intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-03 06:49:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

Dont pass device to Kernel's ctor

Browse Source 
Kernel should reuse devices from program

Related-To: NEO-5001
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
This commit is contained in:
Mateusz Jablonski
2020-11-12 16:31:22 +01:00
committed by Compute-Runtime-Automation
parent 8323ddbb3b
commit a1b5d5a335
42 changed files with 254 additions and 262 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
opencl/source/kernel/kernel.cpp
View File

@@ -64,17 +64,17 @@ class Surface;
uint32_t Kernel::dummyPatchLocation = 0xbaddf00d;
Kernel::Kernel(Program *programArg, const KernelInfo &kernelInfoArg, const ClDevice &deviceArg, bool schedulerKernel)
Kernel::Kernel(Program *programArg, const KernelInfo &kernelInfoArg, bool schedulerKernel)
: slmTotalSize(kernelInfoArg.workloadInfo.slmStaticSize),
isParentKernel((kernelInfoArg.patchInfo.executionEnvironment != nullptr) ? (kernelInfoArg.patchInfo.executionEnvironment->HasDeviceEnqueue != 0) : false),
isSchedulerKernel(schedulerKernel),
program(programArg),
device(deviceArg),
deviceVector(programArg->getDevices()),
kernelInfo(kernelInfoArg) {
program->retain();
imageTransformer.reset(new ImageTransformer);
maxKernelWorkGroupSize = static_cast<uint32_t>(device.getSharedDeviceInfo().maxWorkGroupSize);
maxKernelWorkGroupSize = static_cast<uint32_t>(deviceVector[0]->getSharedDeviceInfo().maxWorkGroupSize);
}
Kernel::~Kernel() {
@@ -166,7 +166,7 @@ cl_int Kernel::initialize() {
const auto &patchInfo = kernelInfo.patchInfo;
reconfigureKernel();
auto &hwInfo = device.getHardwareInfo();
auto &hwInfo = getDevice().getHardwareInfo();
auto &hwHelper = HwHelper::get(hwInfo.platform.eRenderCoreFamily);
auto maxSimdSize = getKernelInfo().getMaxSimdSize();
@@ -240,15 +240,16 @@ cl_int Kernel::initialize() {
// patch crossthread data and ssh with inline surfaces, if necessary
auto perHwThreadPrivateMemorySize = PatchTokenBinary::getPerHwThreadPrivateSurfaceSize(patchInfo.pAllocateStatelessPrivateSurface, getKernelInfo().getMaxSimdSize());
auto rootDeviceIndex = getDevice().getRootDeviceIndex();
if (perHwThreadPrivateMemorySize) {
privateSurfaceSize = KernelHelper::getPrivateSurfaceSize(perHwThreadPrivateMemorySize, device.getSharedDeviceInfo().computeUnitsUsedForScratch);
privateSurfaceSize = KernelHelper::getPrivateSurfaceSize(perHwThreadPrivateMemorySize, getDevice().getSharedDeviceInfo().computeUnitsUsedForScratch);
DEBUG_BREAK_IF(privateSurfaceSize == 0);
if (privateSurfaceSize > std::numeric_limits<uint32_t>::max()) {
retVal = CL_OUT_OF_RESOURCES;
break;
}
privateSurface = device.getMemoryManager()->allocateGraphicsMemoryWithProperties({device.getRootDeviceIndex(), static_cast<size_t>(privateSurfaceSize), GraphicsAllocation::AllocationType::PRIVATE_SURFACE, device.getDeviceBitfield()});
privateSurface = getDevice().getMemoryManager()->allocateGraphicsMemoryWithProperties({rootDeviceIndex, static_cast<size_t>(privateSurfaceSize), GraphicsAllocation::AllocationType::PRIVATE_SURFACE, getDevice().getDeviceBitfield()});
if (privateSurface == nullptr) {
retVal = CL_OUT_OF_RESOURCES;
break;
@@ -256,7 +257,6 @@ cl_int Kernel::initialize() {
const auto &patch = patchInfo.pAllocateStatelessPrivateSurface;
patchWithImplicitSurface(reinterpret_cast<void *>(privateSurface->getGpuAddressToPatch()), *privateSurface, *patch);
}
auto rootDeviceIndex = device.getRootDeviceIndex();
if (patchInfo.pAllocateStatelessConstantMemorySurfaceWithInitialization) {
DEBUG_BREAK_IF(program->getConstantSurface(rootDeviceIndex) == nullptr);
