gpuqsort2

This code is a scalar implementation of a quick sort algorithm that had been ported to GPUs.
Code Snippet

//All variables in this code region are integers.

for(int bx =0; bx < blocks; bx++){
    start = params[bx].from;
    end = params[bx].end;
    int coal = (start&0xf);
    start = start-coal;
    pivot = params[bx].pivot;
    for(int tx = 0; tx < threads; tx++){

        int x = lengths->left[bx] + hist->left[bx*(threads)+tx]-1;
        int y = lengths->right[bx] - hist->right[bx*(threads)+tx];

        if(tx+start<end){
            unsigned int d = data[tx+start];

            if(!(tx<coal)){

                if(d<pivot)
                        data2[x--]=d;
                else
                if(d>pivot)
                        data2[y++]=d;
             }
        }

        for(unsigned int i=start+tx+threads; i<end; i+=threads){
            unsigned int d = data[i];

            if(d<pivot){
                data2[x--]=d;
            }else if(d>pivot)
                data2[y++]=d;
        }
    }
}