Linear algebra reservations, progress and results

[fivemack]

Originally Posted by pinhodecarlos

Will 136^105+105^136 cofactor(15e) fit onto 16GB. If so I would like to reserve it. Also may I request some support on which value of TD should be used? Thank you in advance.

Carlos

I don't think it will fit in 16GB; and I expect that it will take really quite a long time to process, so I'd ideally want a post-processor willing to dedicate a dual Xeon E5v3 to the job; if I can't find one of those I will use my heaviest machine.

[pinhodecarlos]

Originally Posted by fivemack

I don't think it will fit in 16GB; and I expect that it will take really quite a long time to process, so I'd ideally want a post-processor willing to dedicate a dual Xeon E5v3 to the job; if I can't find one of those I will use my heaviest machine.

Thank you. instead I'll take next 11+2_232.

[jyb]

p53 factor: 14487018674856045577516403460196086299057074020508013
p174 factor: 940761631266562254843890915560554694187236007467661447907078365120189071950375177761515637901401566358450238557072111308837554703833860699687205618685114918062617855980937507

Log: http://pastebin.com/km1hzw2y

[pinhodecarlos]

Wed Sep 14 14:26:59 2016  prp88 factor: 3527500241184357141009872283004712704673339795852908184850523961396087333117589259749121
Wed Sep 14 14:26:59 2016  prp115 factor: 1209124221308511118247711000663108819289733412308024474982625427011739779496789556310431795014397908596756219092807

http://pastebin.com/2kPvbA9y

[jyb]

p110 factor: 54268481449280215840198360924838255215252555210331813793874392538308304682756053165706186679970465596514656071
p124 factor: 3180636421215086419421510383588502483627598789271555466138233449167799049803524721986809686719398104069819983642297733288827

Log: http://pastebin.com/xJQMnJ9w

[pinhodecarlos]

Taking C231_129_74 (14e)

[pinhodecarlos]

Originally Posted by pinhodecarlos

Taking C231_129_74 (14e)

Managed to build a 13^2 matrix with TD=120 which LA will take me 170 hours on 4 cores, 33% duplicates. Hence we are on a low 14e crunch participation, LA will be run, but can I request next time, for composites above 200 digits, to aim for more raw relations, push up to 270-280M relations? Also have the poly been verified?

Thank you.

Carlos

[unconnected]

Reserving 11-4_233.

[swellman]

Been working on C269_140_83 for awhile. Finally got it through filtering but now it's almost been a week just building the matrix. I think I'm out on the fringe with my system (16Gb memory) but trying to push through. I'll keep going until the end of the month - if it isn't built by then it will never happen. This effort has taken 5x longer than any previous 32-bit job I've attempted.

Can hardly wait to see the ETA if it makes it that far!

[jyb]

Originally Posted by pinhodecarlos

Managed to build a 13^2 matrix with TD=120 which LA will take me 170 hours on 4 cores, 33% duplicates. Hence we are on a low 14e crunch participation, LA will be run, but can I request next time, for composites above 200 digits, to aim for more raw relations, push up to 270-280M relations? Also have the poly been verified?

Thank you.

Carlos

This is very strange: until recently the numbers on the 14e queue routinely got around 220M relations and managed to post-process just fine. And that included many which were well over 200 digits. I can't see any reason that 250M relations shouldn't have been more than enough. Can anybody explain why there were 33% duplicates? That seems awfully high. By contrast, 11+2_232 has about 21% duplicates, which is much more in line with other jobs I've seen.

I think that routinely going up to 270M relations is a waste; I'd rather take these pathological cases on an individual basis and handle them as necessary. I.e. we can throw more relations at C231_129_74 if you want, and perhaps we'll end up with a net decrease in time, even accounting for having to start the post-processing over. Or perhaps not, but in any case I'd rather not assume that all jobs will be this bad.

Let me know if you want more sieving for this one.

[pinhodecarlos]

Jon, no more relations needed hence I am already running LA. This issue of high duplicates only happens on XYYXF composites above 200 digits. It's the second time this happens to me, first time LA was above 240 hours, this time I've decided to post-process this one.
In the meantime Jon will run 11+2_232 which I've reserved a week back due to be running C231_129_74 instead.