uintptr_t constMemory = isBuiltIn ? (uintptr_t)program->getConstantSurface(rootDeviceIndex)->getUnderlyingBuffer() : (uintptr_t)program->getConstantSurface(rootDeviceIndex)->getGpuAddressToPatch();
@@ -357,7 +357,7 @@ cl_int Kernel::initialize() {
}
if (isParentKernel) {
program->allocateBlockPrivateSurfaces(device);
program->allocateBlockPrivateSurfaces(getDevice());
}
if (program->isKernelDebugEnabled() && getKernelInfo().patchInfo.pAllocateSystemThreadSurface) {
@@ -631,7 +631,7 @@ cl_int Kernel::getSubGroupInfo(cl_kernel_sub_group_info paramName,
if ((paramName == CL_KERNEL_LOCAL_SIZE_FOR_SUB_GROUP_COUNT) ||
(paramName == CL_KERNEL_MAX_NUM_SUB_GROUPS) ||
(paramName == CL_KERNEL_COMPILE_NUM_SUB_GROUPS)) {
if (device.areOcl21FeaturesEnabled() == false) {
if (getDevice().areOcl21FeaturesEnabled() == false) {
return CL_INVALID_OPERATION;
}
}
@@ -646,7 +646,7 @@ cl_int Kernel::getSubGroupInfo(cl_kernel_sub_group_info paramName,
}
numDimensions = inputValueSize / sizeof(size_t);
if (numDimensions == 0 ||
numDimensions > static_cast<size_t>(device.getDeviceInfo().maxWorkItemDimensions)) {
numDimensions > static_cast<size_t>(getDevice().getDeviceInfo().maxWorkItemDimensions)) {
return CL_INVALID_VALUE;
}
}
@@ -660,7 +660,7 @@ cl_int Kernel::getSubGroupInfo(cl_kernel_sub_group_info paramName,
}
numDimensions = paramValueSize / sizeof(size_t);
if (numDimensions == 0 ||
numDimensions > static_cast<size_t>(device.getDeviceInfo().maxWorkItemDimensions)) {
numDimensions > static_cast<size_t>(getDevice().getDeviceInfo().maxWorkItemDimensions)) {
return CL_INVALID_VALUE;
}
}
@@ -735,7 +735,7 @@ void Kernel::substituteKernelHeap(void *newKernelHeap, size_t newKernelHeapSize)
auto &heapInfo = pKernelInfo->heapInfo;
heapInfo.KernelHeapSize = static_cast<uint32_t>(newKernelHeapSize);
pKernelInfo->isKernelHeapSubstituted = true;
auto memoryManager = device.getMemoryManager();
auto memoryManager = getDevice().getMemoryManager();
auto currentAllocationSize = pKernelInfo->kernelAllocation->getUnderlyingBufferSize();
bool status = false;
@@ -744,7 +744,7 @@ void Kernel::substituteKernelHeap(void *newKernelHeap, size_t newKernelHeapSize)
} else {
memoryManager->checkGpuUsageAndDestroyGraphicsAllocations(pKernelInfo->kernelAllocation);
pKernelInfo->kernelAllocation = nullptr;
status = pKernelInfo->createKernelAllocation(device.getDevice());
status = pKernelInfo->createKernelAllocation(getDevice().getDevice());
}
UNRECOVERABLE_IF(!status);
}
@@ -1053,7 +1053,7 @@ inline void Kernel::makeArgsResident(CommandStreamReceiver &commandStreamReceive
if (kernelArguments[argIndex].object) {
if (kernelArguments[argIndex].type == SVM_ALLOC_OBJ) {
auto pSVMAlloc = (GraphicsAllocation *)kernelArguments[argIndex].object;
auto pageFaultManager = device.getMemoryManager()->getPageFaultManager();
auto pageFaultManager = getDevice().getMemoryManager()->getPageFaultManager();
if (pageFaultManager &&
this->isUnifiedMemorySyncRequired) {
pageFaultManager->moveAllocationToGpuDomain(reinterpret_cast<void *>(pSVMAlloc->getGpuAddress()));
@@ -1076,7 +1076,7 @@ inline void Kernel::makeArgsResident(CommandStreamReceiver &commandStreamReceive
}
void Kernel::makeResident(CommandStreamReceiver &commandStreamReceiver) {
auto rootDeviceIndex = device.getRootDeviceIndex();
auto rootDeviceIndex = getDevice().getRootDeviceIndex();
if (privateSurface) {
commandStreamReceiver.makeResident(*privateSurface);
}
@@ -1126,7 +1126,7 @@ void Kernel::makeResident(CommandStreamReceiver &commandStreamReceiver) {
}
void Kernel::getResidency(std::vector<Surface *> &dst) {
