gistfile1.h

kernel void matrixTranspose( global float *a, global float *b, constant float *c, constant int size )
{
	int idx = get_global_id(0);
	int idy = get_global_id(1);
	b[idx * size + idy] = a[idy * size + idx];
	int BLOCKSIZE = 0;
	
	int loops;
	for (loops=0;loops<50;loops++)
	{
		b[idx * size + idy] += a[idy * size + idx] / 2;
	
		b[idx * size + idy] += a[idx * size + idy] / 2;
		
		b[idx * size + idy] += c[idx];
		
		if( b[idx * size + idy] < 50 ){
			b[idx * size + idy] = ( b[idx * size + idy] / 100 ) * ( c[idx] / 20 );
		}
		
		if(idx % 2 == 0){
			b[idx * size + idy] = b[idx * size + idy] - 20 + c[size - idx];
		}
	}
	
	if((idx % (size)) == size - 1){
		int i;
		float som = 0;
		for (i=0; i<(size); i++) {
			som += b[idx * size + i];
		}
		som = som / (size);
		b[idx * size + idy] = som;
	}
}

gistfile2.h

kernel void matrixTranspose( global float *a, global float *b, constant float *c, constant int size, global float *som )
{
	int idx = get_global_id(0);
	int idy = get_global_id(1);
 
	int loops;
 
	float cc = c[idx];
 
	float c2 = cc / 20;
 
	float cidx = c[size - idx];
 
	float ayx = a[idy * size + idx];
 
	b[idx * size + idy] = ayx;
 
	float bxy = b[idx * size + idy];
 
	float axy = a[idx * size + idy];
 
	for (loops=0;loops<50;loops++)
	{
		/* bxy += ayx / 2 + axy / 2 + cc; */
		bxy += native_divide(ayx,2.0f) + native_divide(axy,2.0f) + cc;
		if( bxy < 50 ){
			bxy = native_divide( bxy , 100.0f ) * c2;
		}
 
		if(idx % 2 == 0){
			bxy = bxy - 20 + cidx;
		}
	}
 
	b[idx * size + idy] = bxy;
 
	if(idx == size - 1){
		int i;
		float som = 0;
		for(i=0;i<size;i++)
		{
			float s = b[idx * size +i];
			som += s;
		}
		b[idx * size + idy] = som;
	}
}