[RC5] keyrate stabilizing?

Niles Parr nparr at usa.net
Tue Oct 16 12:21:00 EDT 2001


I was thinking the latest was only slightly optimized and more work was
needed, I could be wrong.  I'm sure there are a lot of people that would
love to be running amd's and G4's but that is not the point.  The point is,
there are many people (myself included) that don't have a choice and are
trying to make the best of it.  If I could buy MP athlons for my entire
company I would.

-----Original Message-----
From: owner-rc5 at lists.distributed.net
[mailto:owner-rc5 at lists.distributed.net]On Behalf Of TD - Sales
International Holland B.V.
Sent: Tuesday, October 16, 2001 4:16 AM
To: rc5 at lists.distributed.net
Subject: Re: [RC5] keyrate stabilizing?


There is a optimized p4 core far as i know, the reason the p4 scores so bad
is because it's missing a hardware instruction called rotate. For some
reason
the people at intel suddenly decided that their cpu didn't need the x86
standard's rotate instruction anymore....... anyways, now they have to
emulate the instruction which is much slower. if you want high rc5 scores
you
better get an amd



On Monday 15 October 2001 14:20, you wrote:
> Could part of the problem be because of no optimized core for newer
> processors (mainly P-4)?  Overall CPU power is increasing but we are
unable
> to utilize it.  By now there has to be enough market penetration of the
P-4
> to make a difference .
>
> -----Original Message-----
> From: owner-rc5 at lists.distributed.net
> [mailto:owner-rc5 at lists.distributed.net]On Behalf Of Bruce Wilson
> Sent: Sunday, October 14, 2001 9:13 PM
> To: rc5 at lists.distributed.net
> Subject: RE: [RC5] keyrate stabilizing?
>
> | > We should be well into 2nd pass of OGR-25 by now.
> | > Could we please have an update like that again? daa?
> |
> | bwilson? anybody?
> |
> |
> | Even better, why can't we put it on the stats page to be automatically
> | updated? I mean, the calculations really aren't hard :)
>
> It's not a question of calculations, it's a question of the volume of
> data.
>
> Every returned stub is stored as a line of text, occupying (perhaps) 50
> bytes.  That's probably low, but it makes calculation easier.
>
> In order to give accurate counts, we currently have to reload all the
> logs since the start of the OGR-(N) project into a database, figure out
> which ID first turned in a stub (pass 1), then figure out if anyone else
> turned in that same stub (pass 2), and if so, did they return the same
> nodecount.
>
> Along the way, we have to take into account retirements.  Any two
> participants who are related by retirement are considered to be the same
> ID.  This means up to 11 emails have to be checked against the person
> who first submitted the record.
>
> We have received a ton of duplicate work on OGR.  Apparently, some
> people think they can get stats credit for submitting the same work more
> than once.  At least temporarily, this is so.  When we discover someone
> is doing this, we usually just take them out of stats completely, the
> same as if they were installing trojans, or had made unauthorized
> installs.
>
> Duplicate or not, we still have to sift through all that work to find
> out what work has or hasn't been processed.  We have gigabytes of data
> to be processed, and it's not currently automated.  We have a script,
> but it's a human script, not a shell script.
>
> I'm not aware of what has been happening internally lately, so I can't
> give you a status update on OGR-25 progress stats.  I remember Daa
> saying back in August he might have time in October.
>
> Someone on this thread commented that we (d.net staff) aren't doing
> enough to push the project forward.  Mea culpa.  I have debated recently
> whether I would be better off taking my name off the list (thereby not
> implying that I am actively involved when I'm not), or staying on to
> avoid the "sky is falling" predictions, but still catching the blame.
> It's discouraging even to me how little progress is visible to the
> outside, and how long it takes to get anywhere.  I truly believe we
> could be going great places, but I can't do it alone, and I can't force
> anyone else to do anything at all.
>
> I do know some client development is in progress, specifically on some
> corrections to OGR.  I also know that we are acutely aware that "whoops,
> we're starting over" is an unacceptable answer on so many levels.  Yes,
> there are some bugs in the OGR core.  Yes, we're working on a
> correction.  Yes, we expect to save all valid work.  We're not sure how
> much of it will be valid, but indications point to saving nearly all of
> it.  That's as specific as I can get - ask a coder for more detail.
>
> And yes, stats are overdue for an update.  (It has nothing to do with
> knowing or not knowing how many nodes there are - we have been reporting
> progress as [stubs accurately completed]/[total stubs] basically
> forever).
>
> I read every message on this list, and it really is discouraging
> sometimes how easily people jump to a conspiracy theory.  We're not a
> corporation (though we are incorporated), we don't gain anything by the
> success of RC5-64 or OGR-(N) (except the good press to support a new
> project), we don't have any agendas that aren't clearly articulated on
> our website.  We're just a group of guys who believe in distributed
> computing, and who want to make it work.  We all wish we could do more.
>
> __
> Bruce Wilson <bwilson at distributed.net>
> PGP KeyID: 5430B995, http://www.toomuchblue.com/
>
> "Quidquid latine dictum sit, altum viditur."
> (Whatever is said in Latin sounds profound.)
>
>
> --
> To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
> rc5-digest subscribers replace rc5 with rc5-digest
>
>
>
> --
> To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
> rc5-digest subscribers replace rc5 with rc5-digest
--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



--
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