intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-03 23:03:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

Store SLM sizes per root device in Kernel

Browse Source 
reduce usages of getDefaultKernelInfo

Related-To: NEO-5001
Signed-off-by: Mateusz Jablonski <mateusz.jablonski@intel.com>
This commit is contained in:
Mateusz Jablonski
2020-12-18 14:47:42 +00:00
committed by Compute-Runtime-Automation
parent 09bdd2ad09
commit be7ae13911
15 changed files with 48 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
opencl/source/kernel/kernel.cpp
View File

@@ -65,8 +65,7 @@ class Surface;
uint32_t Kernel::dummyPatchLocation = 0xbaddf00d;
Kernel::Kernel(Program *programArg, const KernelInfoContainer &kernelInfosArg, bool schedulerKernel)
: slmTotalSize(kernelInfosArg[programArg->getDevices()[0]->getRootDeviceIndex()]->workloadInfo.slmStaticSize),
isParentKernel(kernelInfosArg[programArg->getDevices()[0]->getRootDeviceIndex()]->kernelDescriptor.kernelAttributes.flags.usesDeviceSideEnqueue),
: isParentKernel(kernelInfosArg[programArg->getDevices()[0]->getRootDeviceIndex()]->kernelDescriptor.kernelAttributes.flags.usesDeviceSideEnqueue),
isSchedulerKernel(schedulerKernel),
executionEnvironment(programArg->getExecutionEnvironment()),
program(programArg),
@@ -78,7 +77,9 @@ Kernel::Kernel(Program *programArg, const KernelInfoContainer &kernelInfosArg, b
program->retainForKernel();
imageTransformer.reset(new ImageTransformer);
for (const auto &pClDevice : deviceVector) {
kernelDeviceInfos[pClDevice->getRootDeviceIndex()].maxKernelWorkGroupSize = static_cast<uint32_t>(pClDevice->getSharedDeviceInfo().maxWorkGroupSize);
auto rootDeviceIndex = pClDevice->getRootDeviceIndex();
kernelDeviceInfos[rootDeviceIndex].maxKernelWorkGroupSize = static_cast<uint32_t>(pClDevice->getSharedDeviceInfo().maxWorkGroupSize);
kernelDeviceInfos[rootDeviceIndex].slmTotalSize = kernelInfosArg[rootDeviceIndex]->workloadInfo.slmStaticSize;
}
}
@@ -100,7 +101,7 @@ Kernel::~Kernel() {
}
for (uint32_t i = 0; i < patchedArgumentsNum; i++) {
if (getDefaultKernelInfo().kernelArgInfo.at(i).isSampler) {
if (SAMPLER_OBJ == getKernelArguments()[i].type) {
auto sampler = castToObject<Sampler>(kernelArguments.at(i).object);
if (sampler) {
sampler->decRefInternal();
@@ -372,6 +373,8 @@ cl_int Kernel::initialize() {
if (program->isKernelDebugEnabled() && kernelInfo.patchInfo.pAllocateSystemThreadSurface) {
debugEnabled = true;
}
auto numArgs = kernelInfo.kernelArgInfo.size();
kernelDeviceInfo.slmSizes.resize(numArgs);
isDeviceInitialized.set(rootDeviceIndex);
}
@@ -384,13 +387,11 @@ cl_int Kernel::initialize() {
auto &defaultKernelInfo = getDefaultKernelInfo();
auto numArgs = defaultKernelInfo.kernelArgInfo.size();
kernelArguments.resize(numArgs);
slmSizes.resize(numArgs);
kernelArgHandlers.resize(numArgs);
kernelArgRequiresCacheFlush.resize(numArgs);
for (uint32_t i = 0; i < numArgs; ++i) {
storeKernelArg(i, NONE_OBJ, nullptr, nullptr, 0);
slmSizes[i] = 0;
// set the argument handler
auto &argInfo = defaultKernelInfo.kernelArgInfo[i];
@@ -483,7 +484,6 @@ cl_int Kernel::getInfo(cl_kernel_info paramName, size_t paramValueSize,
const _cl_context *ctxt;
cl_uint refCount = 0;
uint64_t nonCannonizedGpuAddress = 0llu;
auto defaultRootDeviceIndex = getDevices()[0]->getRootDeviceIndex();
auto &defaultKernelInfo = getKernelInfo(defaultRootDeviceIndex);
switch (paramName) {
@@ -1112,7 +1112,7 @@ uint32_t Kernel::getMaxWorkGroupCount(const cl_uint workDim, const size_t *local
availableThreadCount,
dssCount,
dssCount * KB * hardwareInfo.capabilityTable.slmSize,
hwHelper.alignSlmSize(slmTotalSize),
hwHelper.alignSlmSize(kernelDeviceInfos[rootDeviceIndex].slmTotalSize),
static_cast<uint32_t>(hwHelper.getMaxBarrierRegisterPerSlice()),
hwHelper.getBarriersCountFromHasBarriers(barrierCount),
workDim,
@@ -1280,7 +1280,7 @@ cl_int Kernel::setArgLocal(uint32_t argIndex,
storeKernelArg(argIndex, SLM_OBJ, nullptr, argVal, argSize);
slmSizes[argIndex] = argSize;
kernelDeviceInfos[rootDeviceIndex].slmSizes[argIndex] = argSize;
// Extract our current slmOffset
auto slmOffset = *ptrOffset(crossThreadData,
@@ -1291,7 +1291,7 @@ cl_int Kernel::setArgLocal(uint32_t argIndex,
// Update all slm offsets after this argIndex
++argIndex;
while (argIndex < slmSizes.size()) {
while (argIndex < kernelDeviceInfos[rootDeviceIndex].slmSizes.size()) {
const auto &kernelArgInfo = defaultKernelInfo.kernelArgInfo[argIndex];
auto slmAlignment = kernelArgInfo.slmAlignment;
@@ -1306,11 +1306,11 @@ cl_int Kernel::setArgLocal(uint32_t argIndex,
*patchLocation = slmOffset;
}
slmOffset += static_cast<uint32_t>(slmSizes[argIndex]);
slmOffset += static_cast<uint32_t>(kernelDeviceInfos[rootDeviceIndex].slmSizes[argIndex]);
++argIndex;
}
slmTotalSize = defaultKernelInfo.workloadInfo.slmStaticSize + alignUp(slmOffset, KB);
kernelDeviceInfos[rootDeviceIndex].slmTotalSize = defaultKernelInfo.workloadInfo.slmStaticSize + alignUp(slmOffset, KB);
return CL_SUCCESS;
}
@@ -2679,4 +2679,7 @@ void Kernel::setWorkDim(uint32_t rootDeviceIndex, uint32_t workDim) {
uint32_t Kernel::getMaxKernelWorkGroupSize(uint32_t rootDeviceIndex) const {
return kernelDeviceInfos[rootDeviceIndex].maxKernelWorkGroupSize;
}
uint32_t Kernel::getSlmTotalSize(uint32_t rootDeviceIndex) const {
return kernelDeviceInfos[rootDeviceIndex].slmTotalSize;
}
} // namespace NEO

