intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-20 08:36:09 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: use correct allocation type in program init

Browse Source 
Globals surface allocation via USM manager will have correct allocation
type set (instead of just BUFFER) and will use cpu copy when possible.

Related-To: NEO-7796

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
This commit is contained in:
Dominik Dabek
2023-04-04 10:02:03 +00:00
committed by Compute-Runtime-Automation
parent ea15b78a53
commit 1c52017ceb
5 changed files with 16 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
level_zero/core/test/unit_tests/sources/module/test_module.cpp
View File

@@ -2485,8 +2485,8 @@ kernels:
zebin.data(), zebin.size());
auto retVal = moduleTu.processUnpackedBinary();
EXPECT_EQ(retVal, ZE_RESULT_SUCCESS);
EXPECT_EQ(AllocationType::BUFFER, moduleTu.globalConstBuffer->getAllocationType());
EXPECT_EQ(AllocationType::BUFFER, moduleTu.globalVarBuffer->getAllocationType());
EXPECT_EQ(AllocationType::CONSTANT_SURFACE, moduleTu.globalConstBuffer->getAllocationType());
EXPECT_EQ(AllocationType::GLOBAL_SURFACE, moduleTu.globalVarBuffer->getAllocationType());
auto svmAllocsManager = device->getDriverHandle()->getSvmAllocsManager();
auto globalConstBufferAllocType = svmAllocsManager->getSVMAlloc(reinterpret_cast<void *>(moduleTu.globalConstBuffer->getGpuAddress()))->memoryType;

8
shared/source/memory_manager/unified_memory_manager.cpp
View File

@@ -430,7 +430,6 @@ void SVMAllocsManager::removeSVMAlloc(const SvmAllocationData &svmAllocData) {
}
bool SVMAllocsManager::freeSVMAlloc(void *ptr, bool blocking) {
if (SVMDeferFreeAllocs.allocations.size() > 0) {
this->freeSVMAllocDeferImpl();
}
@@ -516,7 +515,6 @@ void SVMAllocsManager::freeSVMAllocImpl(void *ptr, FreePolicyType policy, SvmAll
}
void SVMAllocsManager::freeSVMAllocDeferImpl() {
std::vector<void *> freedPtr;
for (auto iter = SVMDeferFreeAllocs.allocations.begin(); iter != SVMDeferFreeAllocs.allocations.end(); ++iter) {
void *ptr = reinterpret_cast<void *>(iter->second.gpuAllocations.getDefaultGraphicsAllocation()->getGpuAddress());
@@ -764,7 +762,11 @@ AllocationType SVMAllocsManager::getGraphicsAllocationTypeAndCompressionPreferen
if (CompressionSelector::allowStatelessCompression()) {
compressionEnabled = true;
}
allocationType = AllocationType::BUFFER;
if (unifiedMemoryProperties.requestedAllocationType != AllocationType::UNKNOWN) {
allocationType = unifiedMemoryProperties.requestedAllocationType;
} else {
allocationType = AllocationType::BUFFER;
}
}
}
return allocationType;

1
shared/source/memory_manager/unified_memory_manager.h
View File

@@ -123,6 +123,7 @@ class SVMAllocsManager {
Device *device = nullptr;
const RootDeviceIndicesContainer &rootDeviceIndices;
const std::map<uint32_t, DeviceBitfield> &subdeviceBitfields;
AllocationType requestedAllocationType = AllocationType::UNKNOWN;
};
struct SvmCacheAllocationInfo {

8
shared/source/program/program_initialization.cpp
View File

@@ -23,13 +23,13 @@ GraphicsAllocation *allocateGlobalsSurface(NEO::SVMAllocsManager *const svmAlloc
LinkerInput *const linkerInput, const void *initData) {
bool globalsAreExported = false;
GraphicsAllocation *gpuAllocation = nullptr;
auto rootDeviceIndex = device.getRootDeviceIndex();
auto deviceBitfield = device.getDeviceBitfield();
const auto rootDeviceIndex = device.getRootDeviceIndex();
const auto deviceBitfield = device.getDeviceBitfield();
if (linkerInput != nullptr) {
globalsAreExported = constant ? linkerInput->getTraits().exportsGlobalConstants : linkerInput->getTraits().exportsGlobalVariables;
}
const auto allocationType = constant ? AllocationType::CONSTANT_SURFACE : AllocationType::GLOBAL_SURFACE;
if (globalsAreExported && (svmAllocManager != nullptr)) {
RootDeviceIndicesContainer rootDeviceIndices;
rootDeviceIndices.push_back(rootDeviceIndex);
@@ -37,6 +37,7 @@ GraphicsAllocation *allocateGlobalsSurface(NEO::SVMAllocsManager *const svmAlloc
subDeviceBitfields.insert({rootDeviceIndex, deviceBitfield});
NEO::SVMAllocsManager::UnifiedMemoryProperties unifiedMemoryProperties(InternalMemoryType::DEVICE_UNIFIED_MEMORY, rootDeviceIndices, subDeviceBitfields);
unifiedMemoryProperties.device = &device;
unifiedMemoryProperties.requestedAllocationType = allocationType;
auto ptr = svmAllocManager->createUnifiedMemoryAllocation(totalSize, unifiedMemoryProperties);
DEBUG_BREAK_IF(ptr == nullptr);
if (ptr == nullptr) {
@@ -46,7 +47,6 @@ GraphicsAllocation *allocateGlobalsSurface(NEO::SVMAllocsManager *const svmAlloc
UNRECOVERABLE_IF(usmAlloc == nullptr);
gpuAllocation = usmAlloc->gpuAllocations.getGraphicsAllocation(rootDeviceIndex);
} else {
auto allocationType = constant ? AllocationType::CONSTANT_SURFACE : AllocationType::GLOBAL_SURFACE;
gpuAllocation = device.getMemoryManager()->allocateGraphicsMemoryWithProperties({rootDeviceIndex,
true, // allocateMemory
totalSize, allocationType,

