svm.c

#include <stdio.h>
#include <stdlib.h>

#ifdef __APPLE__
#include <OpenCL/opencl.h>
#else
#include <CL/cl.h>
#endif

#define MAX_SOURCE_SIZE (0x100000)

struct node {
    int value;
    node* next;
};

int main(void) {
    int i;
    const int LIST_SIZE = 1024;

    // Load the kernel source code into the array source_str
    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 = (char*)malloc(MAX_SOURCE_SIZE);
    source_size = fread( source_str, 1, MAX_SOURCE_SIZE, fp);
    fclose( fp );

    // Get platform and device information
    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_ALL, 1, &device_id, &ret_num_devices);
    
    // Create an OpenCL context
    cl_context context = clCreateContext(NULL, 1, &device_id, NULL, NULL, &ret);

    // Create a command queue
    cl_command_queue command_queue = clCreateCommandQueue(context, device_id, 0, &ret);

    // Create memory buffers on the device for each vector 
    node **a = (node**)clSVMAlloc(context, CL_MEM_READ_ONLY, 
            LIST_SIZE*sizeof(node), NULL, 0);
    node **b = (node**)clSVMAlloc(context, CL_MEM_READ_ONLY,
            LIST_SIZE*sizeof(node), NULL, 0);
    node **c = (node**)clSVMAlloc(context, CL_MEM_WRITE_ONLY, 
            LIST_SIZE*sizeof(node), NULL, 0);
    
    for (int i = 0; i < LIST_SIZE; ++i) {
        a[i] = (node*)clSVMAlloc(context, CL_MEM_WRITE_ONLY, 
            sizeof(node), NULL, 0);
        b[i] = (node*)clSVMAlloc(context, CL_MEM_WRITE_ONLY, 
            sizeof(node), NULL, 0);
        c[i] = (node*)clSVMAlloc(context, CL_MEM_WRITE_ONLY, 
            sizeof(node), NULL, 0);
    
        a[i]->next = (node*)clSVMAlloc(context, CL_MEM_WRITE_ONLY, 
            sizeof(node), NULL, 0);
        b[i]->next = (node*)clSVMAlloc(context, CL_MEM_WRITE_ONLY, 
            sizeof(node), NULL, 0);
        c[i]->next = (node*)clSVMAlloc(context, CL_MEM_WRITE_ONLY, 
            sizeof(node), NULL, 0);
        
        a[i]->next->next = (node*)clSVMAlloc(context, CL_MEM_READ_ONLY, 
            sizeof(node), NULL, 0);
        b[i]->next->next = (node*)clSVMAlloc(context, CL_MEM_READ_ONLY,
            sizeof(node), NULL, 0);
        c[i]->next->next = (node*)clSVMAlloc(context, CL_MEM_WRITE_ONLY, 
            sizeof(node), NULL, 0);
    }

    // XXX: SVMmap Page 173
    for (int i = 0; i < LIST_SIZE; ++i) {
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, a[i], sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, a[i]->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, a[i]->next->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, b[i], sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, b[i]->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, b[i]->next->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_WRITE, c[i], sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_WRITE, c[i]->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_WRITE, c[i]->next->next, sizeof(node),
                0, NULL, NULL);
    }
    for(int i = 0; i < LIST_SIZE; ++i) { 
        a[i]->value = LIST_SIZE-3;
        a[i]->next->value = LIST_SIZE-2;
        a[i]->next->next->value = LIST_SIZE-1;

        b[i]->value = 3;
        b[i]->next->value = 2;
        b[i]->next->next->value = 1;
    
        c[i]->value = 0;
        c[i]->next->value = 0;
        c[i]->next->next->value = 0;
    }
    
    // : svmunmap
   for (int i = 0; i < LIST_SIZE; ++i) { 
        clEnqueueSVMUnmap(command_queue, a[i],
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, a[i]->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, a[i]->next->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, b[i],
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, b[i]->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, b[i]->next->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, c[i],
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, c[i]->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, c[i]->next->next,
                0, NULL, NULL);
    }


