Re: hardware vs. john the ripper - fun

From: Matt Burnett (marukka@mac.com)
Date: Fri Jan 23 2004 - 15:25:40 EST


Yes the G4 (PowerPC 74XX) and G5 (PowerPC 97X) series chips have a vector
unit. A while ago I began work on a crypt(3) Altivec core but never really
finished it. And you could more or less write a core that would execute 4
crypt(3) ciphers in parallel. If you look around Apple's site
(developer.apple.com) they have plenty of developer docs on the chips. Donšt
bother going to motorola, I ordered their free hard copy developer manuals
years ago and they never came. Although they still may offer some developer
PDFs online.

On 1/23/04 1:33 PM, "Byron Sonne" <blsonne@rogers.com> wrote:

>> Are thoses specs right? Dual 1.4 g4's getting almost 17k cps? Is this right?
>> Its almost 3x a 3.2ghz p4. I It would be interesting to see what a dual g5
>> 2ghz would get if a asm core was written for it.
>
> Especially since, as I understand it, the Gx/PPC line does
> honest-to-goodness real vector processing (is that the altivec stuff?)
> whereas the x86 arch only approximates it using the MMX instruction set,
> and not as well either. Not sure what kinda hassles would be added for
> assembly and ML in order to take advantage of the Mac CPUs special
> abilities.
>
> I suppose that is why Macs are rather nicer at high end video and audio
> tasks but whether that same advantage could be applied to crypto I'm not
> sure. I would appreciate a good schooling on this, if anyone knows
> firsthand or has some good, relevant links.
>
> If I ever get rich, my next, big treat will be a dual G5 running some
> kind of free 'nix.
>

"The ACLU (American Civil Liberties Union) is the most dangerous
organization in the world."
Bill O'Reilly on Fox News during the "No Spin Zone" on January 10th 2004

---------------------------------------------------------------------------
----------------------------------------------------------------------------



This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:53:46 EDT