auto rootDeviceIndex = device.getRootDeviceIndex();
auto rootDeviceIndex = getDevice().getRootDeviceIndex();
if (privateSurface) {
GeneralSurface *surface = new GeneralSurface(privateSurface);
dst.push_back(surface);
@@ -1632,7 +1632,7 @@ void Kernel::unsetArg(uint32_t argIndex) {
void Kernel::createReflectionSurface() {
if (this->isParentKernel && kernelReflectionSurface == nullptr) {
auto &hwInfo = device.getHardwareInfo();
auto &hwInfo = getDevice().getHardwareInfo();
auto &hwHelper = HwHelper::get(hwInfo.platform.eRenderCoreFamily);
BlockKernelManager *blockManager = program->getBlockKernelManager();
uint32_t blockCount = static_cast<uint32_t>(blockManager->getCount());
@@ -1678,7 +1678,7 @@ void Kernel::createReflectionSurface() {
kernelReflectionSize += blockCount * alignUp(maxConstantBufferSize, sizeof(void *));
kernelReflectionSize += parentImageCount * sizeof(IGIL_ImageParamters);
kernelReflectionSize += parentSamplerCount * sizeof(IGIL_ParentSamplerParams);
kernelReflectionSurface = device.getMemoryManager()->allocateGraphicsMemoryWithProperties({device.getRootDeviceIndex(), kernelReflectionSize, GraphicsAllocation::AllocationType::DEVICE_QUEUE_BUFFER, device.getDeviceBitfield()});
kernelReflectionSurface = getDevice().getMemoryManager()->allocateGraphicsMemoryWithProperties({getDevice().getRootDeviceIndex(), kernelReflectionSize, GraphicsAllocation::AllocationType::DEVICE_QUEUE_BUFFER, getDevice().getDeviceBitfield()});
for (uint32_t i = 0; i < blockCount; i++) {
const KernelInfo *pBlockInfo = blockManager->getBlockKernelInfo(i);
@@ -1689,12 +1689,12 @@ void Kernel::createReflectionSurface() {
maxConstantBufferSize,
parentSamplerCount,
*pBlockInfo,
device.getHardwareInfo());
getDevice().getHardwareInfo());
uint32_t offset = static_cast<uint32_t>(offsetof(IGIL_KernelDataHeader, m_data) + sizeof(IGIL_KernelAddressData) * i);
uint32_t samplerHeapOffset = static_cast<uint32_t>(alignUp(kernelDataOffset + sizeof(IGIL_KernelData) + curbeParamsForBlocks[i].size() * sizeof(IGIL_KernelCurbeParams), sizeof(void *)));
uint32_t samplerHeapSize = static_cast<uint32_t>(alignUp(pBlockInfo->getSamplerStateArraySize(device.getHardwareInfo()), Sampler::samplerStateArrayAlignment) + pBlockInfo->getBorderColorStateSize());
uint32_t samplerHeapSize = static_cast<uint32_t>(alignUp(pBlockInfo->getSamplerStateArraySize(getDevice().getHardwareInfo()), Sampler::samplerStateArrayAlignment) + pBlockInfo->getBorderColorStateSize());
uint32_t constantBufferOffset = alignUp(samplerHeapOffset + samplerHeapSize, sizeof(void *));
uint32_t samplerParamsOffset = 0;
@@ -1721,7 +1721,7 @@ void Kernel::createReflectionSurface() {
sshTokenOffsetsFromKernelData[i] + kernelDataOffset,
btOffset,
*pBlockInfo,
device.getHardwareInfo());
getDevice().getHardwareInfo());
if (samplerHeapSize > 0) {
void *pDst = ptrOffset(kernelReflectionSurface->getUnderlyingBuffer(), samplerHeapOffset);
@@ -1752,7 +1752,7 @@ void Kernel::createReflectionSurface() {
if (DebugManager.flags.ForceDispatchScheduler.get()) {
if (this->isSchedulerKernel && kernelReflectionSurface == nullptr) {
kernelReflectionSurface = device.getMemoryManager()->allocateGraphicsMemoryWithProperties({device.getRootDeviceIndex(), MemoryConstants::pageSize, GraphicsAllocation::AllocationType::DEVICE_QUEUE_BUFFER, device.getDeviceBitfield()});
kernelReflectionSurface = getDevice().getMemoryManager()->allocateGraphicsMemoryWithProperties({getDevice().getRootDeviceIndex(), MemoryConstants::pageSize, GraphicsAllocation::AllocationType::DEVICE_QUEUE_BUFFER, getDevice().getDeviceBitfield()});
}
}
}
