[AMBER] sander.MPI gives segfault if compiled with Intel compilers

From: viktor drobot <linux776.gmail.com>
Date: Sun, 15 Sep 2019 16:34:55 +0300

I've compiled Amber18 with Intel compilers (15.0.3) and with
corresponding MKL version linked (lp64 variant, if that matters). I
tested with different OpenMPI versions - 1.8.4, 1.8.4 with MXM
extension, 2.1.1 and 2.1.1 with MXM. All my runs are performed on our
university cluster. The problem is when I try to run computation with
sander.MPI then I get mpirun error with code 11 - segmentation fault.
However, running the same system and config with sander, pmemd or
pmemd.MPI works just fine without any segfaults. I tried running
sander.MPI on 1, 2 and 4 cores and the result always the same - just
silent segfault. There are no special info in mdout file - just seems
like writing wasn't finished. This problem doesn't happens when the same
AMBER distro is compiled with GNU compilers (but without MKL). So I
think this is a bug with sander.MPI module.

It's very pity because I can't use QM/MM with high-performance Intel
compilers and MKL library which gives some boost. What I can do to find
out the cause of this bug? Thank you!

AMBER mailing list
Received on Sun Sep 15 2019 - 07:00:02 PDT
Custom Search