Misc. reserves/statuses/primes for k>2000 & k=1003

[PCZ]

2929*2^735531-1
Hope they keep on coming, tests are taking a lot longer now.

[gd_barnes]

Originally Posted by Flatlander

The 93 ks mentioned here:
http://mersenneforum.org/showpost.ph...7&postcount=12
are all tested from 400k to (at least) 420k.

The idea was to slowly work up to 500k but the top-5000 list is progressing much faster than I anticipated. I'll keep them reserved for now but I've put 5/6 of my cores on port 8000. Maybe I'll find a couple of quads in a skip/dumpster. lol

Don't feel like you have to keep most of your machines on port 8000. If you went to the effort to sieve these k's, then by all means, don't lose their top-5000 primes. That said, you're probably getting more "bang for your buck" on port 8000. The file is probably sieved further and the k's are smaller. The only thing is LLRnet tests slower than manual LLR so that may offset most of the advantage.

Here is an idea that I'll throw out at everyone: If you have sieved a range and then have either lost interest in it or you realize that you aren't going to be able to keep ahead of the top-5000 "curve" which is progressing extremely fast, mostly as a result of PrimeGrid (and somewhat us ), then by all means, let us know and post your sieve file here. We may or may not sieve it further and then could pick it up as a "mini-drive". This is an excellent idea that RPS has done quite frequently.

For such efforts, we'd just manually hand out files so as to not innundate the # of servers used. These would be good for new folks, people with lesser resources, and people with few connected machines.


Gary

[gd_barnes]

Originally Posted by PCZ

Found these 2 a few hours apart :)
2935*2^722817-1
2937*2^726444-1

Originally Posted by PCZ

2929*2^735531-1
Hope they keep on coming, tests are taking a lot longer now.

Very nice grouping of primes Brian! Yep, you may go up to n=800K without another prime or you could find 4-5 more before then. I've found that it's quite a mental adjustment going from testing at n=500K to testing at n>800K. For me, that "cutoff" seems to occur north of n=~850K. I lose patience and want to come back down to searching in the n=500K-700K range. Other folks "live" for those few gargantuan primes (namely n>1M) that pop out quite rarely.


Gary

[Flatlander]

Originally Posted by gd_barnes

...

Here is an idea that I'll throw out at everyone: If you have sieved a range and then have either lost interest in it or you realize that you aren't going to be able to keep ahead of the top-5000 "curve" which is progressing extremely fast, mostly as a result of PrimeGrid (and somewhat us ), then by all means, let us know and post your sieve file here. We may or may not sieve it further and then could pick it up as a "mini-drive". This is an excellent idea that RPS has done quite frequently.
Gary

Okay.

I'll zip the remains and email them to you.
File is n from 420,259 to 500,000. Sieve depth is 4T.
I was leapfrogging with 4 cores so there are gaps at:
421,000 to 421,685
425,000 to 425,386
430,000 to 430,378

iirc I've had 13 primes since n=400,000. It would be nice if a few more primes were squeezed out of the file before they became too small.

[gd_barnes]

Originally Posted by Flatlander

Okay.

I'll zip the remains and email them to you.
File is n from 420,259 to 500,000. Sieve depth is 4T.
I was leapfrogging with 4 cores so there are gaps at:
421,000 to 421,685
425,000 to 425,386
430,000 to 430,378

iirc I've had 13 primes since n=400,000. It would be nice if a few more primes were squeezed out of the file before they became too small.

A clarification is in order for the admins here (namely Karsten):

1. k-range: Chris has searched all 93 k's that includes the entire range of k=3011 to 3199 minus k=3045 and k=3135.

2. n-range: His original testing range was intended to be n=400000-500000.

3. Gap definition: By "gaps", Chris is referring to gaps in the sieve file that he sent me, not gaps in his testing range. Therefore his actual testing n-ranges for the above 93 k's are:

400000-420258
421000-421685
425000-425385
430000-430377

This means his sieve file contains the following n-ranges:
420259-420999
421686-424999
425386-429999
430378-500000

Karsten, I would suggest showing these k's as complete for n=400K-420K. I have asked Chris for results files and will forward those on to you if you would like them.

I will post a low-priority mini-drive for these k's in the next day or 2.


Gary

[Flatlander]

Originally Posted by gd_barnes

...
The only thing is LLRnet tests slower than manual LLR so that may offset most of the advantage.
...
Gary

In my experience LLR3.5 is sometimes faster than LLR3.71. It seems to be when k is above a certain size though I haven't tried to pin it down.
When I started LLRing the 93 ks from n=400,000 I tested the speed of both and found 3.5 to be quicker, so I used that.
When I tested k=125 etc with RPS, LLR3.71 was quicker. But LLR3.5 was quicker when I ran fixed n searches with large ks.

[henryzz]

Originally Posted by Flatlander

In my experience LLR3.5 is sometimes faster than LLR3.71. It seems to be when k is above a certain size though I haven't tried to pin it down.
When I started LLRing the 93 ks from n=400,000 I tested the speed of both and found 3.5 to be quicker, so I used that.
When I tested k=125 etc with RPS, LLR3.71 was quicker. But LLR3.5 was quicker when I ran fixed n searches with large ks.

probably different fft boundaries i would guess

[PCZ]

2933*2^740722-1

[PCZ]

2935*2^775357-1

Something curious.

Between 400k and 500k 5 primes
Between 500k and 600k 5 primes
Between 600k and 700k 5 primes
Between 700k and 800k 5 primes so far. uo to 776k.

[MyDogBuster]

2935*2^775357-1

Something curious.

Between 400k and 500k 5 primes
Between 500k and 600k 5 primes
Between 600k and 700k 5 primes
Between 700k and 800k 5 primes so far. uo to 776k.

That is strange, but it bodes well for the 800's and 900's.

[gd_barnes]

For lack of a better place to put this:

I'm reserving k=1003 for n=500K-800K.

This k is not in any of our drives nor is Peter Benson testing it. It is "officially" known to have been tested up to n=500K and I have results files for it. But like k=289, it is likely a complete double-check up to a very high n-range because it has primes at n=584103 and n=2076535 so it's very possible that it has been tested to n>2.1M.

It should not be too bad to test because it has such a low weight of 571. I'll put one fast core on it and use our nice file sieved to P=25T from the sieving drive.

This will mean when Peter Benson finishes k=1005-1400 (likely 1-2 months) and we finish k=1400-2000 up to n=600K (likely 1 week to finish all stragglers) that all k<2000 will be at n>=600K.

If anyone wants to test this for n=800K-1M, you can put in your reservation now and state it in this thread. If so, I would suggest waiting to start testing until the sieving drive is done. Then you can use a file sieved to P=45T.


Gary