cancel
Showing results for 
Search instead for 
Did you mean: 

OpenCL

apcool
Journeyman III

copy program in opencl....

hi...i m trying to write a opencl program for copy an array into another array....

this is the Kernel which i have written.....

__kernel void vector_add(__global const int *A, __global const int *B) {

    // Get the index of the current element to be processed

    int i = get_global_id(0);

    // Do the operation

    B = A;

}

and the host program for this....

//this program is to copy one arry to another..........

#include <stdio.h>

#include <stdlib.h>

#include <iostream>

#include<time.h>

#ifdef __APPLE__

#include <OpenCL/opencl.h>

#else

#include <CL/cl.h>

#endif

#define MAX_SOURCE_SIZE (0x100000)

double timeDifference = 0;

struct timespec start, stop;

int size=1024;

//#define LIST_SIZE 1024

using namespace std;

int * mean(int c[],int list_size)

{

   int i;

    int *A = (int*)malloc(sizeof(int)*(list_size));

    int *B = (int*)malloc(sizeof(int)*(list_size));

 

   // int *C = (int*)malloc(sizeof(int)*(list_size/2));

  

    // Load the kernel source code into the array source_str

    FILE *fp;

    char *source_str;

    size_t source_size;

    fp = fopen("copy_array_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_GPU, 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

    cl_mem a_mem_obj = clCreateBuffer(context, CL_MEM_READ_ONLY,

            (list_size) * sizeof(int), NULL, &ret);

    cl_mem b_mem_obj = clCreateBuffer(context, CL_MEM_READ_ONLY,

            (list_size) * sizeof(int), NULL, &ret);

  //  cl_mem c_mem_obj = clCreateBuffer(context, CL_MEM_WRITE_ONLY,

          //  (list_size) * sizeof(int), NULL, &ret);

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

    ret = clEnqueueWriteBuffer(command_queue, a_mem_obj, CL_TRUE, 0,

            (list_size) * sizeof(int), c, 0, NULL, NULL);

    //ret = clEnqueueWriteBuffer(command_queue, b_mem_obj, CL_TRUE, 0,

          //  (list_size) * sizeof(int), B, 0, NULL, NULL);

    // 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 = clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *)&a_mem_obj);

    ret = clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *)&b_mem_obj);

   // ret = clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *)&c_mem_obj);

     

      clock_gettime(CLOCK_REALTIME, &start);

  

    // Execute the OpenCL kernel on the list

    size_t global_item_size = (list_size); // Process the entire lists

    size_t local_item_size = 512; // Divide work items into groups of 64

    ret = clEnqueueNDRangeKernel(command_queue, kernel, 1, NULL,

            &global_item_size, &local_item_size, 0, NULL, NULL);

    // Read the memory buffer C on the device to the local variable B

 

    ret = clEnqueueReadBuffer(command_queue, b_mem_obj, CL_TRUE, 0,

            (list_size) * sizeof(int), B, 0, NULL, NULL);

  

/*   // Display the result to the screen

    for(i = 0; i <list_size/2; i++)

        printf("%d\n",C);*/

   /* for(i=0;i<size/2;i++)

   {

       sum+=C;

      // printf(" %d",C);

   }*/

   

      clock_gettime(CLOCK_REALTIME, &stop);

  

    timeDifference += (double) (stop.tv_sec - start.tv_sec)

            + (double) (stop.tv_nsec - start.tv_nsec) / 1000000000;

    

     //printf("\nsum of array %d\n",sum);

        //getchar();    

    // Clean up

    ret = clFlush(command_queue);

    ret = clFinish(command_queue);

    ret = clReleaseKernel(kernel);

    ret = clReleaseProgram(program);

    ret = clReleaseMemObject(a_mem_obj);

    ret = clReleaseMemObject(b_mem_obj);

   // ret = clReleaseMemObject(c_mem_obj);

    ret = clReleaseCommandQueue(command_queue);

    ret = clReleaseContext(context);

    free(A);

   // free(B);

    return B;

}

int main()

{

  int i;

int *c = (int*)malloc(sizeof(int)*(size));int *f;

  for(i=0;i<size;i++)

  {

    // if(i<128)

      // c=0;

    // else

       c=1;

  }

  f=mean(c,size);

  for(i=0;i<size-1000;i++)

   {

    printf(" %d",f);

   }

cout << "parallel Total Required Time:" << timeDifference << endl;

 

  return 0;

}

but it is giving different output......can anybody tell me.....where is the mistake in this program........

0 Kudos
Reply
3 Replies
realhet
Miniboss

Re: copy program in opencl....

Hi,

I'm not sure if this is the error, but all I can find there is that you're writing into a CL_MEM_READ_ONLY buffer. (B)

But maybe you should check the 'ret' values at least with the debugger to ensure all OpenCL commands executed properly.

0 Kudos
Reply
himanshu_gautam
Grandmaster

Re: copy program in opencl....

Hi Ankit,

You wrote: "__kernel void vector_add(__global const int *A, __global const int *B) {"

I don't think B can be a const array, as you have to write into it. Also as suggested by realhet, it is recommended to check the error status from all OpenCL APIs. Check Some simple SDK Samples for more information.

0 Kudos
Reply
realhet
Miniboss

Re: copy program in opencl....

Hi,

That __global const thing sounded weird to me as well, I tested it a bit:

__global const  : wrong

__global : good

__constant :good

__constant const : good

__constant const const : also working (compiler compiles it without problem haha)

So the 'const' keyword is not needed here. And in my understanding it is only there to make the pointer read only, not the data. But that theory was failed when I checked it with __global.

0 Kudos
Reply