mersenneforum.org Queue management for 16e queue
 Register FAQ Search Today's Posts Mark Forums Read

 2021-10-07, 17:54 #78 VBCurtis     "Curtis" Feb 2005 Riverside, CA 117178 Posts That's funny. I don't pay attention to the form of the Kosta numbers so I didn't give a different poly any consideration. I'll test-sieve this sextic tomorrow and send it to the 15e queue; it will fit there without trouble, and that queue is nearly dry anyway.
2021-10-07, 19:38   #79
swellman

Jun 2012

24·7·29 Posts

Quote:
 Originally Posted by VBCurtis That's funny. I don't pay attention to the form of the Kosta numbers so I didn't give a different poly any consideration. I'll test-sieve this sextic tomorrow and send it to the 15e queue; it will fit there without trouble, and that queue is nearly dry anyway.
Thanks for test sieving (again). Are you certain it will fit on 15e? If so I can enqueue the new job file with a new name once you characterize it.

Otherwise I can just modify the existing poly and sieving range if it must remain here on f_small.

2021-10-07, 20:50   #80
jyb

Aug 2005
Seattle, WA

110111101102 Posts

Quote:
 Originally Posted by swellman Thanks for test sieving (again). Are you certain it will fit on 15e? If so I can enqueue the new job file with a new name once you characterize it. Otherwise I can just modify the existing poly and sieving range if it must remain here on f_small.
You may want to run the skew value through cownoise, since it apparently comes up with values different from the standard guideline (which would have given 1.000 for the original poly as well).

2021-10-07, 23:18   #81
charybdis

Apr 2020

547 Posts

Quote:
 Originally Posted by VBCurtis That's funny. I don't pay attention to the form of the Kosta numbers so I didn't give a different poly any consideration.
Same here - well spotted Jon. Didn't even look at the job name; the "35" should have set alarm bells ringing.

The cownoise poly generator appears to automatically generate quartics for exponents divisible by 5, even if they are also divisible by 7, 11 or 13.

Quote:
 Originally Posted by jyb You may want to run the skew value through cownoise, since it apparently comes up with values different from the standard guideline (which would have given 1.000 for the original poly as well).
Cownoise suggests a skew of 1.38. (not 1 because while the algebraic side is symmetrical, the rational side isn't!)

2021-10-08, 21:13   #82
VBCurtis

"Curtis"
Feb 2005
Riverside, CA

11·461 Posts

Quote:
 Originally Posted by swellman Thanks for test sieving (again). Are you certain it will fit on 15e? If so I can enqueue the new job file with a new name once you characterize it.
It'll be on 15e for sure. First-round of test-sieve indicates something like Q from 50-400M will do the job.
sec/rel is a tiny bit faster on 15e than 16f, at Q=200M. The job would be faster on f-small by the margin provided by the need for a lower raw-relations count on f-small; not a waste at all to use 15e.

I'm trying unbalanced lim's and varying some other settings to squeeze out some more yield, trying for q-max under 400M. I expect to post the job sometime Saturday to the 15e thread.

