Debugging -ta=multicore without using the -g flag

Accelerated Computing HPC Compilers Legacy PGI Compilers
1

I had to remove the -g flags to compile with the -ta=multicore flag. It compiles without a problem, but when the executable is ran, it core dumps (seg faults). Normally would run under Valgrind, but with the -g option not being allowed to be enabled alongside the -ta=multicore flag, it is a bit harder to debug. My next option would be to start adding print statements until figuring out around where the program is seg faulting. I was wondering if there is a more efficient manner of debugging the -ta=multicore code generated executable?

Thank you!

2

You can try compiling with -gopt. That might make debugging easier. Also, if you set ACC_NUM_CORES=1, you will run on just one core and that might help to determine if the problem is due to parallelization or not.
Which PGI version are you using? Improvements to -ta=multicore are being made fairly rapidly.