Hi all,
anyone knows if it is possible to implement this line of PTX code:
add.sat.itype d, a, b
in CUDA?
I need to limit result to MAXINT on integer addition overflow, but I would change only the CUDA code.
Thank you.
well, if I change the PTX code, and I compile it into a cubin file, how can I create the .cu.cpp file (as nvcc do it with -cuda parameter) with this custom device code? I tried to use “bin2c”, and I tried to put the result in the old .cu.cpp file, without success…