I am using proth cmov for sieving. What result file should I send to SoB: factexcl.txt, factrange.txt, or both?
Printable View
I am using proth cmov for sieving. What result file should I send to SoB: factexcl.txt, factrange.txt, or both?
The only really necessary file is fact.txt
If you don't have that, it means you haven't found any unique factors yet-keep on sieving, it will appear eventually! :)
Generally, I think it's a good idea (for completeness) to include also factors from factexcl.txt and factrange.txt, but these won't save any tests...
But your factrange.txt file can help us. Please send your factrange.txt files to factrange@yahoo.com. It would be nice, but not necessary to put factrange somewhere in the subject line. Please zip it or compress it with your program of choice. We read .zip, .bz2, .bzip2, .7z, .rar, .tar, .gz, .gzip, etc, and of course, .txt. Thank you.
OK, Joe O, I will send it to you.
Joe O & vjs,
Weekly I have been sending the factrange.txt and factexl.txt to this url: http://www.seventeenorbust.com/sieve/ . The sieve results submission page, do you guys have access to those .txt files?
For your purposes do you still need those .txt files sent to: factrange@yahoo.com ?
e:)
![]()
e,
Yes we still need them set to factrange and we do apply all sob factors to the high-n dat, on a ~weekly basis, quite a few people are sending...
Any factors above 20M are actually ignored by the server, so if you are only sending factrange to sob we can't get those >20M.
You have to PM me with your e-mail I'll put you on the e-mail list, graphs stats etc etc.
Humm, have you been e-mailing those files to factrange? You might be one of the users who we don't know who you are???
If you only sent them to sob and not factrange@yahoo.com, then deleted them don't worry too much. Alot of the factors in factrange are actually excluded and duplicates. Last time I checked only about 5-10% of those were unique, I'd have to check with Joe he does a wonderful job with the db and can probably answer your question better than I can.
VJS' 5-10% may be a little low. The last two files I processed were 4% and 40%.
Input 97 factors Added 4 factors
Input 1963 factors Added 79 factors
Joe O & vjs,
I was mainly sending the factrange.txt and factexcl.txt to sob keep track of the ranges I have completed. I'll keep on sending those .txt file to sob and email those .txt to factrange@yahoo.com every couple of weeks. vjs, does it sound like a plan? As for the previous data's, it is history.
e:)
engracio,
Sounds like a plan to me!
Joe's really the person to answer question regarding where the factors are coming from.
But I do have a hacked together stat of what's been going on, I hope this one is the corrected version.
The dat %'s are a comparison to the number of k/n pairs in the orginal dat.Code:
Lower Upper k/n's k/n's Factors Found Found Found
(n>) (n<) Orginal Remain Found by 10K by 2.5T by 3T+
0 1 28187 27992 195 0 39 156
1 3 53908 53787 121 0 23 98
3 8 131984 131700 284 0 0 284
8 10 53115 52991 124 0 0 124
10 20 265330 264755 575 0 240 335
20 30 648872 311109 337763 331271 3871 2621
30 40 648663 311598 337065 330829 3604 2632
40 50 649463 312441 337022 330923 3500 2599
50 60 649117 319329 329788 318159 5780 5849
60 70 648603 320929 327674 315355 6131 6188
70 80 648590 321341 327249 310861 10282 6106
80 90 648497 320569 327928 310689 11080 6159
90 100 648923 321483 327440 310061 11187 6192
0 1 28187 27992 195 0 39 156
dat % 100 99.31 0.69 0.00 0.14 0.55
1 20 450429 449446 983 0 240 743
dat % 100 99.78 0.22 0.00 0.05 0.16
20 50 1946998 935148 1011850 993023 10975 7852
dat % 100 48.03 51.97 51.00 0.56 0.40
0 50 2479522 1466373 1013149 993023 11277 8849
dat % 100 59.14 40.86 40.05 0.45 0.36
50 100 3243730 1603651 1640079 1565125 44460 30494
dat % 100 49.44 50.56 48.25 1.37 0.94
0 100 5723252 3070024 2653228 2558148 55737 39343
dat % 100 53.64 46.36 44.70 0.97 0.69
Please note that the 10K, 2.5T, 3T+, references don't accurately point out the true p-level.
Found by 2.5T (for example) - We have not sieved to 100M with all p<2.5T, also included would be the factrange submissions as they come in, those values from the maineffort, other higher ranges which have been completed, "frodo's missed factor range, etc"...
I'd like to add at this point people should not consider sieving at higher T values our best bet effort at the moment is to continue with the main effort until the limit of proth is reached or advised otherwise by the project owners.
I think it's also important to point out the n<20M k/n pairs factors etc. The orignal number is not the "orginal number of total k/n pairs possible" it's just the number we started with thanks to the main effort. If you compare 10M<n<20M to 20M<n<30M you can see how much deeper the main effort has sieved.
