Questions and stuff

[kar_bon]

Yep, no special reason.

It's the smallest of the 64 k-values of the Riesel Problem without any prime found for n<3.3M so far.

This k-value (and two others) were done upto n=5M! For more details see here.

[Flatlander]

Ahh yes. The Riesel Problem. It's all coming back to me.

[Flatlander]

Now that GPU sieving has started, will some of the n<1m stuff be further sieved?
I'm guessing there is no hurry with the sieving k400-1000 n>1m, and there is still a ton of LLRing for n<1m that could benefit from a deeper sieve.

[mdettweiler]

Originally Posted by Flatlander

Now that GPU sieving has started, will some of the n<1m stuff be further sieved?
I'm guessing there is no hurry with the sieving k400-1000 n>1m, and there is still a ton of LLRing for n<1m that could benefit from a deeper sieve.

Possibly. The n<1M stuff is already at CPU-optimal depth; while it could be useful to get it to GPU-optimal depth, the n>1M range is more of a priority since we've already used up much of the n<1M sieve file. In the case of k=400-600 (the 5th Drive), we'll be up to 1M probably within a few months.

One range that could possibly benefit from being extended to GPU-optimal depth is k=2000-3400 for n<1M. However, it's still up in the air whether we'll be doing much work on that past the end of the 11th Drive at n=650K (depending on how NPLB's combined firepower looks when we get to that point), so it may or may not be worthwhile.

[Flatlander]

Do you think one more week-long rally will be sufficient to get the 5th drive to 1m by year end? It's getting a bit tight to squeeze two in.

[kar_bon]

Originally Posted by Flatlander

Do you think one more week-long rally will be sufficient to get the 5th drive to 1m by year end? It's getting a bit tight to squeeze two in.

There're ~320,000 candidates left for n<1M on drive #5 and only 84 days left until end of year.

So ~3800 candidates per day to test but only ~500 tests/day currently done.

Assumption (~130,000 paris done in last rally):
One week rally with 100,000 pairs done -> 220,000 pairs left for 77 days -> ~2900 pairs/day left
Two week rally with 200,000 pairs done -> 120,000 pairs left for 70 days -> ~1700 pairs/day left

Both left much more pairs/day than done currently.

It's a hard way to reach this goal!

[TimSorbet]

We could hold a different sort of rally: an extended-length, "finish drive 5 before 2011" drive. It would start as soon as it's set up, and only stop when drive 5 is finished or Jan 1, 2011 comes without finishing it. I wouldn't expect quite as much attention as a normal drive (e.g. not ~14286 per day, but ~3800? perhaps ), but it still might be successful. It would be a rally with a specific goal, instead of just a rally to get some crunching done quickly. Something like this rally. It's an idea.

[mdettweiler]

Originally Posted by Mini-Geek

We could hold a different sort of rally: an extended-length, "finish drive 5 before 2011" drive. It would start as soon as it's set up, and only stop when drive 5 is finished or Jan 1, 2011 comes without finishing it. I wouldn't expect quite as much attention as a normal drive (e.g. not ~14286 per day, but ~3800? perhaps ), but it still might be successful. It would be a rally with a specific goal, instead of just a rally to get some crunching done quickly. Something like this rally. It's an idea.

Indeed, interesting idea.

However, I think the first thing we need to do is get going with another "regular" weeklong rally. It's high time for that, what with 2 months or so having elapsed since the last one. Gary, what do you say we do one the week of October 22-29? That gives us the usual 2 weeks to advertise the rally. If you're available then, we should definitely do it.

Then, if the drive is not finished by December 1, I think we should launch right into a "finish the drive by year end or bust" rally as Mini-Geek suggests. Even if the deadline on the drive looks OK at that point, it would still be fun to have it come to a roaring finish in a rally (rather than just quietly fizzling out under normal crunching load).

If we get the 5th Drive done early enough, we might have time to squeeze in another quick goal...say, getting Doublecheck Drive #1 done by year's end. We were supposed to finish that one last year but it kind of got delayed since nobody's been working on it.

[gd_barnes]

Yes definitely, time for a weeklong rally again.

I'll be out of town from approximately Oct. 18th thru 26th but that shouldn't be a problem unless there is a power blip right before the rally starts. I'll have a friend on call in case that happens. Alternatively we could do one from Oct. 27th to Nov. 3rd.

Yes, I agree and like the idea of a "finish the drive" rally in Dec.

Max, please check with Lennart and Primegrid to make sure that we don't overlap one of their rallies.

[mdettweiler]

Originally Posted by gd_barnes

Yes definitely, time for a weeklong rally again.

I'll be out of town from approximately Oct. 18th thru 26th but that shouldn't be a problem unless there is a power blip right before the rally starts. I'll have a friend on call in case that happens. Alternatively we could do one from Oct. 27th to Nov. 3rd.

Yes, I agree and like the idea of a "finish the drive" rally in Dec.

Max, please check with Lennart and Primegrid to make sure that we don't overlap one of their rallies.

PrimeGrid schedules their rallies ahead of time at the beginning of the year; the list for 2010 can be found at http://www.primegrid.com/challenge/challenge.php.

Either possibility that we've discussed (Oct. 18-26 or Oct. 27-Nov. 3) would be sufficiently far away from the nearest PrimeGrid rally (17 Nov, 24 hours long).

[gd_barnes]

Let's do one Oct. 27th to Nov. 3rd. Does that work for everyone? :-)