I am facing an error when compiling a subroutine. I am wondering what the reason for that could be. The subroutine consists out of about 6 acccelerated loops. One of them shows very poor performacne. However when I compile this version of the subroutien it works.
In order to gain performance in that particular slow loop I did some minor changes in the code, meaning that I replaced some private variables with arrays and then devided this one loop into 3 loops.
When I compile the modified subroutine without -ta=nvidia (runnning on host) it works fine. However when I try to compile it with -ta=nvidia it gives me the following error:
sub1_GPU.f90:
PGF90-F-0000-Internal compiler error. ACC CG: could not find ALV index for symbol 1478 (sub1_GPU.f90: 1121)
Thank you for your reply!