On Tue, Apr 08, 2025, FyD via AMBER wrote:
>CentOS 7.6 + devtoolset8
>openmpi-4.1.8 compiled with devtoolset8
>export TESTsander='/usr/local/amber24///bin/pmemd.MPI' && cd
>hrem_wat_pv && ./Run.rem
>
> Running multipmemd version of pmemd Amber24
> Total processors = 4
> Number of groups = 2
>[...]
>Program received signal SIGSEGV: Segmentation fault - invalid memory
We still occasionally see this error, but it is very sporadic, and seems to
involve some combination of compilers and MPI stack being used. It may be
specific to this test input, but it is hard to reproduce.
Best advice is to just be alert for segmentation faults if you are running
Hamiltonian replica exchange.
[Dan: do you recall what was done about this a year ago?]
>About "DO_PARALLEL set to mpirun -np 4; max is 2" in 2):
>does the later message mean this test crashed because 4 cores are
>requested instead of 2...
No: this should just mean that the test is skipped.
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Apr 08 2025 - 14:00:03 PDT
