Hi David,
The issue went away after the package was compiled with a different version of GCC.
One thing I ran into is that, with CUDA and MPI, all tests passed with DO_PARALLEL="mpirun -np 1", but failed with np > 1. I checked into an output file. The numbers in this file differ a lot form the standard.
In addition, as I asked in the previous email, the file " mm_pbsa_nabnmode" is built in $AMBERHOME/bin after "configure gnu" and "make install", but it disappears after "configure -mpi gnu" and "make install". Is this the way should be?
Thanks,
Huian
-----Original Message-----
From: David A Case [mailto:david.case.rutgers.edu]
Sent: Monday, June 06, 2016 11:06 AM
To: AMBER Mailing List
Subject: Re: [AMBER] Issue of AMBER 2016 with CUDA
On Mon, Jun 06, 2016, Huian Li wrote:
>
> I have built Amber 2016 with CUDA, but "make test" gives me the "error" below consistently for every tests:
>
> possible FAILURE: check trpcage_md.out.dif
Did you examine this file? Does it look like roundoff or like more serious errors? (Other messages can generally be ignored.)
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jun 07 2016 - 08:30:02 PDT
