Re: the bitslice method

Icepick (icepick@pclink.com)
Thu, 29 May 1997 12:23:49 -0500


> From: Darrell Kindred <dkindred@cmu.edu>
> To: deschall@gatekeeper.megasoft.com
> Cc: Andy Church <achurch@dragonfire.net>
> Subject: the bitslice method
> Date: Thursday, May 29, 1997 11:23 AM

[snip]

> Now, a little bit about "bitslicing"...
>
> Andy Church writes:
> > Curiosity attack: What _is_ bitslicing, anyway?
> > What makes it faster than "regular" clients?
>
> The basic idea is that if you have a CPU that operates on
> N-bit integers, you can treat it like N processors each
> operating on single bits. So for a 32-bit CPU, you get 32
> "virtual processors", and for a 64-bit CPU, you get 64.
> The parallel machine you're simulating is of the "SIMD"
> (single instruction, multiple data) model; that is, all the
> "processors" execute the same instruction at each step, but
> they operate on different data.
>

So this begs the question: "Do the 32-bit clietnts use
bit-slicing?"

-Josh

When encryption is outlawed, bayl bhgynjf juyy unir ceuinpl.
Prove DES is weak, break it! http://www.frii.com/~rcv/deschall.htm