1

我有这段代码（从 JOCL 示例修改），其中 CL 内核采用 float 和 char[] 指针：

import static org.jocl.CL.*;
import org.jocl.*;

public class JOCLEncode {

    private static String kernelSource = 
            "__kernel void "+
            "sampleKernel(__global float *c, __global char[] *t)"+
            "{"+
            "    int gid = get_global_id(0);"+
            "    c[gid] = c[gid] + 4;"+
            "}";

    public static void main(String[] a) {

        float dstArray[] = new float[5];
        char charArray[] = new char[2];
        Pointer charsPointer = Pointer.to(charArray);
        Pointer dstPointer = Pointer.to(dstArray);

        final int platformIndex = 0;
        final long deviceType = CL_DEVICE_TYPE_ALL;
        final int deviceIndex = 0;

        CL.setExceptionsEnabled(true);

        int numPlatformsArray[] = new int[1];
        clGetPlatformIDs(0, null, numPlatformsArray);
        int numPlatforms = numPlatformsArray[0];
        System.out.println("Num devices available: " + numPlatformsArray[0]);

        cl_platform_id platforms[] = new cl_platform_id[numPlatforms];
        clGetPlatformIDs(platforms.length, platforms, null);
        cl_platform_id platform = platforms[platformIndex];

        cl_context_properties contextProperties = new cl_context_properties();
        contextProperties.addProperty(CL_CONTEXT_PLATFORM, platform);

        int numDevicesArray[] = new int[1];
        clGetDeviceIDs(platform, deviceType, 0, null, numDevicesArray);
        int numDevices = numDevicesArray[0];

        cl_device_id devices[] = new cl_device_id[numDevices];
        clGetDeviceIDs(platform, deviceType, numDevices, devices, null);
        cl_device_id device = devices[deviceIndex];

        cl_context context = clCreateContext(contextProperties, 1, new cl_device_id[]{device}, null, null, null);

        cl_command_queue commandQueue = clCreateCommandQueue(context, device, 0, null);

        cl_mem memObjects[] = new cl_mem[2];
        memObjects[0] = clCreateBuffer(context, CL_MEM_READ_WRITE, Sizeof.cl_float * 5, null, null);
        System.out.println("Created shared memory");
        memObjects[1] = clCreateBuffer(context, CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, Sizeof.cl_float * 5, charsPointer, null);

        //make and build program
        String[] kernelSrc = {kernelSource};
        cl_program program = clCreateProgramWithSource(context, 1, kernelSrc, null, null);
        clBuildProgram(program, 0, null, null, null, null);
        cl_kernel kernel = clCreateKernel(program, "sampleKernel", null);

         clSetKernelArg(kernel, 0, Sizeof.cl_mem, Pointer.to(memObjects[0]));
         clSetKernelArg(kernel, 1, Sizeof.cl_mem, Pointer.to(memObjects[1]));

         //i think this is how many times it runs
         long global_work_size[] = {5};
         long local_work_size[] = {1};

         //execution ?
         clEnqueueNDRangeKernel(commandQueue, kernel, 1, null, global_work_size, local_work_size, 0, null, null);

         //i think it puts the finished vals in dst
         clEnqueueReadBuffer(commandQueue, memObjects[0], CL_TRUE, 0, 5 * Sizeof.cl_float, dstPointer, 0, null, null);

         clReleaseMemObject(memObjects[0]);
         clReleaseMemObject(memObjects[1]);
         clReleaseKernel(kernel);
         clReleaseProgram(program);
         clReleaseCommandQueue(commandQueue);
         clReleaseContext(context);

         System.out.println("Result: " + java.util.Arrays.toString(dstArray));
    }

}

我认为问题出在内核中，特别是 . char[]，但我无法弄清楚。是它，还是它在构建它？

4

0 回答 0