[Pw_forum] ATLAS DSYGVX problem + mailing list question

[Riccardo Di Meo]

Dear all,

I have compiled pw.x and linked it against the last version of the 
ATLAS/LAPACK packages from netlib. On a specific system, PW (Atlas 
actually...) complains:

 ** On entry to DSYGVX parameter number 20 had an illegal value

This is quite strange however, since with another input (showing the 
same &CONTROL/&SYSTEM/&ELECTRONS cards but a different molecule) the 
executable worked fine for hours without any significant difference 
between the output of the ATLAS executable and the vanilla one. As you 
will notice, the offending system is quite tame.

Is this a known problem? I found a warning about a similar issue in the  
QE user_guide for the AIX architecture, but I don't think it applies to 
my case very well.

The code was compiled with gcc4.4 on a 64bit AMD system, and linked against:

atlas3.6.0
lapack-3.2.1
fftw-2.1.5
openmpi-1.4.2

Uname reports:

Linux hostname 2.6.18-194.el5 #1 SMP Fri Apr 2 14:58:14 EDT 2010 x86_64 
x86_64 x86_64 GNU/Linux

The offending input and resulting output are in attachment.

Thank you for you attention,
Best regards,
Riccardo Di Meo

PS: is there a way to search in the content of the pw_forum archive? I 
found just one page with the backed up traffic of the archive, but no 
mean to look into the files online.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: water.input
Url: http://www.democritos.it/pipermail/pw_forum/attachments/20100525/06d5b688/attachment.asc 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: water.output
Url: http://www.democritos.it/pipermail/pw_forum/attachments/20100525/06d5b688/attachment.txt