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
2025-08-20 16:28:06 +02:00
committed by GitHub
parent 5840940caa
commit f7338717ae
1574 changed files with 162972 additions and 199346 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/clr/opencl/tests/ocltst/module/runtime/OCLLDS32K.cpp
+75 -102
View File
@@ -53,24 +53,23 @@ typedef unsigned int uint32_t;
// 64 threads each handle 8192/64=128 values
#if EMU_ENV
static const char program_source[] = KERNEL(
__kernel void the_kernel(__global const uint *a, __global const uint *b,
__global const uint *c, __global uint *d,
__global uint *e) {
// Reduce size for the emulator
__local uint lds[256];
uint gid = get_global_id(0);
__global const uint* ta = a + 4 * gid;
__global const uint* tb = b + 4 * gid;
__global const uint* tc = c + 4 * gid;
__global uint* td = d + 4 * gid;
uint i;
__kernel void the_kernel(__global const uint* a, __global const uint* b, __global const uint* c,
__global uint* d, __global uint* e) {
// Reduce size for the emulator
__local uint lds[256];
uint gid = get_global_id(0);
__global const uint* ta = a + 4 * gid;
__global const uint* tb = b + 4 * gid;
__global const uint* tc = c + 4 * gid;
__global uint* td = d + 4 * gid;
uint i;
for (i = 0; i < 4; ++i) lds[ta[i]] = tc[i];
for (i = 0; i < 4; ++i) lds[ta[i]] = tc[i];
barrier(CLK_LOCAL_MEM_FENCE);
barrier(CLK_LOCAL_MEM_FENCE);
for (i = 0; i < 4; ++i) td[i] = lds[tb[i]];
} __kernel void the_kernel2(__global uint* d) {
for (i = 0; i < 4; ++i) td[i] = lds[tb[i]];
} __kernel void the_kernel2(__global uint* d) {
__local uint lds[8192];
uint i;
uint gid = get_global_id(0);
@@ -82,15 +81,14 @@ static const char program_source[] = KERNEL(
});
#else
static const char program_source[] = KERNEL(
__kernel void the_kernel(__global const uint *a, __global const uint *b,
__global const uint *c, __global uint *d,
__global uint *e) {
__kernel void the_kernel(__global const uint* a, __global const uint* b, __global const uint* c,
__global uint* d, __global uint* e) {
__local uint lds[8192];
uint gid = get_global_id(0);
__global const uint *ta = a + 128 * gid;
__global const uint *tb = b + 128 * gid;
__global const uint *tc = c + 128 * gid;
__global uint *td = d + 128 * gid;
__global const uint* ta = a + 128 * gid;
__global const uint* tb = b + 128 * gid;
__global const uint* tc = c + 128 * gid;
__global uint* td = d + 128 * gid;
uint i;
for (i = 0; i < 128; ++i) lds[ta[i]] = tc[i];
@@ -98,7 +96,7 @@ static const char program_source[] = KERNEL(
barrier(CLK_LOCAL_MEM_FENCE);
for (i = 0; i < 128; ++i) td[i] = lds[tb[i]];
} __kernel void the_kernel2(__global uint *d) {
} __kernel void the_kernel2(__global uint* d) {
__local uint lds[8192];
uint i;
uint gid = get_global_id(0);
@@ -110,8 +108,7 @@ static const char program_source[] = KERNEL(
});
#endif // EMU_ENV
static void fill(uint32_t *a, uint32_t *b, uint32_t *c, uint32_t *d,
uint32_t *e) {
static void fill(uint32_t* a, uint32_t* b, uint32_t* c, uint32_t* d, uint32_t* e) {
uint32_t i, j, k, t;
static uint32_t p[LDS_SIZE / 4];
static int is_set = 0;
@@ -149,8 +146,8 @@ static void fill(uint32_t *a, uint32_t *b, uint32_t *c, uint32_t *d,
}
}