 2021-10-11, 17:07 #83 swellman     Jun 2012 CB016 Posts 8+3_326 QUEUED AS 8p3_326 8+3_326 is a c193 from the HCN project ready for NFS. It's a SNFS 294 so GNFS it is. Thanks to Max0526, EdH and Plutie for their respective efforts developing more advanced spin tools and helping me try to improve the score of this polynomial. Code: n: 1271515007705102652189952725761191819303345301747516657201545585841235185390068713842394904588907747685724022636366668603115951556046696630803051897639628872247236774090687993328825102782693109 skew: 107487791.894 type: gnfs lss: 0 c0: 133940870445853294835954766480701480672986396520 c1: -1452026674957821687668946144418097210522 c2: -47876173107630258151460327041757 c3: 252440726525492253340481 c4: 4692042918656352 c5: -3027024 Y1: 661059788828815378 Y0: -17582196259493193321859678027857321489 # lognorm 61.62, E 52.82, alpha -8.79 (proj -2.55), 3 real roots # MurphyE(Bf=1.000e+07,Bg=5.000e+06,area=1.000e+16)=1.496e-14 lpbr: 31 lpba: 32 mfbr: 62 mfba: 94 alim: 134000000 rlim: 268000000 rlambda: 2.7 alambda: 3.5 16f_small test sieving results from the algebraic side with Q in blocks of 10k: Code: Q(M) Norm_yield 50 24598 100 25043 150 22717 200 20839 Suggests sieving over a Q range of 50-200M to get 350M raw relations. Caching it here for now. Last fiddled with by swellman on 2021-11-04 at 17:30
 2021-10-28, 00:50 #84 swellman     Jun 2012 324810 Posts HCN 3+2,1890L 3+2,1890L is now ready for GNFS. A sextic of SNFS difficulty 300 and an octic of 240 difficulty were possible but ultimately both proved inferior to the GNFS. Code: n: 12734530900787107377713574161011868289324430536561585108001584634844458530077827798443825688784448329151924256365768391241981555179087907348803260669361241812142769026337465374565031975093261 skew: 26407748.919 lss: 0 type: gnfs c0: 82995649502610121695803377054054562879174460 c1: -68939390311785240168852442729643236969 c2: -1159047757477660892006812203906 c3: 166840082438286438551781 c4: 583630704694594 c5: 15763440 Y0: -4381990159763602434411484418467631498 Y1: 1351413343517779682402767 # MurphyE (Bf=8.590e+09,Bg=4.295e+09,area=5.469e+16) = 1.091e-08 = 1.892e-14 per cownoise lpbr: 31 lpba: 32 mfbr: 62 mfba: 94 rlim: 134000000 alim: 268000000 rlambda: 2.7 alambda: 3.5 Test sieving on 16f_small with Q in blocks of 10K, sieved on the algebraic side: Code: Q(M) Norm_yield 60 28820 100 26626 150 25334 200 23482 Suggesting a Q-range of 60-200M to generate 360M raw relations. Caching here for eventual sieving.
 2021-10-28, 02:00 #85 VBCurtis     "Curtis" Feb 2005 Riverside, CA 11×461 Posts I think if you reverse the lims you'll get much better yield on this one. Smaller lim on sieve side and smaller lim on the 3LP side both suggest reversing which is 134 and which is 268.
 2021-10-28, 02:30 #86 swellman     Jun 2012 24·7·29 Posts Ouch. I’ll verify and correct. Thanks.
2021-10-28, 17:00   #87
swellman

Jun 2012

62608 Posts

Quote:
 Originally Posted by VBCurtis I think if you reverse the lims you'll get much better yield on this one. Smaller lim on sieve side and smaller lim on the 3LP side both suggest reversing which is 134 and which is 268.
A spot check at Q=100M gave a 7% improvement on yield with no measurable difference in sec/rel.

I had always assumed that lim should or must be < 2^(lpb-4), so that in this case with lpba=31 the highest useful alim is 134M. Seems not.

Now fully test sieving with the new parameters (i.e. swapping alim and rlim). Should sieve a bit faster with a tighter Q range, if the spot result holds constant over the sieving range.

Any chance of higher dup rates with alim this high?

2021-10-28, 18:33   #88
VBCurtis

"Curtis"
Feb 2005
Riverside, CA

11·461 Posts

Quote:
 Originally Posted by swellman Any chance of higher dup rates with alim this high?
Aren't you lowering alim to 134M? I'm not sure what you are asking here.

I would also make lpbr = 32; I don't see the point of a 31LP job using the 16e siever. That would remove the violation of your (likely useful) rule of thumb "lpb - limbits should be 4 or more".

What size is this number, anyway?

 Similar Threads Thread Thread Starter Forum Replies Last Post VBCurtis NFS@Home 220 2021-12-04 20:48 VBCurtis NFS@Home 72 2021-12-02 17:10 Rodrigo Software 7 2018-05-25 13:26 debrouxl NFS@Home 10 2018-05-06 21:05 joblack Information & Answers 1 2009-01-06 08:45

All times are UTC. The time now is 17:18.

Sun Dec 5 17:18:14 UTC 2021 up 135 days, 11:47, 1 user, load averages: 1.65, 1.72, 1.65