Archives Discussions

[scharupa]

Hi,

Could please help me how to write below c programm in opencl kernal. I tried different scenarios but nothing worked.

int A[5], temp=0, i;

//scanning A[5] values from user///

for(i=0; i>5; i++)

{ if(A>max){

max=A;}

}

printf("%d", max);

Could you please some give me opencl kernal for the same logic

[nou]

look into reduction example or search parallel reduction algorithm.

[mishka_p]

Hi

Look at "AMD Accelerated Parallel Processing" pdf document, latest version. On page 1-29 is the code for finding the minimum. I think it will help you.

[scharupa]

Thanks for your support.

I got the logic. We need to print K[4] from below kernal which gives larest values.

here is the kernel:

__kernel
void vecadd(__global int *A,
            __global int *B,
            __global int *K) {

   int idx = get_global_id(0);
 int x=0;
    for(int i=4; i>0; i--)
    {
           if(A[idx]>x)
            {
             x=A[idx];   K[idx]=x;
            }
            else { K[idx]=x;}
               A[idx]=A[idx++];  
    }
        
   }

[scharupa]

Sorry, we dont need "__global int *B" in the kernal.

[nou]

why do you decrease? on GPU accessing memory in reverse can have some serious speed penalty