mersenneforum.org (https://www.mersenneforum.org/index.php)
-   NFS@Home (https://www.mersenneforum.org/forumdisplay.php?f=98)
-   -   2018 15e post-processing reservations and results (https://www.mersenneforum.org/showthread.php?t=23411)

 fivemack 2018-12-21 20:28

I've put up two more Lucas Aurifeuillian factors with 32/30 large primes (these ones are 3LP on the rational side because that seemed quicker from trial sieving)

 swellman 2018-12-22 01:05

[QUOTE=fivemack;503586]I've put up two more Lucas Aurifeuillian factors with 32/30 large primes (these ones are 3LP on the rational side because that seemed quicker from trial sieving)[/QUOTE]

I’ve got a pending C190 GNFS 32/32 3LP (-a side) job which test sieves at reasonable speed but your 32/30 strategy intrigues me. Do you think there are any potential pitfalls?

 fivemack 2018-12-22 08:47

[QUOTE=swellman;503623]I’ve got a pending C190 GNFS 32/32 3LP (-a side) job which test sieves at reasonable speed but your 32/30 strategy intrigues me. Do you think there are any potential pitfalls?[/QUOTE]

I would expect 32/30 to work poorly for GNFS, because part of the polynomial selection effort is to make the algebraic and rational sides of similar difficulty - I'm using it for SNFS where the algebraic side is fixed and always small, and the rational side defines the actual number you're factoring.

I tried 32/29 for the L4245AB numbers, but the time per relation went up by 20% and I wasn't able to convince myself that the number of relations needed would go down by more than 10%.

 swellman 2018-12-22 14:52

[QUOTE=fivemack;503650]I would expect 32/30 to work poorly for GNFS, because part of the polynomial selection effort is to make the algebraic and rational sides of similar difficulty - I'm using it for SNFS where the algebraic side is fixed and always small, and the rational side defines the actual number you're factoring.

I tried 32/29 for the L4245AB numbers, but the time per relation went up by 20% and I wasn't able to convince myself that the number of relations needed would go down by more than 10%.[/QUOTE]

Thank you for your insights. I’ll leave the GNFS job as is but will explore 32/30 on some upcoming SNFS jobs.

 fivemack 2018-12-23 19:38

My electricity bills for the compute farm (order of £250 a month) are starting to look disproportionate to my lack of income, so I intend to turn off a few nodes in the New Year and will have less post-processing available. I plan to keep the i9/7940X running linear algebra solidly, which ought to be a meaningful contribution.

 swellman 2018-12-23 22:02

Do what you have to do. Others will take up the slack. Or not. But do not worry - we all contribute as our resources allow.

 fivemack 2018-12-24 11:44

Reserving C217_133_94 for post-processing

 Batalov 2018-12-27 22:12

1 Attachment(s)
L2268 cofactor is finished
[CODE]Thu Dec 27 13:55:24 2018 p115 factor: 4468728470697900840421834967730559072239351609830923968364687516254286475954809097190830791569015578978902211846881
Thu Dec 27 13:55:24 2018 p152 factor: 18296844178068673791615190258094855085834941817043952297534445967738800235208282862863613003855248613009139445140931940077191085694589632995284025835767[/CODE]
[PASTEBIN]u0Cq6pvG[/PASTEBIN]

 Batalov 2018-12-29 18:42

I gave a quick test to L4245A (32/30, aim for 250M relations). (not reserving it, just testing)
Beware that duplication rate is much higher than for other sieving projects: currently from 263M raw rels, only 209M are unique.

 Batalov 2018-12-30 21:58

Reserving Fib(1983) cofactor.

Interesting fact: factoring Fib(1983) cofactor (which is [C]Fib(1983)/Fib(1983/3)[/C] ) is ...factoring [C]lucas(1322)-1[/C]

For odd n , [C]fibonacci(3*n) = fibonacci(n)*(lucas(2*n)-1) = fibonacci(n)*(lucas(n)^2+1)[/C]
For even n, [C]fibonacci(3*n) = fibonacci(n)*(lucas(2*n)+1) = fibonacci(n)*(lucas(n)^2-1) = ...[/C]
(but for even n, it is obvious that this falls apart into fibonacci(3*n/2)*lucas(3*n/2) and so on, by chain)

 Batalov 2018-12-31 04:05

Fib(1983) has a very high duplication rate, so far.
We will probably need to extend the region...
[C]Found 286915235 unique, 72560974 duplicate, and 3787 bad relations.
Largest dimension used: 1034 of 1300
Average dimension used: 875.6 of 1300[/C]

All times are UTC. The time now is 01:43.