Media/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-11-10 05:49:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

Code cleanup - avoid copy 5/n

Browse Source 
Signed-off-by: Kamil Kopryk <kamil.kopryk@intel.com>
This commit is contained in:
Kamil Kopryk
2021-09-07 21:21:19 +00:00
committed by Compute-Runtime-Automation
parent ae88789bce
commit a924b6a304
22 changed files with 81 additions and 80 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
opencl/source/built_ins/vme_dispatch_builder.h
View File

@@ -174,7 +174,7 @@ class VmeBuiltinDispatchInfoBuilder : public BuiltinDispatchInfoBuilder {
return *(RetType *)(vmeKernel->getCrossThreadData() + element.offset);
}
cl_int validateImages(Vec3<size_t> inputRegion, Vec3<size_t> offset) const {
cl_int validateImages(const Vec3<size_t> &inputRegion, const Vec3<size_t> &offset) const {
Image *srcImg = castToObject<Image>((cl_mem)vmeKernel->getKernelArg(srcImgArgNum));
Image *refImg = castToObject<Image>((cl_mem)vmeKernel->getKernelArg(refImgArgNum));
@@ -208,7 +208,7 @@ class VmeBuiltinDispatchInfoBuilder : public BuiltinDispatchInfoBuilder {
return CL_SUCCESS;
}
virtual cl_int validateVmeDispatch(Vec3<size_t> inputRegion, Vec3<size_t> offset, size_t blkNum, size_t blkMul) const {
virtual cl_int validateVmeDispatch(const Vec3<size_t> &inputRegion, const Vec3<size_t> &offset, size_t blkNum, size_t blkMul) const {
{
cl_int imageValidationStatus = validateImages(inputRegion, offset);
if (imageValidationStatus != CL_SUCCESS) {
@@ -383,7 +383,7 @@ class AdvancedVmeBuiltinDispatchInfoBuilder : public VmeBuiltinDispatchInfoBuild
return predictorsBufferExpSize;
}
cl_int validateVmeDispatch(Vec3<size_t> inputRegion, Vec3<size_t> offset, size_t blkNum, size_t blkMul) const override {
cl_int validateVmeDispatch(const Vec3<size_t> &inputRegion, const Vec3<size_t> &offset, size_t blkNum, size_t blkMul) const override {
cl_int basicVmeValidationStatus = VmeBuiltinDispatchInfoBuilder::validateVmeDispatch(inputRegion, offset, blkNum, blkMul);
if (basicVmeValidationStatus != CL_SUCCESS) {
return basicVmeValidationStatus;
@@ -452,7 +452,7 @@ class BuiltInOp<EBuiltInOps::VmeBlockAdvancedMotionEstimateCheckIntel> : public
"block_advanced_motion_estimate_check_intel") {
}
cl_int validateVmeDispatch(Vec3<size_t> inputRegion, Vec3<size_t> offset,
cl_int validateVmeDispatch(const Vec3<size_t> &inputRegion, const Vec3<size_t> &offset,
size_t gwWidthInBlk, size_t gwHeightInBlk) const override {
cl_int basicAdvVmeValidationStatus = AdvancedVmeBuiltinDispatchInfoBuilder::validateVmeDispatch(inputRegion, offset, gwWidthInBlk, gwHeightInBlk);
if (basicAdvVmeValidationStatus != CL_SUCCESS) {

12
opencl/source/command_queue/gpgpu_walker.h
View File

@@ -68,22 +68,22 @@ Vec3<size_t> generateWorkgroupSize(
const DispatchInfo &dispatchInfo);
Vec3<size_t> computeWorkgroupsNumber(
const Vec3<size_t> gws,
const Vec3<size_t> lws);
const Vec3<size_t> &gws,
const Vec3<size_t> &lws);
Vec3<size_t> generateWorkgroupsNumber(
const Vec3<size_t> gws,
const Vec3<size_t> lws);
const Vec3<size_t> &gws,
const Vec3<size_t> &lws);
Vec3<size_t> generateWorkgroupsNumber(
const DispatchInfo &dispatchInfo);
inline uint32_t calculateDispatchDim(Vec3<size_t> dispatchSize, Vec3<size_t> dispatchOffset) {
inline uint32_t calculateDispatchDim(const Vec3<size_t> &dispatchSize, const Vec3<size_t> &dispatchOffset) {
return std::max(1U, std::max(dispatchSize.getSimplifiedDim(), dispatchOffset.getSimplifiedDim()));
}
Vec3<size_t> canonizeWorkgroup(
Vec3<size_t> workgroup);
const Vec3<size_t> &workgroup);
void provideLocalWorkGroupSizeHints(Context *context, DispatchInfo dispatchInfo);

