Individual-k reservations k=300-400 n=600K-1M

[Mini-Geek]

I have resumed work on k=349 and 353 at n=750K. ETA for n=800K is 5/13.

[MyDogBuster]

k369 ---- Tested to n=668K
k371 ---- Tested to n=749K
k375 ---- Tested to n=720K
k377 ---- Tested to n=727K
k379 ---- Tested to n=723K
k381 ---- Tested to n=723K
k383 ---- Tested to n=761K
k385 ---- Tested to n=749K
k391 ---- Tested to n=724K

All but k369 over n=700k Chugging right along

[gd_barnes]

Excellent. My 6 k's just started about 1-1/2 days ago. They're ranging in completion from about n=625K to 632K right now on 1 core each.

So now all k<=1001 are at n>=625K.

We're filling in those holes! :-)

[MyDogBuster]

Gary,

It sure would be nice to finish the individual k effort by year end.

Any thoughts on what would be next?

Ian

[gd_barnes]

Like David said, I'm pretty democratic on this stuff as long as it fits within the very general parameters of the direction of the project. There is k=2000-3000 up to n=1M, there's n=1M-2M on our current drives (this one would be a good starting point), there's even k>3000 if we wanted to find gobs of primes within each n-range that we search. For that matter, I could see the project ultimately filling in all k<10000 but I won't even go there right now. People could ultimately chime in on the direction of how they would like to go.

I just now got a PM from Vaughan expressing an interest in some variety with smaller and larger tests. I've referenced him here for larger tests and port 9000 for smaller tests. I told him he could take my k's from n=700K-1M or just n=900K-1M if he wants to. Regardless, if he throws a fair # of his mean machines on this effort, we should easily clear it out to n=1M by year end. I also suggested for variety that he could help us sieve. If he did that, helping you with n=1M-2M for k=300-400 would be a good place to start or...we could have him start on k=400-1001 (or k=400-2000) for n=1M-2M and then you guys could ultimately combine your sieves. But to do the larger k-range combined would cause k=300-400 to take much longer so we might want to keep them separate. If we got k=300-400 for n=1M-2M fully sieved by year end (optimum depth likely P=150T-200T), that would make things more interesting for the people who like larger tests and have more resources.


Gary

[MyDogBuster]

Sounds like a nice set of choices. I'm game for anything.

[MrOzzy]

What about Sierpinski primes for k=1-1000 or so, or is that way off track?

[Mini-Geek]

Originally Posted by gd_barnes

[bunch of stuff about raising k or n] People could ultimately chime in on the direction of how they would like to go.

[bigger bunch of stuff]


Gary

Hm...this project started to fill in the gaps in ranges, and I think that should still be our priority when possible. Are there any other gaps that we should fill in? If there's no abandoned gaps to attend to, I think we should generally keep with relatively small n's, since we're one of the very few prime-search projects that searches numbers small enough that even people with small resources (like me) will find a prime pretty soon if they stick at it. Besides, even people with larger resources, like you and Free-DC, have fun finding a bunch of primes.
I'm starting to sound a little like Gary!
In short: I vote we first fill gaps where needed, and then raise k instead of n.

[gd_barnes]

Originally Posted by MrOzzy

What about Sierpinski primes for k=1-1000 or so, or is that way off track?

That was mentioned at some point and subsequent to that, PrimeGrid took over the effort. It has since vaulted them to #3 in total primes. I believe they are testing something like k=3-1200 up to n=800K and may have extended that

Basically they either already had the idea or picked up on our idea for the Proth side. It crossed my mind to do it but ultimately I thought it was too big to do both sides for the resources that we had at the time. Had they not picked up such an effort, with our now much larger resources, I'd definitely do it.

Something that I suppose could be considered would be to do the Proth side for k=1200-2000 (assuming PrimeGrid isn't currently considering or started such an effort) but we'd really need to change our project definition to go that route and I wouldn't want our Riesel side to bog down too much as a result of it.

Reference Mini's thoughts, it's a bit of a tough balancing act. I've gotten 2-3 requests and suggestions about doing larger n-ranges, mainly of course from larger searchers. The project does kind of have a "glass cealing" of n<=1M and there really isn't a huge variety of n-ranges to search. (Currently from n=200K to n=~750K on the 6k drive.) But as he said, the design of the original project was to "fiil in" where primes have been missed so expanding upwards by k-value fits a little more the definition of the project. That said...definitions and times do change and to not change with them will cause a project to languish. I'd like to think that we could do both. Karsten is sieving k=2000-3000 for n<1M and Ian is sieving k=300-400 for n=1M-2M, both as low priority "side efforts" at this point.

At this point, we haven't even hit n=650K on all k<=1001 yet. n=650K-1M is a HUGE range...far more time-consuming than what we have tested so far so for now, we can keep all of this in the discussion and "low priority" phase.


Gary

[kar_bon]

that're my thoughts too:

no '+'-side for now. it's much work enough NPLB has reserved plus doublechecks.
another point is, there's no such collection of information of the Sierp-side like mine on Riesel!
and i'm not doing such new collection in the near future because i got work for a long time.

to complete the existing drives will take months of work.

after that (or better just before that point) an expansion upto higher n-level or k-level is the best choice for our work!

the 2000<k<3000 range is searched in these days by many contributors and again many gaps will be left there! so this could be one of our next goals.

[gd_barnes]

Ian, you can mark all of my 6 k's as complete to n=640K. Some are somewhat higher but n=640K is their minimum. No primes found since by start on them at n=620K.
When the 6th drive completes k=600-800 to n=640K, which will happen within about 1 day, all k<=1001 will be complete to n=640K!

Edited by MyDogBuster: Done