@@ -1792,7 +1792,7 @@ size_t Kernel::getInstructionHeapSizeForExecutionModel() const {
}
void Kernel::patchBlocksCurbeWithConstantValues() {
auto rootDeviceIndex = device.getRootDeviceIndex();
auto rootDeviceIndex = getDevice().getRootDeviceIndex();
BlockKernelManager *blockManager = program->getBlockKernelManager();
uint32_t blockCount = static_cast<uint32_t>(blockManager->getCount());
@@ -2201,7 +2201,7 @@ void Kernel::provideInitializationHints() {
}
if (patchInfo.mediavfestate) {
auto scratchSize = patchInfo.mediavfestate->PerThreadScratchSpace;
scratchSize *= device.getSharedDeviceInfo().computeUnitsUsedForScratch * getKernelInfo().getMaxSimdSize();
scratchSize *= getDevice().getSharedDeviceInfo().computeUnitsUsedForScratch * getKernelInfo().getMaxSimdSize();
if (scratchSize > 0) {
context->providePerformanceHint(CL_CONTEXT_DIAGNOSTICS_LEVEL_BAD_INTEL, REGISTER_PRESSURE_TOO_HIGH,
kernelInfo.kernelDescriptor.kernelMetadata.kernelName.c_str(), scratchSize);
@@ -2331,8 +2331,8 @@ void Kernel::resolveArgs() {
}
bool Kernel::canTransformImages() const {
return device.getHardwareInfo().platform.eRenderCoreFamily >= IGFX_GEN9_CORE &&
device.getHardwareInfo().platform.eRenderCoreFamily <= IGFX_GEN11LP_CORE;
return getDevice().getHardwareInfo().platform.eRenderCoreFamily >= IGFX_GEN9_CORE &&
getDevice().getHardwareInfo().platform.eRenderCoreFamily <= IGFX_GEN11LP_CORE;
}
void Kernel::fillWithBuffersForAuxTranslation(MemObjsForAuxTranslation &memObjsForAuxTranslation) {
@@ -2354,7 +2354,7 @@ void Kernel::fillWithBuffersForAuxTranslation(MemObjsForAuxTranslation &memObjsF
}
void Kernel::getAllocationsForCacheFlush(CacheFlushAllocationsVec &out) const {
if (false == HwHelper::cacheFlushAfterWalkerSupported(device.getHardwareInfo())) {
if (false == HwHelper::cacheFlushAfterWalkerSupported(getDevice().getHardwareInfo())) {
return;
}
for (GraphicsAllocation *alloc : this->kernelArgRequiresCacheFlush) {
@@ -2365,7 +2365,7 @@ void Kernel::getAllocationsForCacheFlush(CacheFlushAllocationsVec &out) const {
out.push_back(alloc);
}
auto global = getProgram()->getGlobalSurface(device.getRootDeviceIndex());
auto global = getProgram()->getGlobalSurface(getDevice().getRootDeviceIndex());
if (global != nullptr) {
out.push_back(global);
}

10
opencl/source/kernel/kernel.h
View File

@@ -80,9 +80,7 @@ class Kernel : public BaseObject<_cl_kernel> {
cl_int retVal;
kernel_t *pKernel = nullptr;
auto clDevice = program->getDevices()[0];
pKernel = new kernel_t(program, kernelInfo, *clDevice);
pKernel = new kernel_t(program, kernelInfo);
retVal = pKernel->initialize();
if (retVal != CL_SUCCESS) {
@@ -191,7 +189,7 @@ class Kernel : public BaseObject<_cl_kernel> {
}
const ClDevice &getDevice() const {
return device;
return *deviceVector[0];
}
Context &getContext() const {
@@ -501,7 +499,7 @@ class Kernel : public BaseObject<_cl_kernel> {
void patchWithImplicitSurface(void *ptrToPatchInCrossThreadData, GraphicsAllocation &allocation, const PatchTokenT &patch);
void getParentObjectCounts(ObjectCounts &objectCount);
Kernel(Program *programArg, const KernelInfo &kernelInfoArg, const ClDevice &deviceArg, bool schedulerKernel = false);
Kernel(Program *programArg, const KernelInfo &kernelInfoArg, bool schedulerKernel = false);
void provideInitializationHints();
void patchBlocksCurbeWithConstantValues();
@@ -514,7 +512,7 @@ class Kernel : public BaseObject<_cl_kernel> {
bool allocationForCacheFlush(GraphicsAllocation *argAllocation) const;
Program *program;
Context *context = nullptr;
const ClDevice &device;
const ClDeviceVector &deviceVector;
const KernelInfo &kernelInfo;
std::vector<SimpleKernelArgInfo> kernelArguments;