AI/rocm-systems
2
0 Garpu 0
Kode Masalah Tarik Permintaan Actions Packages Projects Rilis Wiki Kegiatan

SWDEV-470698 - fix formatting, add format check workflow (#657)

Telusuri Sumber
This commit is contained in:
Danylo Lytovchenko authored and GitHub committed 2025-08-20 19:58:06 +05:30
1 orang tua 5840940caa
melakukan f7338717ae
1574 files changed
+162903 -199277
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

projects/clr/opencl/tests/ocltst/module/perf/OCLPerfDevMemReadSpeed.cpp
+40 -54
Melihat File
@@ -37,7 +37,7 @@
#define NUM_SIZES 1
static const unsigned int Sizes[NUM_SIZES] = {256 * 1024 * 1024};
const static char *strKernel =
const static char* strKernel =
"__kernel void read_kernel(__global uint16 *src, ulong size1, uint "
"threads, __global uint* dst\n"
" )\n"
@@ -61,12 +61,11 @@ OCLPerfDevMemReadSpeed::OCLPerfDevMemReadSpeed() { _numSubTests = 1; }
OCLPerfDevMemReadSpeed::~OCLPerfDevMemReadSpeed() {}
static void CL_CALLBACK notify_callback(const char *errinfo,
const void *private_info, size_t cb,
void *user_data) {}
static void CL_CALLBACK notify_callback(const char* errinfo, const void* private_info, size_t cb,
void* user_data) {}
void OCLPerfDevMemReadSpeed::open(unsigned int test, char *units,
double &conversion, unsigned int deviceId) {
void OCLPerfDevMemReadSpeed::open(unsigned int test, char* units, double& conversion,
unsigned int deviceId) {
error_ = CL_SUCCESS;
OCLTestImp::open(test, units, conversion, deviceId);
CHECK_RESULT((error_ != CL_SUCCESS), "Error opening test");
@@ -78,8 +77,7 @@ void OCLPerfDevMemReadSpeed::open(unsigned int test, char *units,
nBytes = Sizes[0];
cl_ulong loopCnt = nBytes / (16 * sizeof(cl_uint));
cl_uint maxCUs;
error_ = _wrapper->clGetDeviceInfo(devices_[_deviceId],
CL_DEVICE_MAX_COMPUTE_UNITS,
error_ = _wrapper->clGetDeviceInfo(devices_[_deviceId], CL_DEVICE_MAX_COMPUTE_UNITS,
sizeof(cl_uint), &maxCUs, 0);
CHECK_RESULT(error_ != CL_SUCCESS, "clGetDeviceInfo failed");
wgs = 64;
@@ -89,16 +87,14 @@ void OCLPerfDevMemReadSpeed::open(unsigned int test, char *units,
inputData = 0x1;
nIter = 1000;
program_ = _wrapper->clCreateProgramWithSource(context_, 1, &strKernel, NULL,
&error_);
program_ = _wrapper->clCreateProgramWithSource(context_, 1, &strKernel, NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "clCreateProgramWithSource() failed");
error_ = _wrapper->clBuildProgram(program_, 1, &devices_[deviceId], NULL,
NULL, NULL);
error_ = _wrapper->clBuildProgram(program_, 1, &devices_[deviceId], NULL, NULL, NULL);
if (error_ != CL_SUCCESS) {
char programLog[1024];
_wrapper->clGetProgramBuildInfo(program_, devices_[deviceId],
CL_PROGRAM_BUILD_LOG, 1024, programLog, 0);
_wrapper->clGetProgramBuildInfo(program_, devices_[deviceId], CL_PROGRAM_BUILD_LOG, 1024,
programLog, 0);
printf("\n%s\n", programLog);
fflush(stdout);
}
@@ -107,44 +103,37 @@ void OCLPerfDevMemReadSpeed::open(unsigned int test, char *units,
kernel_ = _wrapper->clCreateKernel(program_, "read_kernel", &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "clCreateKernel() failed");
srcBuffer_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_ONLY, nBytes,
NULL, &error_);
srcBuffer_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_ONLY, nBytes, NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "clCreateBuffer(srcBuffer) failed");
void *mem;
void* mem;
mem = _wrapper->clEnqueueMapBuffer(cmdQueues_[_deviceId], srcBuffer_, CL_TRUE,
CL_MAP_READ | CL_MAP_WRITE, 0, nBytes, 0,
NULL, NULL, &error_);
CL_MAP_READ | CL_MAP_WRITE, 0, nBytes, 0, NULL, NULL, &error_);
CHECK_RESULT(error_, "clEnqueueMapBuffer failed");
for (unsigned int i = 0; i < nBytes / sizeof(cl_uint); ++i) {
reinterpret_cast<cl_uint *>(mem)[i] = inputData;
reinterpret_cast<cl_uint*>(mem)[i] = inputData;
}
dstBuffer_ = _wrapper->clCreateBuffer(context_, CL_MEM_WRITE_ONLY,
sizeof(cl_uint), NULL, &error_);
dstBuffer_ =
_wrapper->clCreateBuffer(context_, CL_MEM_WRITE_ONLY, sizeof(cl_uint), NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "clCreateBuffer(dstBuffer) failed");
_wrapper->clEnqueueUnmapMemObject(cmdQueues_[_deviceId], srcBuffer_, mem, 0,
NULL, NULL);
_wrapper->clEnqueueUnmapMemObject(cmdQueues_[_deviceId], srcBuffer_, mem, 0, NULL, NULL);
mem = _wrapper->clEnqueueMapBuffer(cmdQueues_[_deviceId], dstBuffer_, CL_TRUE,
CL_MAP_READ | CL_MAP_WRITE, 0,
sizeof(cl_uint), 0, NULL, NULL, &error_);
CL_MAP_READ | CL_MAP_WRITE, 0, sizeof(cl_uint), 0, NULL, NULL,
&error_);
CHECK_RESULT(error_, "clEnqueueMapBuffer failed");
memset(mem, 0, sizeof(cl_uint));
_wrapper->clEnqueueUnmapMemObject(cmdQueues_[_deviceId], dstBuffer_, mem, 0,
NULL, NULL);
_wrapper->clEnqueueUnmapMemObject(cmdQueues_[_deviceId], dstBuffer_, mem, 0, NULL, NULL);
error_ = _wrapper->clSetKernelArg(kernel_, 0, sizeof(cl_mem), &srcBuffer_);
CHECK_RESULT((error_ != CL_SUCCESS), "clSetKernelArg() failed");
error_ =
_wrapper->clSetKernelArg(kernel_, 1, sizeof(cl_ulong), (void *)&loopCnt);
error_ = _wrapper->clSetKernelArg(kernel_, 1, sizeof(cl_ulong), (void*)&loopCnt);
CHECK_RESULT((error_ != CL_SUCCESS), "clSetKernelArg() failed");
error_ = _wrapper->clSetKernelArg(kernel_, 2, sizeof(cl_uint),
(void *)&nWorkItems);
error_ = _wrapper->clSetKernelArg(kernel_, 2, sizeof(cl_uint), (void*)&nWorkItems);
CHECK_RESULT((error_ != CL_SUCCESS), "clSetKernelArg() failed");
error_ =
_wrapper->clSetKernelArg(kernel_, 3, sizeof(cl_mem), (void *)&dstBuffer_);
error_ = _wrapper->clSetKernelArg(kernel_, 3, sizeof(cl_mem), (void*)&dstBuffer_);
CHECK_RESULT((error_ != CL_SUCCESS), "clSetKernelArg() failed");
}
@@ -159,22 +148,21 @@ void OCLPerfDevMemReadSpeed::run(void) {
size_t lws[1] = {wgs};
// warm up
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1,
NULL, gws, lws, 0, NULL, NULL);
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1, NULL, gws, lws, 0,
NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueueNDRangeKernel() failed");
_wrapper->clFinish(cmdQueues_[_deviceId]);
cl_uint *memResult;
memResult = (cl_uint *)malloc(sizeof(cl_uint));
cl_uint* memResult;
memResult = (cl_uint*)malloc(sizeof(cl_uint));
if (0 == memResult) {
CHECK_RESULT_NO_RETURN(0, "malloc failed!\n");
return;
}
memset(memResult, 0, sizeof(cl_uint));
error_ = _wrapper->clEnqueueReadBuffer(cmdQueues_[_deviceId], dstBuffer_,
CL_FALSE, 0, sizeof(cl_uint),
memResult, 0, NULL, NULL);
error_ = _wrapper->clEnqueueReadBuffer(cmdQueues_[_deviceId], dstBuffer_, CL_FALSE, 0,
sizeof(cl_uint), memResult, 0, NULL, NULL);
CHECK_RESULT(error_, "clEnqueueReadBuffer dstBuffer_ failed!");
_wrapper->clFinish(cmdQueues_[_deviceId]);
@@ -190,21 +178,21 @@ void OCLPerfDevMemReadSpeed::run(void) {
timer.Reset();
timer.Start();
double sec2 = 0;
cl_event *events = new cl_event[nIter];
cl_event* events = new cl_event[nIter];
for (unsigned int i = 0; i < nIter; i++) {
error_ = _wrapper->clEnqueueNDRangeKernel(
cmdQueues_[_deviceId], kernel_, 1, NULL, gws, lws, 0, NULL, &events[i]);
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1, NULL, gws, lws, 0,
NULL, &events[i]);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueueNDRangeKernel() failed");
}
_wrapper->clFinish(cmdQueues_[_deviceId]);
timer.Stop();
for (unsigned int i = 0; i < nIter; i++) {
cl_ulong startTime = 0, endTime = 0;
error_ = _wrapper->clGetEventProfilingInfo(
events[i], CL_PROFILING_COMMAND_START, sizeof(cl_ulong), &startTime, 0);
error_ = _wrapper->clGetEventProfilingInfo(events[i], CL_PROFILING_COMMAND_START,
sizeof(cl_ulong), &startTime, 0);
CHECK_RESULT(error_, "clGetEventProfilingInfo failed");
error_ = _wrapper->clGetEventProfilingInfo(
events[i], CL_PROFILING_COMMAND_END, sizeof(cl_ulong), &endTime, 0);
error_ = _wrapper->clGetEventProfilingInfo(events[i], CL_PROFILING_COMMAND_END,
sizeof(cl_ulong), &endTime, 0);
CHECK_RESULT(error_, "clGetEventProfilingInfo failed");
_wrapper->clReleaseEvent(events[i]);
@@ -219,8 +207,8 @@ void OCLPerfDevMemReadSpeed::run(void) {
_perfInfo = (float)perf2;
float perfInfo = (float)perf;
char buf[256];
SNPRINTF(buf, sizeof(buf), " (%8d bytes) i:%4d Wall time Perf: %.2f (GB/s)",
nBytes, nIter, perfInfo);
SNPRINTF(buf, sizeof(buf), " (%8d bytes) i:%4d Wall time Perf: %.2f (GB/s)", nBytes, nIter,
perfInfo);
testDescString = buf;
}
@@ -228,14 +216,12 @@ unsigned int OCLPerfDevMemReadSpeed::close(void) {
if (!skip_) {
if (srcBuffer_) {
error_ = _wrapper->clReleaseMemObject(srcBuffer_);
CHECK_RESULT_NO_RETURN(error_ != CL_SUCCESS,
"clReleaseMemObject(srcBuffer_) failed");
CHECK_RESULT_NO_RETURN(error_ != CL_SUCCESS, "clReleaseMemObject(srcBuffer_) failed");
}
if (dstBuffer_) {
error_ = _wrapper->clReleaseMemObject(dstBuffer_);
CHECK_RESULT_NO_RETURN(error_ != CL_SUCCESS,
"clReleaseMemObject(srcBuffer_) failed");
CHECK_RESULT_NO_RETURN(error_ != CL_SUCCESS, "clReleaseMemObject(srcBuffer_) failed");
}
}