModuleAllocations;
using Program::internalOptionsToExtract;