On Core2, GMP 4.2.2 Xeon code more efficient than Core2 code

[Vincent Lefevre]

I have built GMP 4.2.2 with default options on a Xeon machine and
on an i686 Core2 machine, installed in two different directories
under NFS.

I've tested a code of mine with both the Xeon GMP library and the
Core2 GMP library (checked with ldd). My code uses only mpn_add_n
for the GMP part.

With Xeon GMP code:  25.5 seconds
With Core2 GMP code: 33.7 seconds

Has anyone else noticed a similar problem?

-- 
Vincent Lefèvre <vincent at vinc17.org> - Web: <http://www.vinc17.org/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.org/blog/>
Work: CR INRIA - computer arithmetic / Arenaire project (LIP, ENS-Lyon)