Yes, that does stimulate a thought. Could be that the mpi libraries were
built for the sgi without the -ftz flag. The "all reduce" mpi operation is
used under mpi - that's fp math. In the normal install, I believe sgi
provides these libraries, rather than you needing to build them. That is
how you are getting mpi? If that's the case, I then have to defer the
question to Roberto, and see if maybe SGI forgot the -ftz under C (or the
gcc-equivalent) when they built the library. That would be all that makes
sense to me; the single processor and mpi versions of pmemd are both built
with the -ftz flag, and I presume we are talking one set of executable and
library images on a single altix smp here.
Regards - Bob
----- Original Message -----
From: "Jarrod Smith" <jarrod.smith.vanderbilt.edu>
To: <amber.scripps.edu>
Sent: Wednesday, December 13, 2006 5:13 PM
Subject: Re: AMBER: floating point assist faults on IA64 PMEMD 9
>
> Update: If I run a serial version of pmemd, the FP assist messages
> appear to go away. Hopefully that narrows things down somewhat.
>
> Jarrod
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Thu Dec 14 2006 - 05:00:03 PST
