Parametrization Assistance

RichD · 2018-12-03, 00:30

The top 12th number in the MWRB file has an OPN weight of 2.5M (2465207). It is ready for SNFS. With an SNFS difficulty of 281.5 it seems to be within reach using the 15e queue.

I have little to no experience at this level. What are the best parameters if it could be sieved using 15e?

Code:

n: 605716904027877980774625455520189647387776352555063757365644672493136637525085152114527251672682055452329862008130550673203343550128250999766605061023948523297828457779191592093682881010498969046911261346842026672855745883554109771998292748069377018429964450347583969787
# 732541^47-1, difficulty: 281.51, skewness: 9.49, alpha: 0.00
# cost: 8.45884e+19, est. time: 40280.21 GHz days (not accurate yet!)
skew: 9.494
c6: 1
c0: -732541
Y1: -1
Y0: 82919274927962023982932249248351337261442889121
m: 82919274927962023982932249248351337261442889121
type: snfs

VBCurtis · 2018-12-03, 04:12

I'd try test-sieving alim=rlim=268M, 33LP, 3 large primes on one side (I forget which side to expect to be better, but one test-Q range on each should make it clear).

I'd use 95 and 65 as the lpbr and lpba (or vice versa, depending on which is 3LP side). If yield is above 2.5, you could consider doing a 32/33 hybrid, or dropping the lim on the sieved side to 200M to save memory and maybe squeeze out a bit better sec/rel.

swellman · 2018-12-03, 04:33

I agree the sweet spot is 33-bit LP on 15e.

Besides the suggested parameters given by VBCurtis, you might want to try a/rlim=536M, and variants on that theme (e.g. 3 LP on one side, mixed alim/rlim, etc.). You’re up in thin air with this job - not many tread up here. Test sieving should show the path forward.

RichD · 2018-12-03, 05:38

Thanks for all the feedback. It looks like a lot of test sieving as I expected.

I thought I saw some "non-traditional" parameters like lambda=3.0 for 15e which didn't look familiar. Perhaps a hybrid sieve might be worthwhile. I've had mixed results with those. Then again, I'm not that fluent with cross pollination. Back to the salt mines.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Factoring GCW numbers - call for assistance.	xilman	Factoring	181	2016-07-03 14:32
Moderator assistance	Jim White	Forum Feedback	0	2007-08-31 05:01

2018-12-03, 04:12	#2
VBCurtis "Curtis" Feb 2005 Riverside, CA 2×2,437 Posts	I'd try test-sieving alim=rlim=268M, 33LP, 3 large primes on one side (I forget which side to expect to be better, but one test-Q range on each should make it clear). I'd use 95 and 65 as the lpbr and lpba (or vice versa, depending on which is 3LP side). If yield is above 2.5, you could consider doing a 32/33 hybrid, or dropping the lim on the sieved side to 200M to save memory and maybe squeeze out a bit better sec/rel.

2018-12-03, 04:33	#3
swellman Jun 2012 C14₁₆ Posts	I agree the sweet spot is 33-bit LP on 15e. Besides the suggested parameters given by VBCurtis, you might want to try a/rlim=536M, and variants on that theme (e.g. 3 LP on one side, mixed alim/rlim, etc.). You’re up in thin air with this job - not many tread up here. Test sieving should show the path forward.

2018-12-03, 05:38	#4
RichD Sep 2008 Kansas 2⁶·53 Posts	Thanks for all the feedback. It looks like a lot of test sieving as I expected. I thought I saw some "non-traditional" parameters like lambda=3.0 for 15e which didn't look familiar. Perhaps a hybrid sieve might be worthwhile. I've had mixed results with those. Then again, I'm not that fluent with cross pollination. Back to the salt mines.