[geeks] Itanium 32 bit performace.... hahahaha

[jdboyd at jdboyd.zill.net]

On Tue, Dec 09, 2003 at 12:55:36PM -0500, Phil Stracchino wrote:

> Also take into account that the P4 design is reportedly optimized for
> certain types of operations (though no-one outside Intel seems to be
> really clear on what they are), and as a general-purpose processor, it
> sucks.  Clock for clock, a P4 is slower than a P3, and a P3 is slower
> than an Athlon.

Intel seems to make it fairly clear that it is optimized for integer
operations, and streaming interger and 32bit single precision floating
point math.

So, if you can prevent cache misses, and you have a task that calls for
those sorts of operations, a P4 should theoretically scream. And indeed,
it seems to in certain circumstances.  The secret seems to be to use
Intel supplied libraries at as high a level possible, and as much as
possible, along with also using their compilers if possible.

Of course, that also seems to be the secret to getting the most
performances from Macs, SGIs, and Suns.  However, those three seem more
balanced, and more ready to perform well without as much optimization
work.