opencl使用demo

以下是一个简单的 OpenCL 使用示例，它执行向量加法：

#include <stdio.h>
#include <stdlib.h>
#include <CL/cl.h>

#define LENGTH 1024

const char *kernelSource = 
"__kernel void vectorAdd(__global float *a, __global float *b, __global float *c) {\n"
"    int i = get_global_id(0);\n"
"    c[i] = a[i] + b[i];\n"
"}\n";

int main() {
    cl_platform_id platform;
    cl_device_id device;
    cl_context context;
    cl_command_queue queue;
    cl_program program;
    cl_kernel kernel;
    cl_mem a, b, c;
    cl_int err;
    int i;
    float *A, *B, *C;

    // 初始化 A, B, C 数组
    A = (float*) malloc(sizeof(float) * LENGTH);
    B = (float*) malloc(sizeof(float) * LENGTH);
    C = (float*) malloc(sizeof(float) * LENGTH);
    for (i = 0; i < LENGTH; i++) {
        A[i] = (float) i;
        B[i] = (float) (LENGTH - i);
        C[i] = 0.0f;
    }

    // 获取平台和设备
    err = clGetPlatformIDs(1, &platform, NULL);
    err = clGetDeviceIDs(platform, CL_DEVICE_TYPE_GPU, 1, &device, NULL);

    // 创建 OpenCL 上下文和命令队列
    context = clCreateContext(NULL, 1, &device, NULL, NULL, &err);
    queue = clCreateCommandQueue(context, device, 0, &err);

    // 创建和编译内核程序
    program = clCreateProgramWithSource(context, 1, &kernelSource, NULL, &err);
    err = clBuildProgram(program, 1, &device, NULL, NULL, NULL);

    // 创建内核
    kernel = clCreateKernel(program, "vectorAdd", &err);

    // 创建和设置缓冲区
    a = clCreateBuffer(context, CL_MEM_READ_ONLY, sizeof(float) * LENGTH, NULL, &err);
    b = clCreateBuffer(context, CL_MEM_READ_ONLY, sizeof(float) * LENGTH, NULL, &err);
    c = clCreateBuffer(context, CL_MEM_WRITE_ONLY, sizeof(float) * LENGTH, NULL, &err);
    err = clEnqueueWriteBuffer(queue, a, CL_TRUE, 0, sizeof(float) * LENGTH, A, 0, NULL, NULL);
    err = clEnqueueWriteBuffer(queue, b, CL_TRUE, 0, sizeof(float) * LENGTH, B, 0, NULL, NULL);

    // 设定内核参数
    err = clSetKernelArg(kernel, 0, sizeof(cl_mem), &a);
    err = clSetKernelArg(kernel, 1, sizeof(cl_mem), &b);
    err = clSetKernelArg(kernel, 2, sizeof(cl_mem), &c);

    // 执行内核
    size_t globalSize = LENGTH;
    size_t localSize = 64;
    err = clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &globalSize, &localSize, 0, NULL, NULL);

    // 读取结果
    err = clEnqueueReadBuffer(queue, c, CL_TRUE, 0, sizeof(float) * LENGTH, C, 0, NULL, NULL);

    // 打印结果
    for (i = 0; i < LENGTH; i++) {
        printf("%f + %f = %f\n", A[i], B[i], C[i]);
    }

    // 释放内存和资源
    free(A);
    free(B);
    free(C);
    clReleaseMemObject(a);
    clReleaseMemObject(b);
    clReleaseMemObject(c);
    clReleaseKernel(kernel);
    clReleaseProgram(program);
    clReleaseCommandQueue(queue);
    clReleaseContext(context);

    return 0;
}

这个程序的作用是将两个向量相加，并打印结果。程序首先初始化了三个数组 A, B, C，然后获取 OpenCL 平台和设备，创建上下文和命令队列，创建和编译内核程序，创建内核，创建和设置缓冲区，设定内核参数，执行内核，读取结果，打印结果，最后释放内存和资源。