Thanks all who replied. The info was helpful - the polyhedron site
particularly.
Thanks
Simon
----- Original Message -----
From: "Thorsten Ohl" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Wednesday, November 07, 2001 8:30 AM
Subject: Re: Linux f90 Compiler performance
> Peter Shenkin writes:
>
> > However, NAG's performance, judging from both polyhedron
> > and your own comparison, is not great. Have you found
> > NAG's performance satisfactory in your own hands?
>
> For idiomatic Fortran90 code, NAG performs very well, compared to
> others, including native compilers for Linux. In fact, I was
> pleasantly surprised to see that Intel's compiler can beat NAG
> significantly (for the code that it can compile).
>
> > That is, when you say you "devlelop" with NAG, are you implying that
> > when you want to compile for production, you use something else?
>
> The real production is done by other people who use the code to
> perform large scale simulations. For testing correctness, speed is
> not a big issue. Moreover since our Fortran90 code solves some
> relevant cases that cannot be done (reasonably) with Fortran77, users
> don't complain too much about performance in production either.
> --
> Thorsten Ohl, Physics Department, TU Darmstadt -- [log in to unmask]
> http://heplix.ikp.physik.tu-darmstadt.de/~ohl/ [<=== PGP public key here]
|