AI/rocm-systems
2
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

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

Browse Source
This commit is contained in:
Danylo Lytovchenko authored and GitHub committed 2025-08-20 19:58:06 +05:30
1 parent 5840940caa
commit 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/OCLPerfMapBufferReadSpeed.cpp
+28 -48
View File
@@ -36,11 +36,9 @@
#define NUM_SIZES 4
// 256KB, 1 MB, 4MB, 16 MB
static const unsigned int Sizes[NUM_SIZES] = {262144, 1048576, 4194304,
16777216};
static const unsigned int Sizes[NUM_SIZES] = {262144, 1048576, 4194304, 16777216};
static const unsigned int Iterations[2] = {1,
OCLPerfMapBufferReadSpeed::NUM_ITER};
static const unsigned int Iterations[2] = {1, OCLPerfMapBufferReadSpeed::NUM_ITER};
#define NUM_OFFSETS 1
static const unsigned int offsets[NUM_OFFSETS] = {0};
#define NUM_SUBTESTS (3 + NUM_OFFSETS)
@@ -50,17 +48,15 @@ OCLPerfMapBufferReadSpeed::OCLPerfMapBufferReadSpeed() {
OCLPerfMapBufferReadSpeed::~OCLPerfMapBufferReadSpeed() {}
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 OCLPerfMapBufferReadSpeed::open(unsigned int test, char *units,
double &conversion,
void OCLPerfMapBufferReadSpeed::open(unsigned int test, char* units, double& conversion,
unsigned int deviceId) {
cl_uint numPlatforms;
cl_platform_id platform = NULL;
cl_uint num_devices = 0;
cl_device_id *devices = NULL;
cl_device_id* devices = NULL;
cl_device_id device = NULL;
_crcword = 0;
conversion = 1.0f;
@@ -81,7 +77,7 @@ void OCLPerfMapBufferReadSpeed::open(unsigned int test, char *units,
error_ = _wrapper->clGetPlatformIDs(0, NULL, &numPlatforms);
CHECK_RESULT(error_ != CL_SUCCESS, "clGetPlatformIDs failed");
if (0 < numPlatforms) {
cl_platform_id *platforms = new cl_platform_id[numPlatforms];
cl_platform_id* platforms = new cl_platform_id[numPlatforms];
error_ = _wrapper->clGetPlatformIDs(numPlatforms, platforms, NULL);
CHECK_RESULT(error_ != CL_SUCCESS, "clGetPlatformIDs failed");
#if 0
@@ -91,13 +87,11 @@ void OCLPerfMapBufferReadSpeed::open(unsigned int test, char *units,
#endif
platform = platforms[_platformIndex];
char pbuf[100];
error_ = _wrapper->clGetPlatformInfo(platforms[_platformIndex],
CL_PLATFORM_VENDOR, sizeof(pbuf), pbuf,
NULL);
error_ = _wrapper->clGetPlatformInfo(platforms[_platformIndex], CL_PLATFORM_VENDOR,
sizeof(pbuf), pbuf, NULL);
num_devices = 0;
/* Get the number of requested devices */
error_ = _wrapper->clGetDeviceIDs(platforms[_platformIndex], type_, 0, NULL,
&num_devices);
error_ = _wrapper->clGetDeviceIDs(platforms[_platformIndex], type_, 0, NULL, &num_devices);
// Runtime returns an error when no GPU devices are present instead of just
// returning 0 devices
// CHECK_RESULT(error_ != CL_SUCCESS, "clGetDeviceIDs failed");
@@ -132,19 +126,17 @@ void OCLPerfMapBufferReadSpeed::open(unsigned int test, char *units,
numIter = Iterations[_openTest / (NUM_SIZES * NUM_SUBTESTS)];
devices = (cl_device_id *)malloc(num_devices * sizeof(cl_device_id));
devices = (cl_device_id*)malloc(num_devices * sizeof(cl_device_id));
CHECK_RESULT(devices == 0, "no devices");
/* Get the requested device */
error_ =
_wrapper->clGetDeviceIDs(platform, type_, num_devices, devices, NULL);
error_ = _wrapper->clGetDeviceIDs(platform, type_, num_devices, devices, NULL);
CHECK_RESULT(error_ != CL_SUCCESS, "clGetDeviceIDs failed");
CHECK_RESULT(_deviceId >= num_devices, "Requested deviceID not available");
device = devices[_deviceId];
context_ = _wrapper->clCreateContext(NULL, 1, &device, notify_callback, NULL,
&error_);
context_ = _wrapper->clCreateContext(NULL, 1, &device, notify_callback, NULL, &error_);
CHECK_RESULT(context_ == 0, "clCreateContext failed");
cmd_queue_ = _wrapper->clCreateCommandQueue(context_, device, 0, NULL);
