[RC5] keyrate stabilizing?

Dan Sugalski dan at sidhe.org
Sun Oct 21 18:56:58 EDT 2001


At 01:40 PM 10/20/2001 -0400, Christopher Hicks wrote:
>On Fri, 19 Oct 2001, Dan Sugalski wrote:
>
> > While rotates are useful in this case, and a few other isolated
> > places, this is such an edge case that it's just not worth the silicon
> > that it'd take to implement a rotate instruction. Not having one
> > simplifies the ALU a bunch, and that speeds up the core and leaves
> > more room for other things.  Even if a CPU fakes it, like the newest
> > intel and AMD chips likely do, you still get a win. (Heck, if all
> > tossing rotate out got you was another line in your L1 cache it'd be
> > worth it...)
>
>In this day and age I would think that working well on encryption would be
>more of a priority for CPU designers.

Nope. Encryption is such a very tiny part of the work most CPUs do, and in 
those rare cases where there's an appreciable amount of work that needs 
doing (large webservers with lots of https traffic, or VPN gateway boxes) 
there are custom cards available to offload the work onto if it's deemed 
necessary. (And it rarely is)

If you need lots of encryption work, you go with a custom ASIC or something 
of the sort. General-purpose CPUs are gearing more towards video, and to a 
lesser extent audio, support. Given the number of CPU-hours spent playing 
Diablo or MP3s vs doing real encryption (as opposed to dnet stuff) it makes 
sense.


					Dan

--------------------------------------"it's like this"-------------------
Dan Sugalski                          even samurai
dan at sidhe.org                         have teddy bears and even
                                      teddy bears get drunk

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list