Re: Problems with the use of Feff82 on a Silicon Graphics machine

["John J. Rehr" <jjr@leonardo.phys.washington.edu>]

Hi Jean-Claude,

 Thanks very much for your msg.  FEFF8 was developed on SGI's so
there's a good chance we might have the answer this time.

  Did you also set the stacksize? SGI's sometimes have a limited stack
(64000) which is insufficient for big clusters. The command 

limit

should tell you the various limits. To override, use the following command
which should be enough for 200 atoms. 

limit stacksize 170000

  Let us know if this doesn't solve the problem on the SGIs.
  It's interesting that the -static flag also works, but we haven't
had to use it so far.

 By the way, there is also a way to run the 4 processors in parallel! To do
that you need to go to the Cool Software site
http://www.sgi.com/products/evaluation/
and download and installl the Array 3.4 and Message-Passing toolkit 1.4
packages. Let me know if this interests you and I'll send more details.

  Cheers,
  John


>Our lab has recently acquired a 4-processors Silicon Graphics
>computer. I have compiled feff82 on this machine, using the flag
>given in Appendix B
> f77 -Ofast -LNO:opt=0 feff82.f -o feff8.2
>A short time after the begining of the execution on the SG machine,
>I get a message "segmentation fault".  However, when I add the
>flag -static, this problem disappears.
>I think that problems occur when calling the fms subroutine.
>Can you give me a piece of  advice ?