9
opencl/source/kernel/kernel.h
View File

@@ -182,7 +182,7 @@ class Kernel : public BaseObject<_cl_kernel> {
}
size_t getKernelArgsNumber() const {
return getDefaultKernelInfo().kernelArgInfo.size();
return kernelArguments.size();
}
bool requiresSshForBuffers(uint32_t rootDeviceIndex) const {
@@ -308,11 +308,8 @@ class Kernel : public BaseObject<_cl_kernel> {
static uint32_t dummyPatchLocation;
std::vector<size_t> slmSizes;
uint32_t allBufferArgsStateful = CL_TRUE;
uint32_t slmTotalSize;
bool isBuiltIn = false;
const bool isParentKernel;
const bool isSchedulerKernel;
@@ -406,6 +403,7 @@ class Kernel : public BaseObject<_cl_kernel> {
void setNumWorkGroupsValues(uint32_t rootDeviceIndex, uint32_t numWorkGroupsX, uint32_t numWorkGroupsY, uint32_t numWorkGroupsZ);
void setWorkDim(uint32_t rootDeviceIndex, uint32_t workDim);
uint32_t getMaxKernelWorkGroupSize(uint32_t rootDeviceIndex) const;
uint32_t getSlmTotalSize(uint32_t rootDeviceIndex) const;
protected:
struct ObjectCounts {
@@ -574,6 +572,9 @@ class Kernel : public BaseObject<_cl_kernel> {
size_t numberOfBindingTableStates = 0u;
size_t localBindingTableOffset = 0u;
std::vector<size_t> slmSizes;
uint32_t slmTotalSize = 0u;
std::unique_ptr<char[]> pSshLocal;
uint32_t sshLocalSize = 0u;
char *crossThreadData = nullptr;