Re: [AMBER-Developers] MPICH Failures for REMD - PMEMD

From: Charles Lin <clin92.ucsd.edu>
Date: Mon, 21 Mar 2016 23:16:26 +0000

I just recompiled mpich-3.0.4. Same issue.

145 file comparisons passed
0 file comparisons failed
9 tests experienced an error
Test log file saved as /server-home1/charlie/amber-clean//logs/test_amber_parallel/2016-03-21_16-07-00.log
No test diffs to save!

$ which mpif90
~/mpich-3.0.4/bin/mpif90
$ which mpirun
~/mpich-3.0.4/bin/mpifun
$ mpif90 -show
gfortran -I/server-home1/charlie/mpich-3.0.4/include -I/server-home1/charlie/mpich-3.0.4/include -L/server-home1/charlie/mpich-3.0.4/lib -lmpichf90 -lmpich -lopa -lmpl -lrt -lpthread
$ which mpiexec
~/mpich-3.0.4/bin/mpiexec
$ mpiexec --version
    Version: 3.0.4
$ gcc -v
gcc version 4.8.5 20150623 (Red Hat 4.8.5-4) (GCC)
________________________________________
From: Ross Walker [ross.rosswalker.co.uk]
Sent: Monday, March 21, 2016 3:05 PM
To: AMBER Developers Mailing List
Subject: Re: [AMBER-Developers] MPICH Failures for REMD - PMEMD

Be sure the mpirun / mpiexec you use is the same one as the mpi environment you used to build AMBER with and that the compiler you used for AMBER is the same one (version etc) used to compile the MPI library. Such mismatches is a classic for causing weird segfaults in my experience.

> On Mar 21, 2016, at 15:02, Charles Lin <clin92.ucsd.edu> wrote:
>
> gcc version 4.8.5 20150623 (Red Hat 4.8.5-4) (GCC)
>
> Yea mpich is the only one I'm having errors with. I might have to recompile it later and test it again. If I'm the only one seeing these errors (which only occur for the REMD tests).
> ________________________________________
> From: Daniel Roe [daniel.r.roe.gmail.com]
> Sent: Monday, March 21, 2016 2:47 PM
> To: AMBER Developers Mailing List
> Subject: Re: [AMBER-Developers] MPICH Failures for REMD - PMEMD
>
> Hi,
>
> pmemd.MPI passes all REMD tests for me with the following config:
>
> Fedora 22 (4.4.4-200.fc22.x86_64)
> Compilers: Intel v 14.0
> MPI: mpich 3.0.3
> ./configure -nomtkpp -norism -nofftw3 -noemil --no-updates -mpi intel
>
> In $AMBERHOME/test
> make clean && DO_PARALLEL="mpiexec -n 8" make test.pmemd.REM
>
> Only 1 diff in the constant pH explicit REMD test:
>
> possible FAILURE: check 1AKI.solv10.md1.mdout.0.dif
> 109,110c109,110
> < NSTEP = 30 TIME(PS) = 4400.060 TEMP(K) = 271.67 PRESS = 0.
> < Etot = -52395.6420 EKtot = 11051.9552 EPtot =
> -63447.5972
> ---
>> NSTEP = 30 TIME(PS) = 4400.060 TEMP(K) = 298.07 PRESS = 0.
>> Etot = -51321.7523 EKtot = 12125.8449 EPtot = -63447.5972
>
> Are you using gnu compilers? What version?
>
> -Dan
>
> On Sat, Mar 19, 2016 at 3:57 PM, Charles Lin <clin92.ucsd.edu> wrote:
>> Hi all,
>>
>> I was wondering if anyone was seeing any REMD test case failures on master (segfaults) for PMEMD when using an older version of mpich such as v2.1.4 and v3.0.4. It has also failed for the yum install version of mpich. However these test cases pass with openmpi and mpich 3.2. I am currently using CentOS 7.
>>
>> -Charlie
>> _______________________________________________
>> AMBER-Developers mailing list
>> AMBER-Developers.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
>
>
> --
> -------------------------
> Daniel R. Roe, PhD
> Department of Medicinal Chemistry
> University of Utah
> 30 South 2000 East, Room 307
> Salt Lake City, UT 84112-5820
> http://home.chpc.utah.edu/~cheatham/
> (801) 587-9652
> (801) 585-6208 (Fax)
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers


_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers

_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Mar 21 2016 - 16:30:04 PDT
Custom Search