AMBER Archive (2004)

Subject: Re: AMBER: PMEMD Problems...

From: Robert Duke (
Date: Mon Apr 05 2004 - 23:38:58 CDT

Jon -
Okay, so given that it was not the -c being -C, it is most likely something
else in the input. The fact that you get the "PMEMD Terminated Abnormally!"
message basically means that the code did a "stop" with this message
specified as opposed to a "stop" with nothing after it, ie., it called an
error exit (mexit), and this is most likely due to bad input when it happens
quickly. It is possible that you are attempting to do things that are not
supported by pmemd, so look at what the mdout output tells you. If there is
nothing there, something is odd; otherwise it is some sort of input error.
As far as using the redhat and ifc that you are using, I think they are
probably fine; I don't use 7.3 on my own systems but I believe it is what we
have installed at UNC on the myrinet cluster and blades, and it is widely
used. All the threading grief starts around redhat 8 I believe, and
certainly by redhat 9 (I have pretty much skipped 8, and have recently
tested 9 and rhel 3.1; intel and the release notes for the OS and ifc say
ominous things about threads for these systems, with suggestions for
whacking stack limits and setting thread stacksize environment variables; I
have a "unlimited" stacksize (which really is not unlimited) and as long as
I am fully updated have been fine). If you are not used to constructing
input for sander or pmemd, and you are running amber 7, you can run the
4096wat and dhfr tests under amber7/test to quickly test pmemd. Look at the
makefile there, and add a test.pmemd modelled on test.sander using these two
entries; rename pmemd to sander (alternatively, whack the run scripts in the
subdirs, but that gets to be work), and give it a go (a quick and dirty
test, 'make test.pmemd'). Under amber 8, there are more tests available,
and the test.pmemd target is already in the makefile. You definitely want
to regress pmemd and sander against a few common test cases; in the amber 7
timeframe, a few other tests that should work are under amber7/benchmarks -
dhfr (again), jac, and rt_polymerase. You should be running pmemd 3.1 if
you are not running pmemd 8.
Regards - Bob

----- Original Message -----
From: "Jon Gray" <>
To: <>
Sent: Monday, April 05, 2004 11:56 PM
Subject: Re: AMBER: PMEMD Problems...

> Bob,
> Sorry, that -C was a typo and was meant to be a lowercase -c
> To give you more detail, the system I am attempting to run PMEMD on is a
> 2GHz P4 (single processor). I compiled with ifc 7.1.008 on a redhat 7.3
> distribution. Have you seen a successful run on that version of redhat or
> would you recommend me trying redhat 9? I have also attempted an install
> on a Suse 9.0 system but could not even get amber7 to compile, so did not
> attempt to do PMEMD.
> Also, is this 'Terminated Abnormally' message normal for improper
> parameters, or could this mean the compiled binary is bad? I would like
> figure out whether this is an issue with the system and compiling of PMEMD
> (my area) or an issue with the usage and syntax with the PMEMD program
> (someone elses).
> Thanks again for your help.
> -Jon Gray
> --On Monday, April 05, 2004 5:13 PM -0400 Robert Duke
> wrote:
> > Jonathan -
> > The one thing I see right off is the use of -C to specify input
> > coordinates instead of -c. The input arguments are case sensitive. If
> > you enter "pmemd -help" you will get a list of allowed input parameters,
> > but you will have to read the manual to figure it all out. The basic
> > procedure is to start with a knowledge of sander, look at the pmemd
> > manual section for the delta's in behaviour, and have at it. PMEMD
> > should take close to the same input. If you are dealing with pmemd
> > 3.0-3.1, there is a help facility that is invoked with "pmemd -help",
> > analogous to the help facility that was in sander 6. This has been
> > dropped in both pmemd 8 and sander 8; the sander guys I presume dropped
> > it due to the headaches involved in keeping it up to date; I dropped it
> > in 8 because they did, and also it was a pain to keep up to date in
> > also. You give no info about your systems other than the compiler, so I
> > can't be much more help. If you run under mpi, you of course have to
> > invoke with an mpi-specific command rather than just the executable
> > and there are definite touchy issues involved in getting mpi stuff built
> > under linux. There has been a fairly active discussion on the amber
> > reflector as of late dealing with the problems encountered with the
> > releases of linux and the intel fortran compiler. I have recently
> > everything running with the latest redhat, fully updated (3.x rhel),
> > redhat 9 (also fully updated), ifc 7.1.038 (7.1.040 is broken, ifort 8
> > seems to be perpetually hopeless), and +/- mpich I have gotten
> > it all running on 3.2 GHz dual pentiums, but gave up on a dual opteron
> > (2.2 GHz) that was apparently overheating. You are nowhere near heating
> > problems yet, though ;-)
> > Regards - Bob Duke
> >
> > ----- Original Message -----
> > From: "Jon Gray" <>
> > To: <>
> > Sent: Monday, April 05, 2004 4:38 PM
> > Subject: AMBER: PMEMD Problems...
> >
> >
> >> Hello everyone,
> >>
> >> I have been trying to install PMEMD to be used with our existing amber7
> >> installation for a couple weeks now. At first I was having issues with
> >> version 8 of the intel fortran compiler, but after switching to an
> >> version of 7 it seemed to compile smoothly.
> >>
> >> Although it is compiled now, whenever I execute the pmemd binary file I
> >> just get:
> >>
> >> FORTRAN STOP PMEMD Terminated Abnormally!
> >>
> >> There is no output of an error file whatsoever. This happens when I
> >> execute the command by itself (although then I get an mdout error along
> >> with the Terminated Abnormally message) or when
> >> I try running something that runs fine under sander. Here is line I am
> >> attempting to execute:
> >>
> >> pmemd -O -i sander_V_PMEMD.inp -o a_m.out -p \
> >> -C a_m.V169ps.rst -r a_m.rst -ref a_m.V169ps.rst -x a_m.trj
> >>
> >> Forgive me if I misused some terminology, or this is an obvious
> >> I am not actually in this field but rather just a tech person who has
> >> been assigned this problem :)
> >>
> >> Thank you in advance.
> >>
> >> -Jonathan Gray (Carnegie Mellon University Chemistry Dept)
> >> -----------------------------------------------------------------------
> >> The AMBER Mail Reflector
> >> To post, send mail to
> >> To unsubscribe, send "unsubscribe amber" to
> >>
> >
> >
> > -----------------------------------------------------------------------
> > The AMBER Mail Reflector
> > To post, send mail to
> > To unsubscribe, send "unsubscribe amber" to
> >
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to
> To unsubscribe, send "unsubscribe amber" to

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