@@ -157,24 +149,19 @@ void OCLPerfMapBufferReadSpeed::open(unsigned int test, char *units,
flags |= CL_MEM_ALLOC_HOST_PTR;
} else if (useHostPtr) {
flags |= CL_MEM_USE_HOST_PTR;
hostMem = (char *)malloc(bufSize_ + alignment - 1 + offset);
hostMem = (char*)malloc(bufSize_ + alignment - 1 + offset);
CHECK_RESULT(hostMem == 0, "malloc(hostMem) failed");
alignedMem =
(char *)((((intptr_t)hostMem + alignment - 1) & ~(alignment - 1)) +
offset);
alignedMem = (char*)((((intptr_t)hostMem + alignment - 1) & ~(alignment - 1)) + offset);
}
outBuffer_ =
_wrapper->clCreateBuffer(context_, flags, bufSize_, alignedMem, &error_);
outBuffer_ = _wrapper->clCreateBuffer(context_, flags, bufSize_, alignedMem, &error_);
CHECK_RESULT(outBuffer_ == 0, "clCreateBuffer(outBuffer) failed");
// Force memory to be on GPU, if possible
{
cl_mem memBuffer =
_wrapper->clCreateBuffer(context_, 0, bufSize_, NULL, &error_);
cl_mem memBuffer = _wrapper->clCreateBuffer(context_, 0, bufSize_, NULL, &error_);
CHECK_RESULT(memBuffer == 0, "clCreateBuffer(memBuffer) failed");
_wrapper->clEnqueueCopyBuffer(cmd_queue_, memBuffer, outBuffer_, 0, 0,
bufSize_, 0, NULL, NULL);
_wrapper->clEnqueueCopyBuffer(cmd_queue_, memBuffer, outBuffer_, 0, 0, bufSize_, 0, NULL, NULL);
_wrapper->clFinish(cmd_queue_);
_wrapper->clReleaseMemObject(memBuffer);
@@ -184,15 +171,13 @@ void OCLPerfMapBufferReadSpeed::open(unsigned int test, char *units,
void OCLPerfMapBufferReadSpeed::run(void) {
CPerfCounter timer;
void *mem;
void* mem;
// Warm up
mem =
_wrapper->clEnqueueMapBuffer(cmd_queue_, outBuffer_, CL_TRUE, CL_MAP_READ,
0, bufSize_, 0, NULL, NULL, &error_);
mem = _wrapper->clEnqueueMapBuffer(cmd_queue_, outBuffer_, CL_TRUE, CL_MAP_READ, 0, bufSize_, 0,
NULL, NULL, &error_);
CHECK_RESULT(error_, "clEnqueueMapBuffer failed");
error_ = _wrapper->clEnqueueUnmapMemObject(cmd_queue_, outBuffer_, mem, 0,
NULL, NULL);
error_ = _wrapper->clEnqueueUnmapMemObject(cmd_queue_, outBuffer_, mem, 0, NULL, NULL);
CHECK_RESULT(error_, "clEnqueueUnmapBuffer failed");
error_ = _wrapper->clFinish(cmd_queue_);
CHECK_RESULT(error_, "clFinish failed");
@@ -200,13 +185,11 @@ void OCLPerfMapBufferReadSpeed::run(void) {
timer.Reset();
timer.Start();
for (unsigned int i = 0; i < numIter; i++) {
mem = _wrapper->clEnqueueMapBuffer(cmd_queue_, outBuffer_, CL_TRUE,
CL_MAP_READ, 0, bufSize_, 0, NULL, NULL,
&error_);
mem = _wrapper->clEnqueueMapBuffer(cmd_queue_, outBuffer_, CL_TRUE, CL_MAP_READ, 0, bufSize_, 0,
NULL, NULL, &error_);
CHECK_RESULT(error_, "clEnqueueMapBuffer failed");
error_ = _wrapper->clEnqueueUnmapMemObject(cmd_queue_, outBuffer_, mem, 0,
NULL, NULL);
error_ = _wrapper->clEnqueueUnmapMemObject(cmd_queue_, outBuffer_, mem, 0, NULL, NULL);
CHECK_RESULT(error_, "clEnqueueUnmapBuffer failed");
error_ = _wrapper->clFinish(cmd_queue_);
CHECK_RESULT(error_, "clFinish failed");
@@ -234,21 +217,18 @@ void OCLPerfMapBufferReadSpeed::run(void) {
SNPRINTF(str, sizeof(str), "(GB/s)");
}
char buf[256];
SNPRINTF(buf, sizeof(buf), " (%8d bytes) i: %4d %29s ", bufSize_, numIter,
str);
SNPRINTF(buf, sizeof(buf), " (%8d bytes) i: %4d %29s ", bufSize_, numIter, str);
testDescString = buf;
}
unsigned int OCLPerfMapBufferReadSpeed::close(void) {
if (outBuffer_) {
error_ = _wrapper->clReleaseMemObject(outBuffer_);
CHECK_RESULT_NO_RETURN(error_ != CL_SUCCESS,
"clReleaseMemObject(outBuffer_) failed");
CHECK_RESULT_NO_RETURN(error_ != CL_SUCCESS, "clReleaseMemObject(outBuffer_) failed");
}
if (cmd_queue_) {
error_ = _wrapper->clReleaseCommandQueue(cmd_queue_);
CHECK_RESULT_NO_RETURN(error_ != CL_SUCCESS,
"clReleaseCommandQueue failed");
CHECK_RESULT_NO_RETURN(error_ != CL_SUCCESS, "clReleaseCommandQueue failed");
}
if (context_) {
error_ = _wrapper->clReleaseContext(context_);