Aaron -
In our (at least some subset of the amber dev group) experience to date, the
intel compilers have had superior performance and usability compared to the
stuff from PGI; PGI may be getting better, I have not used their stuff that
much in the last year myself, and I would think they would be highly
motivated to fix problems. Intel compilers are NOT free for everybody, so
beware of the conditions on the licenses (we buy copies for amber dev, for
instance). There is the occasional intel compiler bug, but the ia32 stuff
is reasonably solid; the itanium 2 stuff has been flakey at points but
presumably also is getting better over the last year. I have also had good
experience with Pathway on opterons, but have not recently done comparisons
there either (very usable, performance in the same ballpark with intel). I
did have users (in Europe) complain to me about a year ago that Pathscale
was giving them grief over licensing issues or some such.
Regards - Bob Duke
----- Original Message -----
From: "Aaron Greenwood" <agreenwo.uci.edu>
To: <amber.scripps.edu>
Sent: Tuesday, October 23, 2007 1:03 PM
Subject: AMBER: Portland Group vs. Intel Compilers.
>
> Even through the Intel C and FORTRAN compilers are without cost there
> seems to be a lot more problems using them to compile AMBER than using
> Portland Group compilers. I am not trying to start a religious war but
> would appreciate comments on the pros and cons of each compiler.
> Best regards,
>
> Aaron
>
> --
>
> Aaron Greenwood
> Systems Administrator
> McPherson/Poulos Groups
> University of California
> Irvine, CA 92697
>
> Office: 9327 Battle Creek Street NW
> Albuquerque, NM 87114
> Home Tele: 505.890.8218
> Cell: 951.818.0594
> agreenwo.uci.edu
> greenwood.abqnorthwest.com
> http://crystal.bio.uci.edu/agreenwo
>
>
> -----------------------------------------------------------------------
> 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 Wed Oct 24 2007 - 06:07:51 PDT
