Re: RC5 contest news

Charlie Hedlin (chedlin@stac.org)
Tue, 4 Mar 1997 10:34:41 -0600 (CST)


On Tue, 4 Mar 1997, Agus Budy Wuysang wrote:

> > Are you sure you have the turbo button on the PPro 200?? I have a Pentium
> > 133 with 32M and I get the following:

My AMD K5-PR75 gets it in aprox 13.7 seconds
The AMD 5x86-133 does it in 11
i486dx2-50 in 30
i486dx2-66 in about the same
now our SGI o2 does it in 7.something, our (old) IBM rs6000's in 60.
For the Linux machines (AMD and i486) I used the target build of gcc_486.
It used alot more optimizations than the origional poster used, I would
say that is much of the difference. The SGI took 15 seconds with no opts,
and the rs6000's took 120.

> >
> > rc5-56-client: Performance testing with 1000000 crypts
> > rc5-56-client: Complete in 7.792 seconds. (128328.65 keys/sec)
> >
>
> PPro 200 (256k, 32M) ~ 10 secs
> P166 (1024k, 256M) ~ 19 secs
> P100 (256k, 64M) ~ 40 secs
>
> > >
> > > gcc 2.7.2.1 parameters:
> > >
> > > -O3 -m486 -pipe -fomit-frame-pointer -funroll-loops
> > > -ffast-math -fno-strength-reduce
> > >