int openclInit()
{
cl_int ret;
openclRetTackle( clGetPlatformIDs(1, &cpPlatform, NULL), "clGetPlatFormIDs");
openclRetTackle( clGetDeviceIDs(cpPlatform, CL_DEVICE_TYPE_CPU,1,&cdDevice,NULL), "clGetDeviceIDs");
cxGPUContext = clCreateContext(0, 1, &cdDevice, NULL, NULL, &ret);
#if (defined CL_DEBUG) || (defined CL_VERBOSE)
printCLDeviceInfo();
#endif
openclRetTackle( ret , "clCreateContext" );
cqCommandQueue = clCreateCommandQueue(cxGPUContext, cdDevice, 0, &ret);
openclRetTackle( ret , "clCreateCommandQueue");
return CL_SUCCESS;
}
int main() {
std::ios::sync_with_stdio();
cl_int err = 0;
cl_uint num_devices;
cl_uint num_plats;
cl_platform_id *plat;
cl_device_id final_device;
err = clGetPlatformIDs(0, NULL, &num_plats);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
if (num_plats > 1) printf("There are %u platforms\n", num_plats);
else if (num_plats == 1) printf("There's exactly one platform, namely:\n");
else printf("There are no available platforms"), exit(0);
plat = (cl_platform_id*) malloc(sizeof(cl_platform_id) * num_plats);
err = clGetPlatformIDs(num_plats, plat, NULL);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
for (int i = 0; i < num_plats; i++) {
char plat_info[1024];
err = clGetPlatformInfo(plat[i], CL_PLATFORM_VERSION, 1024, plat_info, NULL);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
printf("Platform %d:\n", i);
printf(" * %s\n", plat_info);
err = clGetDeviceIDs(plat[i], CL_DEVICE_TYPE_GPU, 0, NULL, &num_devices);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
printf("There are %u devices found, namely:\n", num_devices);
cl_device_id devices[num_devices];
err = clGetDeviceIDs(plat[i], CL_DEVICE_TYPE_ALL, num_devices, devices, NULL);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
for (int j = 0; j < num_devices; j++) {
cl_device_type current_type;
clGetDeviceInfo(devices[j], CL_DEVICE_TYPE, sizeof(cl_device_type), ¤t_type, NULL);
if (current_type == CL_DEVICE_TYPE_GPU) {
printf("Found GPU\n");
final_device = devices[j];
}
printCLDeviceInfo(devices[j],false);
if (j != num_devices - 1) putchar('\n');
}
}
printf("\n");
size_t global = 8092;
unsigned int count = global * N;
// Fill the array
float* input_data = (float*)malloc(count * sizeof(float));
srand(time(NULL));
for (int i = 0; i < count; i++) input_data[i] = rand() / (float)RAND_MAX;
// Load the kernel
const char* program_source = readFile("kernel/square.cl");
printf("Loaded kernel program source:\n%s\n", program_source);
// Set OpenCL context
cl_context context = clCreateContext(0, 1, &final_device, NULL, NULL, &err);
if (err != CL_SUCCESS) printf("ERROR at line %u: Failed to make OpenCL context\n", __LINE__);
// Create command queue
cl_command_queue commands = clCreateCommandQueue(context, final_device, 0, &err);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
// Create the program from source and build it.
cl_program program = clCreateProgramWithSource(context, 1, (const char **) &program_source, NULL, &err);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
err = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
if (err != CL_SUCCESS) {
size_t len;
char buffer[2048];
printf("Error: Failed to build program executable!\n");
clGetProgramBuildInfo(program, final_device, CL_PROGRAM_BUILD_LOG, sizeof(buffer), buffer, &len);
printf("%s\n", buffer);
}
cl_kernel kernel = clCreateKernel(program, "square", &err);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
// Create the buffers for input and output arrays.
cl_mem read_buffer = clCreateBuffer(context, CL_MEM_READ_ONLY, sizeof(float)*count, NULL, &err);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
cl_mem write_buffer = clCreateBuffer(context, CL_MEM_WRITE_ONLY, sizeof(float)*count, NULL, &err);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
// Copy the input data (blocking)
err = clEnqueueWriteBuffer(commands, read_buffer, CL_TRUE, 0, sizeof(float)*count, input_data, 0, NULL, NULL);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
// Set the kernel arguments
err = 0;
err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &read_buffer);
err |= clSetKernelArg(kernel, 1, sizeof(cl_mem), &write_buffer);
err |= clSetKernelArg(kernel, 2, sizeof(unsigned int), &count);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
std::clock_t c_start = std::clock();
// Execute the kernel
err = clEnqueueNDRangeKernel(commands, kernel, 1, NULL, &global, NULL, 0, NULL, NULL);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
// Give it time to finish
clFinish(commands);
float* output_data = (float*)malloc(count * sizeof(float));
err = clEnqueueReadBuffer(commands, write_buffer, CL_TRUE, 0, sizeof(float)*count, output_data, 0, NULL, NULL);
if (err != CL_SUCCESS) printf("ERROR at line %u\n", __LINE__);
std::clock_t c_end = std::clock();
printf("Test duration (OpenCL): %f ms\n", 1000.0 * (c_end - c_start) / CLOCKS_PER_SEC);
float *test_result = (float*)malloc(count * sizeof(float));
c_start = std::clock();
for (int i = 0; i < count; i++) {
float num = input_data[i];
test_result[i] = num*num;
}
c_end = std::clock();
printf("Test duration (regular): %f ms\n", 1000.0 * (c_end - c_start) / CLOCKS_PER_SEC);
printf("Validating..:\n");
// Validate everything
int correct_opencl = 0;
int correct_regular = 0;
for (int i = 0; i < count; i++) {
if (output_data[i] == input_data[i]*input_data[i]) correct_opencl++;
if (test_result[i] == input_data[i]*input_data[i]) correct_regular++;
}
printf("%d/%d correct results (OpenCL)\n", correct_opencl, count);
printf("%d/%d correct results (regular)\n", correct_regular, count);
}