[RC5] RC5-64 HAS BEEN SOLVED
bwilson at distributed.net
Fri Sep 27 00:36:32 EDT 2002
-----BEGIN PGP SIGNED MESSAGE-----
| > Not only are we looking forward to moving on to RC5-72 but
| > we're
| I would personally not participate in this project, since I
| think it will
| take much longer than RC5-64.
Actually, when you take our current keyrate and project it onto
RC5-72, the estimate is lower than our estimate of the time to
complete RC5-64 when we started that project.
Sorry, I don't have exact numbers here on the train as I'm composing
this. I believe in 1997 we thought it would take about 20 years to
do RC5-64. Today we think RC5-72 will take about 10-12 years. Both
estimates are based on 0 growth over our current keyrate at the time
of the estimate. Faster/stronger processors, optimized cores and
greater participation will all tend to shorten this number over time.
Yes, these numbers take into account that RC5-72 has 256 times as
much work to be done as RC5-64, and that RC5-72 will require 50% more
registers than RC5-64 (three registers instead of two). Our coders
tell us that on some platforms, the core will achieve about 99% of
the performance of the RC5-64 core! (Go coders!)
I can't give you a release date for the new clients (I don't know
myself), but I can tell you that I heard the coders talking about
having testable clients, and are working on putting together a test
network. The testing could turn something up that sends us back to
the drawing boards (the nature of coding), but we might also have
beta clients before very long.
We know not everyone is excited about OGR, and that's fine - we plan
to provide alternatives. Getting RC5-72 operational is our highest
priority - we're working very hard to get the new clients and proxies
out as soon as humanly possible without compromising the results. As
Nugget hinted in the announcement, we're also in the process of
investigating cores for new, non-OGR, non-RC5 projects. Naturally,
these developments were put on hold when we found out about the RC5
win. RC5 has been our mainstay for years, and a major draw for new
participants (especially the stats-hungry kind), so we're not likely
to drop it.
We've also been working on updates to the OGR core. Our best brains
tell us that the net result of the problems in the OGR core don't
mean we have incorrect results, only incomplete results. The core is
capable of correctly detecting the shortest ruler, but not of
accurately counting the total number of nodes in each stub. Stubs
over a certain length would be discarded without being counted. So
while we can be reasonably sure of finding a shorter ruler if it
exists, we can't be sure of the counts for all stubs. This means if
someone else calculates the same stubs using a different method, they
won't get the same totals, calling our results into question. (I may
be oversimplifying here - this is my best understanding of the issues
- - perhaps someone else in DCTI can jump in with corrections.)
This is a very exciting time for us, and even a little scary - we
want the new RC5 running as badly as you do! Stick with us, and we
promise you won't be disappointed.
Bruce Wilson <bwilson at distributed.net>
PGP KeyID: 5430B995, http://www.toomuchblue.com/
Build a man a fire and he'll be warm for a day.
Set a man on fire, he'll be warm for the rest of his life.
-----BEGIN PGP SIGNATURE-----
Version: PGP 7.0.4
-----END PGP SIGNATURE-----
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest
More information about the rc5