On Tue, 10 Sep 2002, Mark Hartner wrote:
> MPICH 1.2.4 and LAM 6.5.6 both use WORDS_BIGENDIAN
>
>
> lam_config.h does:
>
> #ifndef WORDS_BIGENDIAN
> #define WORDS_BIGENDIAN 0
> #endif
>
> and the MPE source does:
>
> #ifndef WORDS_BIGENDIAN
>
> See the problem here? For a little endian machine the LAM source #def's
> WORDS_BIGENDIAN, and then MPICH sees that it is defined and compiles for a
> big endian machine.
>
> We are using a "little endian" intel box and noticed that the MPE traces
> were complete garbage. Replacing WORDS_BIGENDIAN with MPE_WORDS_BIGENDIAN
> in the MPICH source fixed the problem and now LAM/MPE are working fine.
Grumble... It does look like LAM is stepping on MPE by doing the #ifndef
.. #define trick. It appears that you have made it work, so I obviously
don't need to suggest a quick fix.
And we'll remove the offending code for the next release of LAM/MPI.
Brian
--
Brian Barrett
LAM/MPI developer and all around nice guy
Have a LAM/MPI day: http://www.lam-mpi.org/
_______________________________________________
This list is archived at http://www.lam-mpi.org/MailArchives/lam/
|
