Re: ifort and CVF
From: andy2o (andy2O_at_hotmail.com)
Date: 09/01/04
- Next message: FCC: "Re: ftn95 with /cfpp"
- Previous message: Richard E Maine: "Re: ifort and CVF"
- In reply to: Gib Bogle: "ifort and CVF"
- Next in thread: Steve Lionel: "Re: ifort and CVF"
- Reply: Steve Lionel: "Re: ifort and CVF"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: 1 Sep 2004 08:39:30 -0700
Gib Bogle <bogle@ihug.too.much.spam.co.nz> wrote in message news:<ch408v$7s8$1@lust.ihug.co.nz>...
> The only thing that would persuade me to make the
> change is a clear performance improvement in solving flow problems using
> finite elements. Am I likely to get this?
>
> Thanks
> Gib
Intel provide a free trial of ifort available at www.intel.com, so why
not do some benchmarking of your code? That has to be the best way to
see if *your* code will benefit.
I can say is that ifort for Linux has beaten the other compilers I use
by a large margin for most of the mathematical codes I've used it for
(I've never had a chance to use CVF).
However it will also depend whether your core computation routines
(usually linear system solvers from libraries for me) will benefit.
Two examples:
1)I use GOTO's optimised BLAS library in many programs - that library
is already highly optimised and most of my code's time is spent in
calls to library routines. Because of this the surrounding code (which
is the only bit I actually compile) does little of the hard work and
differences between compilers have little effect.
2)All compilers seem to do well on my Lattice Boltzman CFD code - the
main computational routine is simple enough that they all get it
right.
HTH
Andy
- Next message: FCC: "Re: ftn95 with /cfpp"
- Previous message: Richard E Maine: "Re: ifort and CVF"
- In reply to: Gib Bogle: "ifort and CVF"
- Next in thread: Steve Lionel: "Re: ifort and CVF"
- Reply: Steve Lionel: "Re: ifort and CVF"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|