4
shared/test/unit_test/program/program_initialization_tests.cpp
View File

@@ -85,6 +85,7 @@ TEST(AllocateGlobalSurfaceTest, GivenSvmAllocsManagerWhenGlobalsAreExportedThenM
ASSERT_NE(nullptr, svmAllocsManager.getSVMAlloc(reinterpret_cast<void *>(static_cast<uintptr_t>(alloc->getGpuAddress()))));
EXPECT_TRUE(alloc->isMemObjectsAllocationWithWritableFlags());
EXPECT_EQ(DEVICE_UNIFIED_MEMORY, svmAllocsManager.getSVMAlloc(reinterpret_cast<void *>(alloc->getGpuAddress()))->memoryType);
EXPECT_EQ(AllocationType::CONSTANT_SURFACE, alloc->getAllocationType());
svmAllocsManager.freeSVMAlloc(reinterpret_cast<void *>(static_cast<uintptr_t>(alloc->getGpuAddress())));
alloc = allocateGlobalsSurface(&svmAllocsManager, device, initData.size(), 0u, true /* constant */, &linkerInputExportGlobalVariables, initData.data());
@@ -92,6 +93,7 @@ TEST(AllocateGlobalSurfaceTest, GivenSvmAllocsManagerWhenGlobalsAreExportedThenM
ASSERT_EQ(initData.size(), alloc->getUnderlyingBufferSize());
EXPECT_EQ(0, memcmp(alloc->getUnderlyingBuffer(), initData.data(), initData.size()));
EXPECT_EQ(nullptr, svmAllocsManager.getSVMAlloc(reinterpret_cast<void *>(static_cast<uintptr_t>(alloc->getGpuAddress()))));
EXPECT_EQ(AllocationType::CONSTANT_SURFACE, alloc->getAllocationType());
device.getMemoryManager()->freeGraphicsMemory(alloc);
alloc = allocateGlobalsSurface(&svmAllocsManager, device, initData.size(), 0u, false /* constant */, &linkerInputExportGlobalConstants, initData.data());
@@ -99,6 +101,7 @@ TEST(AllocateGlobalSurfaceTest, GivenSvmAllocsManagerWhenGlobalsAreExportedThenM
ASSERT_EQ(initData.size(), alloc->getUnderlyingBufferSize());
EXPECT_EQ(0, memcmp(alloc->getUnderlyingBuffer(), initData.data(), initData.size()));
EXPECT_EQ(nullptr, svmAllocsManager.getSVMAlloc(reinterpret_cast<void *>(static_cast<uintptr_t>(alloc->getGpuAddress()))));
EXPECT_EQ(AllocationType::GLOBAL_SURFACE, alloc->getAllocationType());
device.getMemoryManager()->freeGraphicsMemory(alloc);
alloc = allocateGlobalsSurface(&svmAllocsManager, device, initData.size(), 0u, false /* constant */, &linkerInputExportGlobalVariables, initData.data());
@@ -108,6 +111,7 @@ TEST(AllocateGlobalSurfaceTest, GivenSvmAllocsManagerWhenGlobalsAreExportedThenM
EXPECT_NE(nullptr, svmAllocsManager.getSVMAlloc(reinterpret_cast<void *>(static_cast<uintptr_t>(alloc->getGpuAddress()))));
EXPECT_TRUE(alloc->isMemObjectsAllocationWithWritableFlags());
EXPECT_EQ(DEVICE_UNIFIED_MEMORY, svmAllocsManager.getSVMAlloc(reinterpret_cast<void *>(alloc->getGpuAddress()))->memoryType);
EXPECT_EQ(AllocationType::GLOBAL_SURFACE, alloc->getAllocationType());
svmAllocsManager.freeSVMAlloc(reinterpret_cast<void *>(static_cast<uintptr_t>(alloc->getGpuAddress())));
}