On Tue, Feb 22, 2005 at 02:02:39PM -0500, Peter Shenkin wrote:
> Does anyone have any data (anecdotal or otherwise) about
> relative speeds of 32- and 64-bit executables (both compiled
> with optimization using Intel compilers) on the new Intel
> Nocona processors?
The Polyhedron website has 32-bit and 64-bit results for Opteron, for
the PGI and Intel compilers. You can't make a fair comparison with
PathScale's compiler as in general we haven't worked hard enough on
our 32-bit code. However, it is the case that our SPECfp and SPECint
submissions do have some 32-bit runs, because in some cases the
increased pointer size of 64-bit mode outweighs the advantage of more
registers.
I would expect Nocona to be pretty similar.
-- greg
|