4
opencl/source/command_queue/hardware_interface.h
View File

@@ -94,8 +94,8 @@ class HardwareInterface {
uint32_t &interfaceDescriptorIndex,
const DispatchInfo &dispatchInfo,
size_t offsetInterfaceDescriptorTable,
Vec3<size_t> &numberOfWorkgroups,
Vec3<size_t> &startOfWorkgroups);
const Vec3<size_t> &numberOfWorkgroups,
const Vec3<size_t> &startOfWorkgroups);
static WALKER_TYPE<GfxFamily> *allocateWalkerSpace(LinearStream &commandStream,
const Kernel &kernel);

16
opencl/source/command_queue/hardware_interface_base.inl
View File

@@ -193,18 +193,18 @@ void HardwareInterface<GfxFamily>::dispatchKernelCommands(CommandQueue &commandQ
}
//Get dispatch geometry
uint32_t dim = dispatchInfo.getDim();
Vec3<size_t> gws = dispatchInfo.getGWS();
Vec3<size_t> offset = dispatchInfo.getOffset();
Vec3<size_t> startOfWorkgroups = dispatchInfo.getStartOfWorkgroups();
auto dim = dispatchInfo.getDim();
const auto &gws = dispatchInfo.getGWS();
const auto &offset = dispatchInfo.getOffset();
const auto &startOfWorkgroups = dispatchInfo.getStartOfWorkgroups();
// Compute local workgroup sizes
Vec3<size_t> lws = dispatchInfo.getLocalWorkgroupSize();
Vec3<size_t> elws = (dispatchInfo.getEnqueuedWorkgroupSize().x > 0) ? dispatchInfo.getEnqueuedWorkgroupSize() : lws;
const auto &lws = dispatchInfo.getLocalWorkgroupSize();
const auto &elws = (dispatchInfo.getEnqueuedWorkgroupSize().x > 0) ? dispatchInfo.getEnqueuedWorkgroupSize() : lws;
// Compute number of work groups
Vec3<size_t> totalNumberOfWorkgroups = dispatchInfo.getTotalNumberOfWorkgroups();
Vec3<size_t> numberOfWorkgroups = dispatchInfo.getNumberOfWorkgroups();
const auto &totalNumberOfWorkgroups = dispatchInfo.getTotalNumberOfWorkgroups();
const auto &numberOfWorkgroups = dispatchInfo.getNumberOfWorkgroups();
UNRECOVERABLE_IF(totalNumberOfWorkgroups.x == 0);
UNRECOVERABLE_IF(numberOfWorkgroups.x == 0);

4
opencl/source/command_queue/hardware_interface_bdw_and_later.inl
View File

@@ -67,8 +67,8 @@ inline void HardwareInterface<GfxFamily>::programWalker(
uint32_t &interfaceDescriptorIndex,
const DispatchInfo &dispatchInfo,
size_t offsetInterfaceDescriptorTable,
Vec3<size_t> &numberOfWorkgroups,
Vec3<size_t> &startOfWorkgroups) {
const Vec3<size_t> &numberOfWorkgroups,
const Vec3<size_t> &startOfWorkgroups) {
auto walkerCmdBuf = allocateWalkerSpace(commandStream, kernel);
WALKER_TYPE<GfxFamily> walkerCmd = GfxFamily::cmdInitGpgpuWalker;

