I failed to debug Cuda kernel function. I tried to add Â¨-GÂ¨ option to nvcc but failed. Some errors exist when linking fortran object code to cuda object code. For example, undefined reference to $gpu_registers. How to debug device kernel code? Thanks a lot.
as far as I know, all it should take to debug device code is to add the flag “-deviceemu” to the nvcc command line;
however, I did not succeed so far; see my thread at http://forums.nvidia.com/index.php?showtop…t=&#entry438171
Hope you have more luck,
and let me know when you succeed!
p.s.: which environement (Win+VS, Linux…?) are you using?