I have some problems with unroling loops, AMD retun clBuildProgram, with Intel OCL works fine
code:
__kernel void conv(...)
{
#pragma unroll (SIZECOEF+8)
for (j = 1; j <= SIZECOEF+8; j++)
{
....
}
}
compiler parameters
-DSIZEFAC=288 -DSIZECOEF=49 -DMAX_N_PAR=200 -DLMAX=6 -DMMAX=6