AMBER Archive (2006)

Subject: RE: AMBER: <kein Betreff>

From: Ross Walker (
Date: Mon Nov 13 2006 - 10:44:03 CST

Dear Michel

> recently we have compiled the multiprocessorversion of Amber8
> but unfortunately a rather strange problem has occured. As I
> am accustumed to know with Amber 7 the .out and .inf files
> are written and readable during a job.
> Surley you can influence this with ntpr and ntwx.
> But with that multiprocessorversion of Amber8, the .out and
> .inf files are written not until the job is finished. It's a
> little trouble not to know what's going on during the run;
> our boss ment, it could be a problem of a wrong adjustment
> of the buffer; perhaps files are first written until they are
> large enough to overcome the buffer capacity (because the
> .rst and .trj files are written during the jobs)

This buffering is a function of your system setup and is not easily
controlled by amber. Likely your mpi implementation is buffering all output
to try to boost performance. I would take a look at the mpi and queuing
system documentation for your system and see if you can change the buffer
sizes for I/O. However, decreasing these may cause issues with performance.

Amber does flush the i/o buffers after writing so you can't really force it
to flush them again. I suspect that your system is somehow ignoring the
flush command.

You could try closing the file in dynlib.f after the information has been
written and then opening it again append mode. This might fix it. I would
start by seeing if there are any options in your mpi implementation. You
could try writing a short hello world program in mpi and see if you get the
same behaviour here.

All the best

|\oss Walker

| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- |
| | PGP Key available on request |

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.

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to