Skip to Content.
Sympa Menu

forum - Re: Re: Re: Re: Re: [abinit-forum] A problem for ABINIT-5.1.4

forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

Re: Re: Re: Re: Re: [abinit-forum] A problem for ABINIT-5.1.4


Chronological Thread 
  • From: "Anglade Pierre-Matthieu" <anglade@gmail.com>
  • To: forum@abinit.org
  • Subject: Re: Re: Re: Re: Re: [abinit-forum] A problem for ABINIT-5.1.4
  • Date: Wed, 13 Sep 2006 11:22:44 +0200
  • Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=BwrPhjfxFx9psOKLBa54FGD/P+qfxk0vjgi0pSpQylYTpua+Jx/mtAqAdFT3jiS7PZP74v75xvyFryLctpk8irjY645AuHfPihltYeigujdi1lkb/Jl6EWsHwy+bSbLqdlemW9+qISlyw2oLJ/N4ZVN4ISMFPKxr/sZheMZs14U=

The IBM xlf compiler is quite touchy all the more when combined with
their home unix...
From your previous mails and since you have not answered my question I
assume your abinit binary don't pass the self test.

On the one hand, I will assume that you have carrefully tryed the
solutions proposed in the page you've linked; i.e. by cleaning every
object file then compiling again the whole abinit.
Then I recommand that you try the 5.2.2 version taking advantage of
the compilation sequence discussed with Deyu Lu And Yann Pouillon on
this forum. I've made the complete sequence of self test on this
version with both no and "full" optimisation and it works fine on a
machine which I guess is pretty similar to yours.

On the other hand, supposing you got a working binary and a real
memory problem, we may still glimpse a drop of hope... First you may
have a machine with a really small amount of memory . This would
result in this error message. In this case there is some options
within abinit like "mkmem" which may help reduce the memory cost of
your calculations. Second, it may happens that AIX imposes hard limits
on the ressource available to processes/users. Those limits may not be
commensurate with the true machine limits. The best way to determine
those limits is to make a small sample program where you allocate an
increasing amount of memory. Then you'll see when it stops and know
wether this limit is large enough for the calculation intended with
abinit. If for instance you find that the limit is a ridiculous 256MB
per process while your machine hold a hefty 64GB you may call again
your "machine consulter" and ask for and obtain a change on this
limits.

regards

PMA

On 9/13/06, septhiroth0629@hotmail.com <septhiroth0629@hotmail.com> wrote:
Dear Pierre-Matthieu:

I mailed the error message to the P690 machine consulter.
He told me that according to
http://www.ncsa.uiuc.edu/UserInfo/Resources/Hardware/IBMp690/FAQ/p690Jobs.html#memory_limits_batch
We can solve the problem by add -bmaxstack:0x80000000 -bmaxdata:0x80000000
or -q64 when we compile.
But it still had the same problem.

W.C Lo



--
Pierre-Matthieu Anglade



Archive powered by MHonArc 2.6.16.

Top of Page