Re: [AMBER] Could not read velocities from restart file

From: Ross Walker <>
Date: Wed, 11 Apr 2012 12:37:04 -0700

Hi David,

> I applied the bug fixes for AmberTools 1.5 and AMBER11.

Does this include the patch I provided? - Did you install this in a
different location from your original? - so we can test both with and
without the patch?

> export DO_PARALLEL='mpirun -np 4'
> >
> bash: export: `-np': not a valid identifier
> >
> bash: export: `4'': not a valid identifier

This is very strange and is coming from your shell. Are you sure you are
using the quotes correctly? - Maybe try "" instead of '' - I have a feeling
you are using the back sliding quote which is on the same key as a ~ (on a
US keyboard) - this is the WRONG quote to use. On a US keyboard use either '
or " which is on the same key next to : and ;.
> When I attempt to run the inpcrd.rst that Ross gave,
>$ mpiexec -np 8
> pmemd.MPI
> > -O -c inprcd.rst
> >
> application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0

Lets forget parallel for the moment and just focus on serial. Firstly does
the problem you reported with the restart file not being fully written occur
in serial on the CPU as well as the GPU code or do you only see it in GPU

> or just a single processor:
>$ pmemd -O -c
> inprcd.rst
> >
> > Unit 9 Error on OPEN:
> > inprcd.rst
> >
> > STOP PMEMD Terminated Abnormally!

The file is called 'inpcrd.rst' you have the c and r reversed above, not
sure if that is an actual typo in your runs or a typo in the email.

Can you try the following please if you can.

1) Run the example I gave using the original executable in serial on CPU and
verify that the restart writing issue exists. That is the restart file
during a run when you do 'ls -la' is smaller in size than the inpcrd.rst

2) If 1 shows the problem then try running again with the new executable
built from code with the patch I gave. Does it work this time?

If neither show the problem then try the GPU code and see if the issue is
restricted to just that. - Note for the serial GPU runs you might want to
reduce NTPR, NTWR and NSTLIM by an order of magnitude so you aren't waiting
forever for it to reach 10,000 steps.

All the best

|\oss Walker

| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
| | |
| Tel: +1 858 822 0854 | EMail:- |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

AMBER mailing list
Received on Wed Apr 11 2012 - 13:00:06 PDT
Custom Search