Also alot of those k/n pairs eliminated <20M are actually those included from the maineffort. Those less than 1M are our effort and factrange of course. (Joe correct me here if my comments are incorrect)
699598206253927|4847*2^12764247+1
699599049182837|22699*2^15108238+1
well, everything is allright
I was just reading the thread in the main forum regarding error testing.
Do new users of the sieve need to start with a known small range to verify that their computers are working properly? Or is the math not as much a problem with sieving as it is in the PRP effort?
Stromkarl
That's a good idea, but we haven't done it. Just reserve a decent size range and "start your engines". Welcome to sieving! And yes, there is not as much a problem with sieving as in the PRP effort. Here you may just miss a factor. If you miss too many, gap analysis will find it, and someone will post about it. In PRP if you miss a prime, it will be a long time before double checking finds it.Quote:
Originally posted by Stromkarl
I was just reading the thread in the main forum regarding error testing.
Do new users of the sieve need to start with a known small range to verify that their computers are working properly? Or is the math not as much a problem with sieving as it is in the PRP effort?
Stromkarl
Doublechecking is really something that Joe, myself and a few nameless others are working on.
We are not truly doublechecking but sieving from 991<n<50M, where as the main effort is sieving from the secondpass n ~1.5M<n<20M currently. In the past several dats have been used 300K<n<3<, 3M<n<20M, and 1M<n<20M. Our large dat spans all of these dats and is about 15% slower than the current dat. In doing so our large range does infact find factors missed by previous programs and dats as well as find factors below and above.
There have been quite a few factors missed, nothing to get excited about but enough to keep Joe and I going and interested. Currently the worst thing to do would be recheck one of the old ranges using the current 1.5M<n<20M dat your chances of fiding a factor is better reserving a new range.
Welcome to sieve, there is alot to learn in this project if your interested and alot of question still remain.
I have gone to this site and it doesn't show anything for 678000-678500 yet. How often is it updated? Does it include all 78 factors I have submitted already from factexcl.txt and 1 from factrange.txt that was below 1T and the 1 I have submitted from fact.txt?Quote:
Originally posted by vjs
<snip>
If people are interested in where you and others are Mike has a fantastic page...
http://www.aooq73.dsl.pipex.com/gaps...n3_20_ps0u.htm
The above is a link shows all the gaps..
<snip>
I have also found 5 factors above 20T. I will submit them to the factrange[at]yahoo[dot]com email address when the range is done, if Joe_O and vjs still want them.
Stromkarl
Try these pages for more a more detailed look at the gaps...
http://www.aooq73.dsl.pipex.com/gaps...n3_20_p04u.htm
Are you sieving in that range? or how did you find them. Joe, myself and a few others are already working on all ranges less than 50T.Quote:
I have also found 5 factors above 20T.
Please e-mail the factors but if they are within the range of 20-23T it's already been sieved and we have found quite a few between 1M<n<20M.
If anyone is planning on resieving any range please check with us first we have a better system and my have done the range already... Thanks.
VJS, I think Stromkarl means he has found 5 factors with n>20 million in his factrange.txt :D
Ahhh...
20M 20T are totally different I hate these k,K,M,m,N,n,p,P,T,G,Y people including myself get them confused.
Yes if you found factors for numbers like
678XXXXXXXXXXXXXXXX | 67097*2^ 21,156,785+1
Yes we still need them and send them to factrange yahoo com thanks for helping out.
we also need factors like
678XXXXXXXXXXXXXXXX | 24737*2^991+1
as well so just send your entire factrange.txt file.
Sorry, brain lapse there. It is indeed 20M, not 20T. I also found one around 186k also. I submitted that one to SOB, but I will submit it to you also.
Stromkarl
I have to stop seiving. I am not sure how well I did, but will post my results here.
file name: SoB.bat <contents below>
./NbeGon_010_osx -s=SoB.dat -f=SoB.del -d=1.42 -p=630688573751303-631000000000000
file name: SoB.del <contents below>
4758824368237 | 22699*2^2421118+1
I am assuming that I did not finish my range. Please assign the remainder to another siever.
thxbai
ceselb,
In the future could you *, +, >, or note in the main thread those ranges reserved with a 991<n<50M dat please.
740000-742000 Nuri (991<n<50M dat) was omitted and it's possible that I will eventually I'll miss one of these reservations.
Examples
747000-761600 engracio (on 991>50m sob.dat)
738500-739000 Silverfish (with 991-50M dat - ETA:End of April)
Also do you have a working copy of chucks new sieve program for testing or has this effort entirely been dropped. I havn't heard anything from him this year!!!
I'd like to test it against our dat and dat's for missed factors etc. I could use the most recent build for either the 32 or 64-bit clients... if there is a difference.
If you don't want to give it out would you test a small range for me?
P.S. Sorry for messing up the thread I just want to make sure you see this post.