修复 opencl 程序的框图以及可能的程序本身

使用 OpenCL 并使用本地内存编写程序。 任务：b=min(A+C)

我有一个准备好的框图，但老师说它按顺序显示所有内容，并且我们有一个并行程序，我不知道如何修复框图。

我还想知道我是否正确使用本地内存编写了程序。

链接到带有框图的图像： https://www.imghippo.com/i/1704956923.jpg

#define _CRT_SECURE_NO_WARNINGS
#define CL_USE_DEPRECATED_OPENCL_1_2_APIS
#include <CL/cl.h>
#include <CL/cl_platform.h>
#include <iostream>
#include <cstdlib>
#include <stdio.h>

#define MAX_SOURCE_SIZE (0x100000)
#define LIST_SIZE 64

int main(void) {
    int* A = new int[LIST_SIZE];
    int* C = new int[LIST_SIZE];
    int* B = new int[LIST_SIZE];
    srand(time(0));
    for (int i = 0; i < LIST_SIZE; i++) {
        A[i] = (rand() % 200) / 2.;
        C[i] = (rand() % 200) / 2.;
    }

    FILE* fp;
    char* source_str;
    size_t source_size;

    fp = fopen("kernel.cl", "r");
    if (!fp) {
        fprintf(stderr, "Failed to load kernel.\n");
        exit(1);
    }
    source_str = new char[MAX_SOURCE_SIZE];
    source_size = fread(source_str, 1, MAX_SOURCE_SIZE, fp);
    fclose(fp);

    cl_platform_id platform_id = NULL;
    cl_device_id device_id = NULL;
    cl_uint ret_num_devices;
    cl_uint ret_num_platforms;
    cl_int ret = clGetPlatformIDs(1, &platform_id, &ret_num_platforms);
    ret = clGetDeviceIDs(platform_id, CL_DEVICE_TYPE_DEFAULT, 1, &device_id, &ret_num_devices);

    cl_context context = clCreateContext(NULL, 1, &device_id, NULL, NULL, &ret);

    cl_command_queue command_queue = clCreateCommandQueueWithProperties(context, device_id, 0, &ret);

    cl_mem a_mem_obj = clCreateBuffer(context, CL_MEM_READ_ONLY, LIST_SIZE * sizeof(int), NULL, &ret);
    cl_mem c_mem_obj = clCreateBuffer(context, CL_MEM_READ_ONLY, LIST_SIZE * sizeof(int), NULL, &ret);
    cl_mem b_mem_obj = clCreateBuffer(context, CL_MEM_WRITE_ONLY, LIST_SIZE * sizeof(int), NULL, &ret);

    ret = clEnqueueWriteBuffer(command_queue, a_mem_obj, CL_TRUE, 0, LIST_SIZE * sizeof(int), A, 0, NULL, NULL);
    ret = clEnqueueWriteBuffer(command_queue, c_mem_obj, CL_TRUE, 0, LIST_SIZE * sizeof(int), C, 0, NULL, NULL);

    cl_program program = clCreateProgramWithSource(context, 1, (const char**)&source_str, (const size_t*)&source_size, &ret);

    ret = clBuildProgram(program, 1, &device_id, NULL, NULL, NULL);

    cl_kernel kernel = clCreateKernel(program, "krn", &ret);

    size_t local_item_size = 64;
    ret = clSetKernelArg(kernel, 0, sizeof(cl_mem), (void*)&a_mem_obj);
    ret = clSetKernelArg(kernel, 1, sizeof(cl_mem), (void*)&c_mem_obj);
    ret = clSetKernelArg(kernel, 2, sizeof(cl_mem), (void*)&b_mem_obj);
    ret = clSetKernelArg(kernel, 3, local_item_size * sizeof(int), NULL);

    size_t global_item_size = LIST_SIZE;
    ret = clEnqueueNDRangeKernel(command_queue, kernel, 1, NULL, &global_item_size, &local_item_size, 0, NULL, NULL);

    ret = clEnqueueReadBuffer(command_queue, b_mem_obj, CL_TRUE, 0, LIST_SIZE * sizeof(int), B, 0, NULL, NULL);

    for (int i = 0; i < LIST_SIZE; i++) {
        printf("%d. sum: %d\n", i, B[i]);
    }

    int min = B[0];
    int minIndex = 0;
    for (int i = 0; i < LIST_SIZE; i++) {
        if (B[i] < min) {
            min = B[i];
            minIndex = i;
        }
    }
    printf("\n\n%d. Minimum min(A+C) = %d\n", minIndex, min);

    ret = clFlush(command_queue);
    ret = clFinish(command_queue);
    ret = clReleaseKernel(kernel);
    ret = clReleaseProgram(program);
    ret = clReleaseMemObject(a_mem_obj);
    ret = clReleaseMemObject(c_mem_obj);
    ret = clReleaseMemObject(b_mem_obj);
    ret = clReleaseCommandQueue(command_queue);
    ret = clReleaseContext(context);
    delete[] source_str;
    delete[] A;
    delete[] B;
    delete[] C;
    return 0;
}

内核.cl：

__kernel void krn(__global const int* A, __global const int* C, __global int* B) {
    int i = get_global_id(0); 
    __local int localA[128];
    __local int localC[128];
    localA[i] = A[i];
    localC[i] = C[i];
    barrier(CLK_LOCAL_MEM_FENCE);
    B[i] = localA[i] + localC[i];
}

我尝试过的：

画了一个错误的框图并写了一个程序