static int check(const uint32_t *a, const uint32_t *b, const uint32_t *c,
const uint32_t *d, const uint32_t *e) {
static int check(const uint32_t* a, const uint32_t* b, const uint32_t* c, const uint32_t* d,
const uint32_t* e) {
uint32_t i, j, t;
uint32_t lds[LDS_SIZE / 4];
@@ -162,8 +159,8 @@ static int check(const uint32_t *a, const uint32_t *b, const uint32_t *c,
for (i = 0; i < LDS_SIZE / 4; ++i) {
t = lds[b[i]];
if (d[i] != t) {
printf("mismatch group %u thread %u element %u: %u instead of %u\n", j,
i / 128, i % 128, d[i], t);
printf("mismatch group %u thread %u element %u: %u instead of %u\n", j, i / 128, i % 128,
d[i], t);
return EXIT_FAILURE;
}
}
@@ -180,12 +177,11 @@ static int check(const uint32_t *a, const uint32_t *b, const uint32_t *c,
#define E_SIZE 32
#endif
void OCLLDS32K::setup_run(const char *cmplr_opt) {
void OCLLDS32K::setup_run(const char* cmplr_opt) {
cl_ulong lsize;
const char *ps[2];
error_ =
_wrapper->clGetDeviceInfo(devices_[_deviceId], CL_DEVICE_LOCAL_MEM_SIZE,
sizeof(lsize), &lsize, NULL);
const char* ps[2];
error_ = _wrapper->clGetDeviceInfo(devices_[_deviceId], CL_DEVICE_LOCAL_MEM_SIZE, sizeof(lsize),
&lsize, NULL);
if (lsize < LDS_SIZE) {
fprintf(stderr, "Passed! Test does not support 32kb of lds space!");
return;
@@ -193,25 +189,21 @@ void OCLLDS32K::setup_run(const char *cmplr_opt) {
// create the program
ps[0] = program_source;
program_ =
_wrapper->clCreateProgramWithSource(context_, 1, ps, NULL, &error_);
program_ = _wrapper->clCreateProgramWithSource(context_, 1, ps, NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "clCreateProgramWithSource failed");
// build the program
error_ = _wrapper->clBuildProgram(program_, 1, &devices_[_deviceId],
cmplr_opt, NULL, NULL);
error_ = _wrapper->clBuildProgram(program_, 1, &devices_[_deviceId], cmplr_opt, NULL, NULL);
if (error_ != CL_SUCCESS) {
char build_log[16384];
size_t log_sz;
fprintf(stderr, "build program failed, err=%d\n", error_);
error_ = _wrapper->clGetProgramBuildInfo(
program_, devices_[_deviceId], CL_PROGRAM_BUILD_LOG, sizeof(build_log),
build_log, &log_sz);
error_ = _wrapper->clGetProgramBuildInfo(program_, devices_[_deviceId], CL_PROGRAM_BUILD_LOG,
sizeof(build_log), build_log, &log_sz);
if (error_ != CL_SUCCESS)
fprintf(stderr, "failed to get build log, err=%d\n", error_);
else
fprintf(stderr, "----- Build Log -----\n%s\n----- ----- --- -----\n",
build_log);
fprintf(stderr, "----- Build Log -----\n%s\n----- ----- --- -----\n", build_log);
return;
}
@@ -224,46 +216,35 @@ void OCLLDS32K::setup_run(const char *cmplr_opt) {
CHECK_RESULT((error_ != CL_SUCCESS), "create a kernel failed");
// allocate the buffer memory objects
a_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_ONLY, A_SIZE, NULL,
&error_);
a_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_ONLY, A_SIZE, NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "create a buffer a failed");
buffers_.push_back(a_buf_);
b_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_ONLY, B_SIZE, NULL,
&error_);
b_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_ONLY, B_SIZE, NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "create a buffer b failed");
buffers_.push_back(b_buf_);
