Thanks Tim that did it. It was the compile options I had wrong
Owen
At 02:39 24/02/2005, Tim Prince wrote:
>At 09:24 AM 2/23/2005, Owen Brazell wrote:
>
>>We have a strange behaviour when running our code on AMD Opterons.
>>
>>The code is linked with lam 6.5.9 on a 32 bit RHEL 3 system and works
>>fine on those system types. When we try and run it on an AMD system which
>>is x86_64 RHEL 3 update 3. We obviously have to build 6.5.9 for that
>>system. lamboot etc works OK as does the mpirun. However the process's
>>never talk to each other. is there a special way I need to build 6.5.9 on
>>the AMD systems to allow this to work ?
>>
>>SPAM: Spamnix identified this message as spam.
>>SPAM: Score: 51
>
>If you are running the 32-bit binary, you must build lam on the x86-64
>system with consistent 32-bit options, and keep it segregated from 64-bit
>builds e.g.
>../configure --with-cc='gcc -m32 --with-cxx='g++ -m32' --with-fc=<your
>32-bit Fortran compiler> --prefix=/usr/local/lam/32
>
>then of course set your PATH consistent with the lam installation
>prefix. In order to enable this, you must install all the "legacy" and
>"compatibility" software development and library support options of EL3.
>
>If this is not the situation you face, please explain.
>
>I have been Googling in vain for an explanation of x86-64 32- vs 64-bit
>software installation which I could furnish as an preliminary explanation
>to those who are new to x86-64, and I have run into a lack of interest on
>the part of my colleagues toward vetting my own effort.
>
>Don't know what you did to taunt the spam filters. html, whatever. Best
>assure that your postings aren't trapped by them.
>
>
>Tim Prince
>_______________________________________________
>This list is archived at http://www.lam-mpi.org/MailArchives/lam/
Owen Brazell
Systems and Deployment manager
Abingdon Technology Center
Lambourn Court,
Wyndyke Furlong
Abingdon
Oxfordshire
OX14 1UJ, U.K.
Tel: 01235 857930
Fax: 01235 535565
Mobile 07715 117178
|