4
opencl/source/command_queue/hardware_interface_xehp_and_later.inl
View File

@@ -57,8 +57,8 @@ inline void HardwareInterface<GfxFamily>::programWalker(
uint32_t &interfaceDescriptorIndex,
const DispatchInfo &dispatchInfo,
size_t offsetInterfaceDescriptorTable,
Vec3<size_t> &numberOfWorkgroups,
Vec3<size_t> &startOfWorkgroups) {
const Vec3<size_t> &numberOfWorkgroups,
const Vec3<size_t> &startOfWorkgroups) {
using COMPUTE_WALKER = typename GfxFamily::COMPUTE_WALKER;

6
opencl/source/command_queue/local_work_size.cpp
View File

@@ -452,13 +452,13 @@ Vec3<size_t> generateWorkgroupSize(const DispatchInfo &dispatchInfo) {
return (dispatchInfo.getEnqueuedWorkgroupSize().x == 0) ? computeWorkgroupSize(dispatchInfo) : dispatchInfo.getEnqueuedWorkgroupSize();
}
Vec3<size_t> computeWorkgroupsNumber(const Vec3<size_t> gws, const Vec3<size_t> lws) {
Vec3<size_t> computeWorkgroupsNumber(const Vec3<size_t> &gws, const Vec3<size_t> &lws) {
return (Vec3<size_t>(gws.x / lws.x + ((gws.x % lws.x) ? 1 : 0),
gws.y / lws.y + ((gws.y % lws.y) ? 1 : 0),
gws.z / lws.z + ((gws.z % lws.z) ? 1 : 0)));
}
Vec3<size_t> generateWorkgroupsNumber(const Vec3<size_t> gws, const Vec3<size_t> lws) {
Vec3<size_t> generateWorkgroupsNumber(const Vec3<size_t> &gws, const Vec3<size_t> &lws) {
return (lws.x > 0) ? computeWorkgroupsNumber(gws, lws) : Vec3<size_t>(0, 0, 0);
}
@@ -466,7 +466,7 @@ Vec3<size_t> generateWorkgroupsNumber(const DispatchInfo &dispatchInfo) {
return generateWorkgroupsNumber(dispatchInfo.getGWS(), dispatchInfo.getLocalWorkgroupSize());
}
Vec3<size_t> canonizeWorkgroup(Vec3<size_t> workgroup) {
Vec3<size_t> canonizeWorkgroup(const Vec3<size_t> &workgroup) {
return ((workgroup.x > 0) ? Vec3<size_t>({workgroup.x, std::max(workgroup.y, static_cast<size_t>(1)), std::max(workgroup.z, static_cast<size_t>(1))})
: Vec3<size_t>(0, 0, 0));
}

4
opencl/source/helpers/dispatch_info.h
View File

@@ -32,9 +32,9 @@ class DispatchInfo {
using EstimateCommandsMethodT = size_t(size_t, const HardwareInfo &, bool);
DispatchInfo() = default;
DispatchInfo(ClDevice *device, Kernel *kernel, uint32_t dim, Vec3<size_t> gws, Vec3<size_t> elws, Vec3<size_t> offset)
DispatchInfo(ClDevice *device, Kernel *kernel, uint32_t dim, const Vec3<size_t> &gws, const Vec3<size_t> &elws, const Vec3<size_t> &offset)
: pClDevice(device), kernel(kernel), dim(dim), gws(gws), elws(elws), offset(offset) {}
DispatchInfo(ClDevice *device, Kernel *kernel, uint32_t dim, Vec3<size_t> gws, Vec3<size_t> elws, Vec3<size_t> offset, Vec3<size_t> agws, Vec3<size_t> lws, Vec3<size_t> twgs, Vec3<size_t> nwgs, Vec3<size_t> swgs)
DispatchInfo(ClDevice *device, Kernel *kernel, uint32_t dim, const Vec3<size_t> &gws, const Vec3<size_t> &elws, const Vec3<size_t> &offset, const Vec3<size_t> &agws, const Vec3<size_t> &lws, const Vec3<size_t> &twgs, const Vec3<size_t> &nwgs, const Vec3<size_t> &swgs)
: pClDevice(device), kernel(kernel), dim(dim), gws(gws), elws(elws), offset(offset), agws(agws), lws(lws), twgs(twgs), nwgs(nwgs), swgs(swgs) {}
ClDevice &getClDevice() const { return *pClDevice; }