Re: AMBER: lib-4324 : UNRECOVERABLE library error

From: Thomas Cheatham <>
Date: Wed, 21 Dec 2005 20:16:22 -0700 (Mountain Standard Time)

> Now I did a hydrogen minimize using amber7, the input file is the following:
> Control section
> &cntrl
> ntpr = 5,
> cut = 10.0,
> ntb = 0,
> maxcyc = 5000,
> ncyc = 100,
> ntr = 1,
> imin = 1,
> restraint_wt=500,
> restraintmask='!.H=',
> &end
> But the system always terminated the job abnormally, the error information is :
> lib-4324 : UNRECOVERABLE library error
> Encountered during a namelist READ from unit 5
> Fortran unit 5 is connected to a sequential formatted text file: ""
> IOT Trap
> MPI: MPI_COMM_WORLD rank 0 has terminated without calling MPI_Finalize()
> MPI: aborting job
> MPI: Received signal 6

If I remember correctly, the "restaintmask" commands replacing the archaic
GROUP input were not added until AMBER 8.0. What complicates diagnosing
your particular problem is that the error message is not very instructive.

Without knowing the history (i.e. GROUP input vs. restraintmask), I would
first try to diagnose the problem by running the code interactively on a
single processor (not using MPI) as a test.

Anytime you see the problem "UNRECOVERABLE library error ... Encountered
during a namelist READ" this normally means that some variable is in the
namelist (i.e. one of the keywords specified with the &cntrl or &wt or
other namelists) that is not recognized. It could be spelled wrong or not
present in that version of AMBER. This problem can be particularly acute
if you are using inputs from amber6->8 with sander versions that do not
match as many variables changed or were deprecated. Some
compilers/machines are graceful with respect to namelist errors; most are
not so in normal usage you will have to systematically check.

I believe however in this case that it is the restraintmask and related
variables; for AMBER 7.0, see the GROUP input specification in the
appendix and follow this for setting atoms to be restrained when NTR=1.

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Thu Dec 22 2005 - 03:53:01 PST
Custom Search