Huhu!
We're using Amber 11 (including all bugfixes up to last week) and
observed a compiler-dependent error while parsing a DISANG file in
sander. For ICC 12.1, things work properly, while for GCC 4.6.2, sander
crashes. Details below.
We've assembled a small test case. It is performing a minimization of
some system with sander, using this input (min2.in):
> &cntrl
> imin = 1,
> maxcyc = 6000,
> ncyc = 3000,
> ntb = 1,
> ntr = 0,
> cut = 8,
> nmropt = 1,
> &end
> &wt
> type ='END',
> &end
> LISTOUT=POUT
> DISANG=../rest
The `rest` file is attached.
I'm running the minimization using the following command:
> sander -O -i ../min2.in -o min2.out -p ../top.top -c ../min1.crd -r min2.rst
Result on Amber 11 compiled with GCC 4.6.2 on CentOS 5.7
========================================================
Sander crashes "immediately".. presumably while parsing the `rest` file.
min2.out:
> [...]
> Here is the input file:
> [...]
> --------------------------------------------------------------------------------
> 1. RESOURCE USE:
> --------------------------------------------------------------------------------
>
> | Flags:
> getting new box info from bottom of inpcrd
> | INFO: Old style inpcrd file read
>
> | peek_ewald_inpcrd: Box info found
> |Largest sphere to fit in unit cell has radius = 21.977
> Error: Invalid atom or grouping specified in restraint.
> restraint = "torsion [:70.C1, :70.O5, :70.C5, :70.C4]"
Result on Amber 11 compiled with ICC 12.1 on Ubuntu 10.04
=========================================================
The minimization works.
Final remarks
=============
I've compiled both of these Amber environments and also performed the
automatic tests in $AMBERHOME/test. As far as I remember, this problem
has not been catched by an Amber test case.
I have to add that although I cannot provide a repro, our group observed
this problem already with Amber 10 and other (older) compilers. There, I
have been told that the occurrence of the error depended on
- the operating system or
- the hardware
but definitely not on the executable, because it was always the same.
However, with Amber 11, the sander binary created with ICC 12.1 seems to
work fine on each system I could test it on, while the sander binary
created with GCC 4.6.2 seems to fail on all of these systems.
I hope that this problem can be easily tracked down. If I can be of any
further help, please let me know what I can do.
Thanks,
Jan-Philip
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
- text/plain attachment: rest
Received on Fri Jan 27 2012 - 06:00:03 PST
