[quote=henryzz;213516]it's complex
i first inputed undersieved files to the server and several times at unknown depths(of testing) used the remove candidates with factors from sieve file feature of prpnet
i can provide the factor files i used to remove the factors
i don't think i have the first set of undersieved files
the reason that i started with undersieved files was that i expected to find some primes(and didn't find as many as i thought(wrongly)) which would speed up sieving lots
AFAIK i have provided the most sieved files that i have with all factors removed
sorry to cause confusion
any idea on how to do this in future?[/quote]
Ah, okay...no problem. Lennart did the same thing for his S25 ranges and therefore I've sent those on to Gary sorted but not checked against the original sieve file. Normally we'd prefer that factors be removed at definite n-range cutoffs so that we can match things up (to prevent server error and/or human error from causing anything to have been accidentally missed along the way) but in this case it's OK to let them slide since you guys are both pretty familiar with how this stuff works and the chance of error is rather small.

As for how to do it in the future, I can't speak for Gary, but I personally would prefer removing factors at definite cutoffs so that we can make absolutely sure that nothing is missing. Having to forego checking once in a while is OK, but definitely not ideal.

BTW, if anyone else ends up removing factors throughout like that in the future, please give me a heads-up to that effect when you post your results--it will save me a lot of time in processing to know that I don't have to even try to match up what will surely be a futile endeavor. :smile:

I cant see why it wouldn't be possible as long as you have the most sieved sieve file. What you need to do is make sure all the candidates in the sieve file have results not the other way around. I suppose your problem really is that your current script won't do that.

[quote=henryzz;213526]I cant see why it wouldn't be possible as long as you have the most sieved sieve file. What you need to do is make sure all the candidates in the sieve file have results not the other way around. I suppose your problem really is that your current script won't do that.[/quote]
Yes, you're right. I suppose I could write up a script to remove all results from a results file that aren't present in a sieve file, then check what's left with the latest sieve file. However, it does seem like a clumsy workaround to the problem, and it still doesn't solve the problem of when someone sieves throughout and removes n-ranges from the lower end of the sieve file as they're sieved to optimal or tested (as I believe may have happened with Lennart's S25 range, since the 3.6T sieve file he sent me was only for 52K-100K, while for the lower ranges he just had a 1.2T file).

I'll have to do some thinking about how to best deal with this. One possibility I've thought of is to set up a MySQL database in which to put all the results, then have scripts pull them out and verify them as needed. Even though we don't have a whole stats system set up for CRUS at this time, this would still be adequate for processing purposes. The main tricky thing is that I'd have to write a lot of scripts from scratch (since all my scripts now deal with flat text files)--but I think it might be worth it in the long run due to the much greater flexibility of a database. Once it's all set up and the scripts are written, a DB would simplify processing a great deal.

In the meantime, though, I'll go ahead and process your results without checking them. Any solution I come up with to this problem will probably not be available within the next few days. :smile:

Riesel base 1025
 

I have to return a computer (my old one at work), one in which I was using for R1025. I completed it up to n=45916 with no primes found. Here are the residues.

[quote=henryzz;213516]it's complex
i first inputed undersieved files to the server and several times at unknown depths(of testing) used the remove candidates with factors from sieve file feature of prpnet
i can provide the factor files i used to remove the factors
i don't think i have the first set of undersieved files
the reason that i started with undersieved files was that i expected to find some primes(and didn't find as many as i thought(wrongly)) which would speed up sieving lots
AFAIK i have provided the most sieved files that i have with all factors removed
sorry to cause confusion
any idea on how to do this in future?[/quote]


Hint: Don't do that! It's dangerous and cannot be subsequently verified if necessary in the future unless you keep the intial sieve file, all factors, and all results, which would be a headache worth of files to keep and match up in the future.

If you're sieving n=1K to 100K, sieve the entire thing to an optimum depth for n=1K-10K, break that off and test, remove k's from the remainder, sieve n=10K-100K to an optimum depth for n=10K-25K, break that off and test, remove k's from the remainder, and do the same for n=25K-50K and then n=50K-100K. (Even if using PRPnet, which will "remember" which k's have primes and so won't test them, you still need to remove them because otherwise, much additional sieving time is used.)

Don't just guess at an undersieve and hope for some primes. If you don't want to spend so much time sieving such a large n-range as n=1K-100K for all k's to an optimum depth for n=1K-10K at first, then sieve only n=1K-25K to an optimum depth for n=1K-10K, break that off and test and then remove k's, sieve, and test n=10K-25K. THEN do a brand new sieve for remaining k's for n=25K-100K and do the final 2 steps above for n=25K-50K and 50K-100K.

In other words, don't remove factors throughout. Pick specific break-off points.

Max, you will need to account for specific breakoff points in sieving. It's a must-have because if people aren't doing it when the high n-value to low n-value ratio is > ~3 to 1 (for anything n>~10K), they are wasting quite a bit of CPU resources. The key that I'm recommending to David here is that the breakoffs be minimized but specific; not at just random points throughout the process.

