User subroutines and STAR-HPC v3.2

Lars Ola Liavåg

Hello everybody,

I've got myself an apparently well-working newxyz.f file to handle the piston motion of my ICE simulation. That is, it works as long as I run sequential. When I run parallel, I get warning#113 upon execution of my cell layer deactivation events, stating that the cell side areas of the deactivated cells are non-zero. Plotting the resulting mesh in prostar shows that although the cell layers seemingly have been deactivated, the vertices of the layers have not been collapsed onto each other but continue to move as if there were no deactivation events. As a result, I get very distorted cells and bogus results.

Interestingly, I get warning #113 in sequential mode too but only when I run single precision and there are no evident problems, just the warning. Running sequential with double precision gives no no warnings either. Why then should this warning appear in parallel and double precision? And what are the vertex problems about?

An acquaintant of me mentioned that v3.2 is very picky about variable declarations hand he'd found that he couldn't make some subroutines work in parallel unless using implicit declarations of real and double precision variables. Could this be part of the answer. Has anybody got any experience to share with me on this issue?

Regs. Lars Ola