OpenCL实现向量加的简单例子

看了两天OpenCL的入门书，对OpenCL编程的流程有了大概的了解。下面是一个实现向量加的例子，从网上拿过来的，刚开始有些问题，修改以后可以跑通。代码如下：

 

 #include <stdio.h>#include <stdlib.h>#include <CL/cl.h>#define LEN(arr) (sizeof(arr) / sizeof(arr[0]))//kernel source codeconst char* src[] = {"__kernel void vec_add(__global const float *a, __global const float *b, __global float *c)/n","{/n","int gid = get_global_id(0);/n","c[gid] = a[gid] + b[gid];/n","}/n"};int main(){cl_uint numPlatforms;cl_platform_id platform = NULL;cl_int status = clGetPlatformIDs(0, NULL, &numPlatforms);if(status != CL_SUCCESS){printf("error: getting platforms/n");exit(1);}cl_platform_id *platforms = (cl_platform_id *)malloc(numPlatforms*sizeof(cl_platform_id));status = clGetPlatformIDs(numPlatforms, platforms, NULL);if(status != CL_SUCCESS){printf("error: getting platform IDs/n");exit(1);}platform = platforms[0];cl_context_properties cps[3] ={CL_CONTEXT_PLATFORM,(cl_context_properties) platform, 0};cl_context_properties *cprops =(NULL == platform) ? NULL :cps;//create OpenCL contextcl_context context = clCreateContextFromType(cprops, CL_DEVICE_TYPE_GPU, NULL, NULL, &status);if(status != CL_SUCCESS){printf("create context error/n");exit(1);}//get device id from contextsize_t cb;status = clGetContextInfo(context, CL_CONTEXT_DEVICES, 0, NULL, &cb);if(status != CL_SUCCESS){printf("get context info 1 error/n");exit(1);}cl_device_id *devices= malloc(cb);status = clGetContextInfo(context, CL_CONTEXT_DEVICES, cb, devices, NULL);if(status != CL_SUCCESS){printf("get context info 2 error/n");exit(1);}//create a command queuecl_command_queue cmd_queue = clCreateCommandQueue(context, devices[0], 0, NULL);//create kernelcl_program program = clCreateProgramWithSource(context, LEN(src),src, NULL, NULL);status = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);cl_kernel kernel = clCreateKernel(program, "vec_add", NULL);//host initializesize_t n = 5;float srcA[] = {1, 2, 3, 4, 5};float srcB[] = {5, 4, 3, 2, 1};float dst[n];//set the input and output arguments of kernelcl_mem memobjs[3];memobjs[0] = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, sizeof(float)*n, srcA, NULL);memobjs[1] = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, sizeof(float)*n, srcB, NULL);memobjs[2] = clCreateBuffer(context, CL_MEM_WRITE_ONLY, sizeof(float)*n, NULL, NULL);//set "a", "b", "c" vector argumentstatus = clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *)&memobjs[0]);status |= clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *)&memobjs[1]);status |= clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *)&memobjs[2]);size_t global_work_size[1] = {n};//execute kernelstatus = clEnqueueNDRangeKernel(cmd_queue, kernel, 1, NULL, global_work_size, NULL, 0, NULL, NULL);//read output arraystatus = clEnqueueReadBuffer(cmd_queue, memobjs[2], CL_TRUE, 0, n*sizeof(cl_float), dst, 0, NULL, NULL);for(int i=0; i<n; ++i){printf("-> %.2f/n", dst[i]);}return 0;}

 

从以上的代码可以基本得出OpenCL编程的流程：

1、获得平台，clGetPlatformIDs

2、创建上下文，clCreateContextFromType，这个函数的第一个参数为NULL时在AMD的平台上运行不能通过，必须为该函数指定第一个参数。

3、通过上下文得到设备信息，clGetContextInfo

4、为相应设备创建comandQueue， clCreateCommandQueue

5、创建源程序，build源程序，生成kernel

6、分配buffer空间，逐个设置程序参数。

7、执行kernel，clEnqueueNDRangeKernel

8、从buffer读回数据，clEnqueueReadBuffer