The ABINIT Users Mailing List ( CLOSED )

[Samuel Sarholz <sarholz@rz.rwth-aachen.de>]

Hi Alain,

I used the option "-fp-model precise" and similar ones for the other 
compilers.
The man page of the Intel compiler states that this only "Enables 
value-safe optimizations on floating-point data"

Regarding the speed so far didn't do a fair comparison, because I had 
only the Intel compiler optimize for Xeon other with the others I 
optimized for Opteron, but ran on Xeon.
Speed wise it is Intel faster as Sun faster as GNU.
My experience is that it is very code dependent with compiler is faster.

The main reason why I tried different compiler was that I have a problem 
with the Intel compiler and -O3. I get "wrong" results and core dumps.
(I guess I will do some more tests, it looks like an Intel compiler 
problem to me)

config/optflags locks like someone spend some serious thought on
which flags to use. Unfortunately I am used that many project don't 
optimize much (and it looks that nobody thinks about it) and usually put 
in highest optimization while turning off/down floating point 
optimization. So far that worked well.

Anyway thanks again for all the helpful hints.

Best regards
Samuel

--
Dipl.-Inform. Samuel Sarholz    -   High Performance Computing,
RWTH Aachen University, Center for Computing and Communication,
Seffenter Weg 23, 52074 Aachen (Germany), Office: 2.13
Tel: +49 241/80-24915 - Fax: +49 241/80-22134
mailto:sarholz@rz.rwth-aachen.de www.rz.rwth-aachen.de

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature