SSE & compilers

[Pedro Díaz Jiménez]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The Intel Compiler does a pretty good job optimizing for the Pentium  family 
(It can also compile/optimize for the Itanium, but don't have a such a beast 
to test it). Right now is in beta stage and a trial-ware version can be 
downloaded from Intel's webpage (available until September)

An unfinished (finished probably this october - not too much spare time here 
:-) review of the C/C++ compiler can be found here 
http://planetcluster.org/reviews.php?op=showcontent&id=1

And in the comments there is a link to another review of both the Fortran and 
C/C++ compiler from Computational Battery

- From the tests I've made so far (little ones: mandelbrot, sorting, (alleged) 
RC4 cipher, factorials and a couple of real world apps: lame and bzip2) speed 
achievement is important in C code. You can instruct the compiler to generate 
MMX & SSE code, while deciding if you want to retain compatibility for non 
MMX/SSE capable chips. All those tests do not include an interesting feature 
of the compiler: profile based optimization (they will, someday)

Another tests not included yet in the review are my implementation of the IBM 
Mars (AES finalist) cipher (25 Mbps gcc optimized - 40 Mbps icc optimized, 
all in ECB mode) and testing with the PovRay raytracer (not significant 
improvements)

The compiler lacks of some gnu gcc extensions, making dificult/impossible to 
test some of the programs I would wanted (GnuPG and MySQL). Due to this it is 
also imposible to compile the Linux Kernel right now (Intel says it will fix 
this in the final release)

Take a look at 
http://www.planetcluster.org/storage/icc_review/ where you will find the 
results of the bechnmarks and an interesting Intel paper about the 
optimization features in their compiler



Cheers
Pedro

BTW:
Guys, there is an interesting poll at Planet Cluster (homepage). I'll 
appreciate your votes

For last, but no least, Rajkumar Buyya (IEEE TFCC) and I are preparing a 
cluster FAQ. It is unfinished, and can be found here:
http://planetcluster.org/modules.php?op=modload&name=FAQ&file=index&myfaq=yes&id_cat=2&categories=About+clusters+in+general
(yeah, I should make a link or something :-)

The FAQ is about clusters in general, not specific to Beowulf clusters.We 
will appreciate your comments and suggestions


On Tuesday 21 August 2001 13:19, Dan Kirkpatrick wrote:
> 1. What do you know about SSE ? Apparently the p III and P 4 have extra
> hardware for fp work which is ignored by current compilers but can buy big
> factors in speed...
> we can code in assembler for it pretty easily if the PIII prcocessors have
> it.  Comments?
>
> 2. Apparently there are several optimizing compilers out there (like
> portland) which do better than gcc.  Any suggestions?  Information on
> costs?
>
> Thanks!
> Dan
>
>
> =======================================================
> Dan Kirkpatrick                   dkirk at physics.syr.edu
> Computer Systems Manager
> Department of Physics
> Syracuse University, Syracuse, NY
> http://www.physics.syr.edu/help/    Fax:(315) 443-9103
> =======================================================
>
>
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit
> http://www.beowulf.org/mailman/listinfo/beowulf

- -- 

/*
 * Pedro Diaz Jimenez: pdiaz88 at terra.es, pdiaz at acm.asoc.fi.upm.es 
 *                                                  
 * GPG KeyID: E118C651                              
 * Fingerprint: 1FD9 163B 649C DDDC 422D  5E82 9EEE 777D E118 C65
 *                                                  
 * http://planetcluster.org                         
 * Clustering & H.P.C. news and documentation       
 *
 * "La sabiduria me persigue, pero yo soy mas rapido"
 */

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.4 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE7gowhnu53feEYxlERArGrAJ41vKjIrbGhkQOwYwHfAdAunpWD2QCg6QlG
Ep//EqUuRjasVdZZqHd8UfU=
=yQR+
-----END PGP SIGNATURE-----