c_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_ONLY, C_SIZE, NULL,
&error_);
c_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_ONLY, C_SIZE, NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "create a buffer c failed");
buffers_.push_back(c_buf_);
d_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_WRITE, D_SIZE, NULL,
&error_);
d_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_WRITE, D_SIZE, NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "create a buffer d failed");
buffers_.push_back(d_buf_);
e_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_WRITE, E_SIZE, NULL,
&error_);
e_buf_ = _wrapper->clCreateBuffer(context_, CL_MEM_READ_WRITE, E_SIZE, NULL, &error_);
CHECK_RESULT((error_ != CL_SUCCESS), "create a buffer e failed");
buffers_.push_back(e_buf_);
// set the args values
error_ =
_wrapper->clSetKernelArg(kernel_, 0, sizeof(cl_mem), (void *)&a_buf_);
error_ |=
_wrapper->clSetKernelArg(kernel_, 1, sizeof(cl_mem), (void *)&b_buf_);
error_ |=
_wrapper->clSetKernelArg(kernel_, 2, sizeof(cl_mem), (void *)&c_buf_);
error_ |=
_wrapper->clSetKernelArg(kernel_, 3, sizeof(cl_mem), (void *)&d_buf_);
error_ |=
_wrapper->clSetKernelArg(kernel_, 4, sizeof(cl_mem), (void *)&e_buf_);
error_ = _wrapper->clSetKernelArg(kernel_, 0, sizeof(cl_mem), (void*)&a_buf_);
error_ |= _wrapper->clSetKernelArg(kernel_, 1, sizeof(cl_mem), (void*)&b_buf_);
error_ |= _wrapper->clSetKernelArg(kernel_, 2, sizeof(cl_mem), (void*)&c_buf_);
error_ |= _wrapper->clSetKernelArg(kernel_, 3, sizeof(cl_mem), (void*)&d_buf_);
error_ |= _wrapper->clSetKernelArg(kernel_, 4, sizeof(cl_mem), (void*)&e_buf_);
CHECK_RESULT((error_ != CL_SUCCESS), "setkernelArg failed!");
error_ =
_wrapper->clSetKernelArg(kernel2_, 0, sizeof(cl_mem), (void *)&d_buf_);
error_ = _wrapper->clSetKernelArg(kernel2_, 0, sizeof(cl_mem), (void*)&d_buf_);
CHECK_RESULT((error_ != CL_SUCCESS), "setkernelArg failed!");
}
@@ -273,22 +254,21 @@ void OCLLDS32K::cleanup_run() {
}
}
void OCLLDS32K::exec_kernel(void *a_mem, void *b_mem, void *c_mem, void *d_mem,
void *e_mem) {
void OCLLDS32K::exec_kernel(void* a_mem, void* b_mem, void* c_mem, void* d_mem, void* e_mem) {
size_t global_work_size[1];
size_t local_work_size[1];
// Send data to device
error_ = _wrapper->clEnqueueWriteBuffer(
cmdQueues_[_deviceId], a_buf_, CL_TRUE, 0, A_SIZE, a_mem, 0, NULL, NULL);
error_ = _wrapper->clEnqueueWriteBuffer(cmdQueues_[_deviceId], a_buf_, CL_TRUE, 0, A_SIZE, a_mem,
0, NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueWritebuffer failed");
error_ = _wrapper->clEnqueueWriteBuffer(
cmdQueues_[_deviceId], b_buf_, CL_TRUE, 0, B_SIZE, b_mem, 0, NULL, NULL);
error_ = _wrapper->clEnqueueWriteBuffer(cmdQueues_[_deviceId], b_buf_, CL_TRUE, 0, B_SIZE, b_mem,
0, NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueWritebuffer failed");
error_ = _wrapper->clEnqueueWriteBuffer(
cmdQueues_[_deviceId], c_buf_, CL_TRUE, 0, C_SIZE, c_mem, 0, NULL, NULL);
error_ = _wrapper->clEnqueueWriteBuffer(cmdQueues_[_deviceId], c_buf_, CL_TRUE, 0, C_SIZE, c_mem,
0, NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueWritebuffer failed");
// set work-item dimensions
@@ -296,60 +276,55 @@ void OCLLDS32K::exec_kernel(void *a_mem, void *b_mem, void *c_mem, void *d_mem,
local_work_size[0] = LOCAL_WORK_SIZE;
// execute kernel
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1,
NULL, global_work_size,
local_work_size, 0, NULL, NULL);
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1, NULL,
global_work_size, local_work_size, 0, NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueueNDRangeKernel failed");
// execute kernel
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1,
NULL, global_work_size,
local_work_size, 0, NULL, NULL);
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1, NULL,
global_work_size, local_work_size, 0, NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueueNDRangeKernel failed");
// execute kernel
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1,
NULL, global_work_size,
local_work_size, 0, NULL, NULL);
error_ = _wrapper->clEnqueueNDRangeKernel(cmdQueues_[_deviceId], kernel_, 1, NULL,
global_work_size, local_work_size, 0, NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueueNDRangeKernel failed");
// read results
error_ = _wrapper->clEnqueueReadBuffer(cmdQueues_[_deviceId], d_buf_, CL_TRUE,
0, D_SIZE, d_mem, 0, NULL, NULL);
error_ = _wrapper->clEnqueueReadBuffer(cmdQueues_[_deviceId], d_buf_, CL_TRUE, 0, D_SIZE, d_mem,
0, NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueueReadBuffer failed");
error_ = _wrapper->clEnqueueReadBuffer(cmdQueues_[_deviceId], e_buf_, CL_TRUE,
0, E_SIZE, e_mem, 0, NULL, NULL);
error_ = _wrapper->clEnqueueReadBuffer(cmdQueues_[_deviceId], e_buf_, CL_TRUE, 0, E_SIZE, e_mem,
0, NULL, NULL);
CHECK_RESULT((error_ != CL_SUCCESS), "clEnqueueReadBuffer failed");
error_ = _wrapper->clFinish(cmdQueues_[_deviceId]);
CHECK_RESULT((error_ != CL_SUCCESS), "clFinish failed");
}
const char *OCLLDS32K::kernel_src = "";
const char* OCLLDS32K::kernel_src = "";
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) {}
OCLLDS32K::OCLLDS32K() { _numSubTests = 1; }
OCLLDS32K::~OCLLDS32K() {}
void OCLLDS32K::open(unsigned int test, char *units, double &conversion,
unsigned int deviceId) {
void OCLLDS32K::open(unsigned int test, char* units, double& conversion, unsigned int deviceId) {
_deviceId = deviceId;
testID_ = test;
OCLTestImp::open(test, units, conversion, _deviceId);
}
void OCLLDS32K::run(void) {
void *a;
void *b;
void *c;
void *d;
void *e;
const char *cmplr_opt = NULL;
void* a;
void* b;
void* c;
void* d;
void* e;
const char* cmplr_opt = NULL;
int j, nj;
double f, dj, p;
@@ -383,14 +358,12 @@ void OCLLDS32K::run(void) {
// printf("Testing " TEST_NAME " on %s\n", argv[1]);
for (j = 0; j < nj; ++j) {
fill((uint32_t *)a, (uint32_t *)b, (uint32_t *)c, (uint32_t *)d,
(uint32_t *)e);
fill((uint32_t*)a, (uint32_t*)b, (uint32_t*)c, (uint32_t*)d, (uint32_t*)e);
// printf("%s Test %d: ", sDevice, j);
exec_kernel(a, b, c, d, e);
CHECK_RESULT((error_ != CL_SUCCESS), "exec_kernel failed!");
CHECK_RESULT((check((uint32_t *)a, (uint32_t *)b, (uint32_t *)c,
(uint32_t *)d, (uint32_t *)e) < 0),
CHECK_RESULT((check((uint32_t*)a, (uint32_t*)b, (uint32_t*)c, (uint32_t*)d, (uint32_t*)e) < 0),
" Failed!\n");
f = (j + 1) * dj;
if (nj > 1 && f >= p) {