The method in the 2nd para. above is almost exactly what I do with a small exception: I script everything to n=2500, sieve n=2.5K-25K, break off n=2.5K-10K, etc. and continue as shown above. I usually stop at n=25K but if I was going to n=100K, the 2nd para. above is how I would do it; that is subsequently start a brand new sieve for n=25K-100K. IMHO, there are just too many k's that are eliminated at the very low n-ranges to justify sieving all of them for n=1K-100K at once.


Gary

[quote=rogue;213035]Primes found:

[code]
6357*928^15040-1
6567*928^15115-1
9387*928^15142-1
1103*928^15193-1
1292*928^15244-1
4271*928^15247-1
19884*928^15433-1
29661*928^15530-1
5534*928^15747-1
5127*928^15878-1
23600*928^15901-1
30956*928^15933-1
19911*928^15954-1
26217*928^15995-1
14612*928^16075-1
5756*928^16082-1
8282*928^16106-1
15663*928^16128-1
11388*928^16148-1
5966*928^16211-1
20831*928^16206-1
4853*928^16245-1
16518*928^16256-1
31133*928^16290-1
28632*928^16371-1
18173*928^16409-1
21555*928^16425-1
16043*928^16502-1
28086*928^16565-1
7001*928^17075-1
4664*928^17127-1
[/code]

I am ending my effort on this base. If anyone wants a file with the remaining candidates for n < 25000, let me know.[/quote]


Please state your exact search depth. If you'd like for the sieve file to possibly be used in the future, I'll need to post it on the pages. Otherwise I virtually guarantee that it will be forgotten. Please post it here with k's removed that already have primes and with its actual sieve depth in the file. The latter is frequently needed to see if it has been sieved to an optimum depth, which can vary widely with future software and hardware improvements.

Edit: k=4271 and 5534 already had primes at n=9557 and n=9921 respectively. So there are 29 primed k's for the range and 711 k's remaining at n=~17127.


Gary

Riesel Base 654
 

Riesel Base 654
Conjectured k = 261
Covering Set = 5, 131
Trivial Factors k == 1 mod 653(653)

Found Primes: 239k's File attached

Remaining k's: 14k's - Tested to n=25k
30*654^n-1
44*654^n-1
53*654^n-1
56*654^n-1
79*654^n-1
100*654^n-1
114*654^n-1
124*654^n-1
132*654^n-1
136*654^n-1
204*654^n-1
219*654^n-1
236*654^n-1
239*654^n-1

k=4, 9, 49, 64, 144, 169 Proven composite by partial algebraic factors

Base Released

It is with a mild sense of foreboding that announce my intention of attacking Sierpinski base 928 ([URL="http://www.mersenneforum.org/showpost.php?p=207252&postcount=214"]last attempted here[/URL])

There are 686k's remaining at n=10,000, I'm hoping to make that number a little smaller! Initial sieving has commenced, I shall post occasional updates. If people think this is foolhardy, well, that's your prerogative. If you think it's foolhardy, but wish to offer advice, please PM me :) This comes under the heading of extreme whimsy (that and wanting to stop cluttering up the forum with the 1 k-remaining reservations.)

[quote=paleseptember;213939]It is with a mild sense of foreboding that announce my intention of attacking Sierpinski base 928 ([URL="http://www.mersenneforum.org/showpost.php?p=207252&postcount=214"]last attempted here[/URL])

There are 686k's remaining at n=10,000, I'm hoping to make that number a little smaller! Initial sieving has commenced, I shall post occasional updates. If people think this is foolhardy, well, that's your prerogative. If you think it's foolhardy, but wish to offer advice, please PM me :) This comes under the heading of extreme whimsy (that and wanting to stop cluttering up the forum with the 1 k-remaining reservations.)[/quote]


No problem and it's not foolhardy at all as long as you are aware that it will likely take at least a full CPU year to finish. (rough estimate) The only recommendation that I'll give is to put at least a full quad-core on it unless you are very patient. The main thing to be aware of is that base 928 takes much longer to test at the same n-depth than bases in the 200s and 300s.

Many people on the project like to use a personal PRPnet server for this kind and scope of effort. It allows easy management of your cores. Feel free to post questions about it. Mark (Rogue) created it. The latest version seems to work quite well.

BTW, I like your 1k remaining work. That's why we have the thread. Never feel like you're cluttering up the forum with it. :smile:


Gary

[quote=gd_barnes;213957]Many people on the project like to use a personal PRPnet server for this kind and scope of effort. It allows easy management of your cores. Feel free to post questions about it. Mark (Rogue) created it. The latest version seems to work quite well.[/quote]
See [url=http://www.mersenneforum.org/showpost.php?p=209872&postcount=8]here[/url]--I have a standing offer to host private LLRnet/PRPnet servers for anyone interested at NPLB or CRUS (heck, I don't mind even if you want to load in stuff from another project). This can take quite a bit of the hassle out of running a server since I've already got the infrastructure and processes set up so that adding another server over on this end is hardly a big deal. :smile: