TF Limits to Release At

[chalsall]

Originally Posted by davieddy

Your 1) merely justifies my assertion that the 200 new LL assignments per day can comfortably be GPUed to 72, with something (albeit insufficient) to spare for the ~1000/day (regrettable) expiries/returnees. So settle for 70 bits ATM.

With all due respect, I don't understand why you are so fixated on "new" assignments vs. "recycled" (read: expired). At the end of the day, all of these have not had a LL test completed, and need to be worked.

Originally Posted by davieddy

For 5 years I have monitored the Primenet Summary hourly. This may sound sad, but I have learned much from it, although Chalsall's spider has rendered this summary even more opaque than it used to be:(

If you want to remove my spider's impact from the data from PrimeNet hourly summary, simply subtract the totals from the Available Trial Factoring Assignments from the PrimeNet report. My report is updated in "real-time".

Originally Posted by davieddy

Big mystery: when (if ever) do his grabbed exponents get released to the masses? This is part of the counter-productive business.

It's not a "big mystery". Once TFing has been completed to 72 "bits" (and P-1 work completed if not already done) the candidates are returned to PrimeNet 50 at a time every five minutes (except at :00 and :05 after the hour). Observation for completion is done every 20 minutes (:10, :30 and :50).

The Overall System Progress report clearly shows how many have been returned to PrimeNet (9,455 in the LL range, 10,280 in the DC range as of right now). And, for the record, we are averaging over 250 a day at the moment (and Xyzzy hasn't checked in for a while...).

Originally Posted by davieddy

As I anticipated, 26000 primenet TF assignments 50-60M have ground to a halt, presumably lower than 72 bits. Time these were expired, and I would suggest Chalsall allocates them and the 58M - 60M range to GPUers.

As the G72 workers clean up the lower ranges, we will continue up the ranges and take care of the higher ones.

Have patience Grasshopper....

[diamonddave]

Originally Posted by davieddy

First of all, due respect to Chalsall for making this sort of data so easy to come by.
Your 1) merely justifies my assertion that the 200 new LL assignments per day can comfortably be GPUed to 72, with something (albeit insufficient) to spare for the ~1000/day (regrettable) expiries/returnees. So settle for 70 bits ATM.

I could have swore you were a big "depth first" apologist. Am I reading you right?

[davieddy]

Originally Posted by diamonddave

I could have swore you were a big "depth first" apologist. Am I reading you right?

Great post as usual, Dave.

I think the short answer is "You can't win 'em all".

Optimal search strategy is not simple, even when you concentrate on one goal. The case I'm talking about is another Mersenne prime before I die.

David

PS "70 bits" referred to the n times regurgitated expos, which are too numerous to handle ATM.

[davieddy]

Originally Posted by diamonddave

I could have swore you were a big "depth first" apologist. Am I reading you right?

If you have reserved say 50 expos that need TFing to 72,
you can do them whichever way you like or most efficient.

If such expos are urgent (e.g. holding up LL progress) then
don't hoard/sit on more than a month's worth of the (expletive deleted)
There are other kids on the block.

David

[diamonddave]

Originally Posted by davieddy

If you have reserved say 50 expos that need TFing to 72,
you can do them whichever way you like or most efficient.

If such expos are urgent (e.g. holding up LL progress) then
don't hoard/sit on more than a month's worth of the (expletive deleted)

David

I have 11 days worth of exponent reserved... Can't talk for the others.

FYI, by default GPU272 TF assignment expire after a month. I also don't know of a way to reserve an exponent for more then a month or to extend an assignment.

[chalsall]

Originally Posted by davieddy

If such expos are urgent (e.g. holding up LL progress) then don't hoard/sit on more than a month's worth of the (expletive deleted) There are other kids on the block.

If you actually used the system, rather than simply complain about it, you would know that the system does not even allow someone to reserve more than a month's worth of work based on their past performance.

Currently only a single person has any TF reservations more than a month old, and that's because they lost remote access to the machine doing the work, and won't be able to regain access for another day or so. Further, almost all of these are in the DC range.

[Dubslow]

@OnTopic: Chalsall, will 58.5+M be taken to 73 before release? And at what point should we just revert to using PrimeNet's TF assignments? I'd think we're almost there. George, what's the current status of PrimeNet TF assignments?

[Prime95]

Originally Posted by Dubslow

George, what's the current status of PrimeNet TF assignments?

I just changed the server to only hand out TF assignments between 58M and 60M to manual testers

[Dubslow]

Well okay then chalsall. It seems, for the next few weeks at least, that exponents >=58M should be released back to PrimeNet ASAP.

[Prime95]

Originally Posted by Dubslow

Well okay then chalsall. It seems, for the next few weeks at least, that exponents >=58M should be released back to PrimeNet ASAP.

???? The change means that 58M - 60M is pretty much completely under GPUto72 control.

[Dubslow]

Well if we can get the TF work through manual assignments, why would we need to go through GPU272? That reserves it as an LL, which is unnecessary due to your changes?

Edit: In other words, if GPU272 reserves them all, then the LL wavefront will be artificially pushed to 60M, where there's even less factoring done.