On 16 Feb 2002, at 12:26, Mary Conner wrote: > Trial factoring is well ahead of LL testing, but the gap is closing. > Yesterday was the first day in a long time where the net number of > checkouts for factoring exceeded those for first time LL's. That is due > to the fact that one team is having a factoring challenge among their > membership due to begin soon, and large numbers of factoring assignments > are being checked out in preparation. It isn't a trend that is going to > persist. Ongoing, more factoring assignments have to be done than LL > assignments because factoring eliminates a substantial number of exponents > from LL testing.
Ah, but, factoring is actually moving ahead anyway in terms of CPU years required for LL to catch up - because larger exponents take more time to LL test. My impression is that the gap between the head of the factoring assignment queue and the head of the LL testing assignment queue has remained fairly steady at 4 to 5 million for at least three years. Nevertheless I've just tripled my factoring effort from a single PIII- 500 by switching one CPU on a dual processor PIII-1000 system from ECM to factoring. The other CPU is of course running LL tests. > > The situation has improved recently from two earlier surges in LL testing. > One accompanied the discovery of M39, the other was a post-Christmas surge > (GIMPS participants with shiny new fast computers replacing slower ones, > presumably). Many of those who joined up after M39 have dropped out, and > the crest of the wave of their expiries has passed (over 150 assignments > started on Nov. 14th expired in one day, for instance). For awhile, the > balance was around 300 net LL's checked out to 150 factoring assignments. > That has narrowed to around 250 to 200, and sometimes comes fairly even, > although to keep up, GIMPS needs more factoring assignments done than > LL's. Another point, surely, is that factoring assignments "suddenly slowed down" when we reached the changeover point between 65 bit and 66 bit trial factoring depth in the low 18 millions. > > The cutoffs for those computers set to "do the work that makes the most > sense" can be adjusted to put more computers on trial factoring work if LL > testing starts to overrun trial factoring. The cutoffs could also move > some computers doing first time LL onto doing double checks, although too > much would have the leading edge of DC's running into the trailing edge of > LL's, and I understand that PrimeNet has difficulty with that situation. Eh? The only PrimeNet problem I'm aware of with in this respect is that it can't cope with having LL and DC assignments for the same exponent out at the same time. > Right now DC checkouts are going very slowly, less than 100 a day most > days. Well, some of us try - personally I have several systems which would be running LL assignments if left to "makes most sense" running DCs instead. The difficulty here (it's purely a matter of perception, not a technical problem) is that, by pushing up the number of systems that get DC assignments by default, we risk alienating new users who find that they're not being given "exciting" assignments, and those with slower systems who find that even DC assignments are taking well over a month. > > Something that would likely help a lot would be to add factoring > capability and PrimeNet access code to those programs that don't yet have > it so that more computers can reasonably participate. Yes. There are also a lot of non-Intel un*x systems out there, most of which would be happier with DC than LL assignments, that we could probably attract if we had an automatic interface for obtaining assignments from PrimeNet and submitting results of completed tests. At least two reasonably efficient and reliable clients (Mlucas and Glucas) are already available for a good range of host systems; I'm sure the perceived complications of using the manual assignment forms are responsible for at least some users not participating. Regards Brian Beesley _________________________________________________________________________ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