    // Copy the lists A and B to their respective memory buffers

    // Create a program from the kernel source
    cl_program program = clCreateProgramWithSource(context, 1, 
            (const char **)&source_str, (const size_t *)&source_size, &ret);

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

    // Create the OpenCL kernel
    cl_kernel kernel = clCreateKernel(program, "vector_add", &ret);

    // Set the arguments of the kernel
    ret = clSetKernelArgSVMPointer(kernel, 0, (void *)a);
    ret = clSetKernelArgSVMPointer(kernel, 1, (void *)b);
    ret = clSetKernelArgSVMPointer(kernel, 2, (void *)c);
    
    void *ptr_arr[];
    for (int i = 0; i  < LIST_SIZE; i++) {
        ptr_arr[0+i*9] = a[i];
        ptr_arr[1+i*9] = b[i];
        ptr_arr[2+i*9] = c[i];
        ptr_arr[3+i*9] = a[i]->next;
        ptr_arr[4+i*9] = a[i]->next->next;
        ptr_arr[5+i*9] = b[i]->next;
        ptr_arr[6+i*9] = b[i]->next->next;
        ptr_arr[7+i*9] = c[i]->next;
        ptr_arr[8+i*9] = c[i]->next->next;
    }
    // : set kern arg exec info ???:value? Page:214
    ret = clSetKernelExecInfo(kernel, CL_KERNEL_EXEC_INFO_SVM_PTRS,
            9*LIST_SIZE*sizeof(void *), ptr_arr);
    
    // XXX: flatten, move data from our pool to buffer
    // XXX: kernel rewriting
    
    // Execute the OpenCL kernel on the list
    size_t global_item_size = LIST_SIZE; // Process the entire lists
    size_t local_item_size = 64; // Process in groups of 64
    ret = clEnqueueNDRangeKernel(command_queue, kernel, 1, NULL, 
            &global_item_size, &local_item_size, 0, NULL, NULL);

    // : svnmap
    for(int i = 0; i < LIST_SIZE; i++) {
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, a[i], sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, a[i]->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, a[i]->next->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, b[i], sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, b[i]->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_READ, b[i]->next->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_WRITE, c[i], sizeof(node),
                0, NULL, NULL);
       clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_WRITE, c[i]->next, sizeof(node),
                0, NULL, NULL);
        clEnqueueSVMMap(command_queue, CL_TRUE, 
                CL_MAP_WRITE, c[i]->next->next, sizeof(node),
                0, NULL, NULL);
    } 
    // Display the result to the screen
    for (int i = 0; i < LIST_SIZE = 3; ++i) { 
        printf("%d + %d = %d\n", a[i]->value, b[i]->value, c[i]->value);
        printf("%d + %d = %d\n", a[i]->next->value, b[i]->next->value, c[i]->next->value);
        printf("%d + %d = %d\n", a[i]->next->next->value, b[i]->next->next->value, c[i]->next->next->value);
    }

    // : svmunmap
    for (int i = 0; i < LIST_SIZE; ++i) {
        clEnqueueSVMUnmap(command_queue, a[i],
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, a[i]->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, a[i]->next->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, b[i],
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, b[i]->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, b[i]->next->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, c[i],
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, c[i]->next,
                0, NULL, NULL);
        clEnqueueSVMUnmap(command_queue, c[i]->next->next,
                0, NULL, NULL);
    }
   
    // Clean up
    ret = clFlush(command_queue);
    ret = clFinish(command_queue);
    ret = clReleaseKernel(kernel);
    ret = clReleaseProgram(program);
    for (int i = 0; i < LIST_SIZE; i++){
        ret = clSVMFree(context, a[i]->next->next);
        ret = clSVMFree(context, b[i]->next->next);
        ret = clSVMFree(context, c[i]->next->next);
        ret = clSVMFree(context, a[i]->next);
        ret = clSVMFree(context, b[i]->next);
        ret = clSVMFree(context, c[i]->next);
        ret = clSVMFree(context, a[i]);
        ret = clSVMFree(context, b[i]);
        ret = clSVMFree(context, c[i]);
    }
    ret = clSVMFree(context, a);
    ret = clSVMFree(context, b);
    ret = clSVMFree(context, c);
    return 0;
}