Re: sander_classic, erratic on IBM

From: David Case <case_at_scripps.edu>
Date: Thu 30 May 2002 07:13:44 -0700

On Thu, May 30, 2002, Giovanni Fossati wrote:
>
> First of all it decides, autonomously, to open also the mdvel file (NB:
> ntwv=0), and moreover it never writes anything in mdcrd and mden!

Weird. Assuming that ntwv really is zero at the beginning, it sounds like
something is being over-written in memory. The file open happens in
routine OUTOPN() in dynlib.f. You might put some debugging statements there
to trace what happens with the opens.

>
> I have a second question, this time about another weird behavior on a Linux
> box. I am using the Intel compiler on this one, which comes with optimized
> lapack libraries too.
> First of all when compiling the whole package I was either getting an error
> for 'multiple definitions' of 'second' (it is defined in sys.f, and in the
> Intel lapack library), or -when I was using a tweaked sys.f- and 'undefined
> symbol' error....


In this case, I would suggest just using the Fortran lapack libraries
provided (i.e. don't use the optimized library). Sander doesn't make much
use of LAPACK anyway, so there should not be much of a performance hit.

You could also replace the amber "second()" routine with a different name
whereever it occurs. We'll probably do that in the next release, to avoid
name conflicts like the one you report.

> Now, I try to run sander_classic and I get "Illegal instruction", immediately.
> After some struggle with the ddd debugger (I am not big in debugging) it seems
> that the problem has to do with -probably- "second" again.
> In fact, the debugger diagnosis is:
>
> program received signal SIGILL, Illegal instruction.
> timit (index=Cannot access memory at address 0x7fffffff
> ) at _misc_.f:98
> 98 in _misc_.f
>
> And "timit" calls "second".
> I don't get it...on at least 4 other similar Linux boxes, in all cases with
> Intel compiler, I did not have a problem.

Again, I suggest trying to not use the optimized LAPACK libraries, and see
if that helps.

...good luck...dac


-- 
==================================================================
David A. Case                     |  e-mail:      case_at_scripps.edu
Dept. of Molecular Biology, TPC15 |  fax:          +1-858-784-8896
The Scripps Research Institute    |  phone:        +1-858-784-9768
10550 N. Torrey Pines Rd.         |  home page:                   
La Jolla CA 92037  USA            |    http://www.scripps.edu/case
==================================================================
Received on Thu May 30 2002 - 07:13:44 PDT
Custom Search