[QUOTE=MisterBitcoin;437599]I wanna reserve K25 to K100 from R287. I just downloaded the data from CRUS and its now running in LLR.

As info: "Just" 222 k´s remaining.[/QUOTE]

I recommended to use BOINC (SRBase) to keep all cores busy. There is a lot of work left and I have increased the work of 50 WUs per core. Iam not happy to run only one k in a base, better to run all.

[QUOTE=rebirther;437612]I recommended to use BOINC (SRBase) to keep all cores busy. There is a lot of work left and I have increased the work of 50 WUs per core. Iam not happy to run only one k in a base, better to run all.[/QUOTE]

You might be right. And, as allready said, it might be better to run this @SRBase cause the large amount of k´s.
My mistake. (NVM: n=25000 to 25100 checked, no primes found)

BTW: Just noticed that LLR is single-threaded. :O

Maybe it´s possible that I can test the k´s 28 and 134 to n=100k. These are only a few (8k) tasks, this will take about 1-3 days.

Well, I just wanted to test somethink at by own, so that I can say "I checked k28 from R278". :)

[QUOTE=MisterBitcoin;437628]You might be right. And, as allready said, it might be better to run this @SRBase cause the large amount of k´s.
My mistake. (NVM: n=25000 to 25100 checked, no primes found)

BTW: Just noticed that LLR is single-threaded. :O

Maybe it´s possible that I can test the k´s 28 and 134 to n=100k. These are only a few (8k) tasks, this will take about 1-3 days.

Well, I just wanted to test somethink at by own, so that I can say "I checked k28 from R278". :)[/QUOTE]

I strongly discourage cherry picking, i.e. picking a small subset of k on a conjecture like this. It creates more work for Gary to maintain k tested to different n for the same conjecture and it just makes a mess of the web pages. S3 and conjectures with thousands (or more) remaining k are an exception.

If you are PRP testing on multiple cores I strongly recommend PRPNet. I don't recommend because I wrote it, but because it really does make your life easier when running on multiple cores or multiple computers. This way you could sieve all k for the conjecture to 50000 or 100000, dump the remaining candidates in a PRPNet server and run all k's to the same n efficiently, even if you decide to stop before getting to n=100000.

[QUOTE=rogue;437632]I strongly discourage cherry picking, i.e. picking a small subset of k on a conjecture like this. It creates more work for Gary to maintain k tested to different n for the same conjecture and it just makes a mess of the web pages. S3 and conjectures with thousands (or more) remaining k are an exception.

If you are PRP testing on multiple cores I strongly recommend PRPNet. I don't recommend because I wrote it, but because it really does make your life easier when running on multiple cores or multiple computers. This way you could sieve all k for the conjecture to 50000 or 100000, dump the remaining candidates in a PRPNet server and run all k's to the same n efficiently, even if you decide to stop before getting to n=100000.[/QUOTE]


Well, you might be right. It´s even a lot of work if I would sort the k´s from the big file. If Gary would allow it, I could work on an single k´ed conjecture, like R505 or R504. This might be more effictive.

[QUOTE=MisterBitcoin;437634]Well, you might be right. It´s even a lot of work if I would sort the k´s from the big file. If Gary would allow it, I could work on an single k´ed conjecture, like R505 or R504. This might be more effictive.[/QUOTE]

I responded to your PM. If you want to start a new base consider the "recommended bases and efforts" thread. As others have stated, it does make a mess of things to only run a few k's in a larger conjecture but we do make some exceptions on huge-conjectured (conjecture k>~1M) smaller bases (bases <~100).

Here, conjectures with one k remaining like these are fine. My suggestion before starting is to run a test at the lower end of your n-range and a test at the higher end of your n-range. Then average the two times, add about 25% (rough guess if the n-max / n-min is ~2, i.e. n=200K-400K), and then multiply by the number of tests. That will give you a good idea of how much CPU time it will take. Then you can divide it by your number of cores to see what your total "wall" time will be if running the cores 24x7. That should give you a good idea if you think you can stick with it.

Virtually all low-hanging fruit on the project has been done. Any work now is fairly time-consuming.

I will wait until I hear more from you before reserving anything for you.

[QUOTE=MisterBitcoin;437634]Well, you might be right. It´s even a lot of work if I would sort the k´s from the big file. If Gary would allow it, I could work on an single k´ed conjecture, like R505 or R504. This might be more effictive.[/QUOTE]

I assume that you are going off of [URL="http://www.noprimeleftbehind.net/crus/vstats_new/crus-unproven.htm"]this[/URL] to find something to work on.

If you want to help an existing subproject, there is a PRPNet port (or two) available. There is also lots of work in BOINC.

If you want to "chart your own course" and have more control over the work, I suggest the following. Find a conjecture with 2 k remaining and Tested N of 100000. Sort by ascending relative difficulty. Choose the first conjecture from the top of the list that has a sieve file on the corresponding reservation page. You might need to sieve more deeply, but it will save you a couple of steps. Using PRPNet, test what is left after sieving.

One advantage of this is that if you get bored of PRP testing, you can stop and submit the the residues and the remaining k/n for someone else to finish. I would expect that with 4 cores that one could do a range of 50,000 for one of the conjectures I suggested in less than two months. If a prime is found, there is a chance that it will make it into the Top 5000.

Reserving R270 to n=100k (50-100k) for BOINC

R355 tested to n=100k (25-100k)

33 primes found, 56 remain

19134*355^25664-1
7692*355^31565-1
46154*355^32424-1
16532*355^33123-1
42290*355^33185-1
9542*355^35429-1
38424*355^36071-1
41328*355^36494-1
40868*355^38883-1
25370*355^39419-1
42276*355^42375-1
10994*355^45021-1
42362*355^45296-1
14352*355^46805-1
1626*355^47267-1
18990*355^51365-1
22088*355^55461-1
7062*355^57062-1
45788*355^62690-1
41204*355^63342-1
13436*355^64984-1
15954*355^65319-1
31100*355^66115-1
10734*355^71316-1
44522*355^73239-1
7958*355^78596-1
32652*355^79817-1
24170*355^80670-1
33368*355^81218-1
38768*355^89019-1
33840*355^92385-1
26790*355^93371-1
19848*355^95459-1

Results emailed - Base released

Reserving R368 to n=400k (200-400k) for BOINC

R255 tested to n=10k (2.5-10k)

666 primes found, 731 remain

Results emailed - Base released

R270 tested to n=100k (50-100k)

34 primes found, 69 remain

16328*270^53068-1
10513*270^53267-1
10894*270^53498-1
13035*270^54632-1
3992*270^54902-1
19769*270^56509-1
5266*270^59396-1
15102*270^61161-1
13820*270^61616-1
9811*270^63066-1
2045*270^63131-1
3599*270^64759-1
21212*270^65247-1
6969*270^65737-1
3200*270^67129-1
6325*270^67570-1
4254*270^68117-1
15591*270^69413-1
6983*270^72405-1
7255*270^73126-1
18344*270^74431-1
11929*270^75257-1
5006*270^75637-1
8542*270^78741-1
9825*270^80094-1
4195*270^80151-1
10817*270^81410-1
2588*270^81676-1
5359*270^84972-1
15213*270^85053-1
11580*270^87864-1
4*270^89661-1
9296*270^93802-1
4573*270^99828-1

Results emailed - Base released