Running GGNFS - Page 73

EdH · 2013-12-29, 20:26

I have a machine that is returning only ~12% as many relations as others. I'm curious as to why and whether I can do anything about it. I tried it and another machine with identical poly files several times with the same results each time. I have the poly and one of the runs from each machine below:

For some reason, I incorrectly believed that B²'s sievers did not run on AMD. This erroneous idea must have been from an earlier experience with one of my antiques.

B²'s sievers are, indeed, returning many more relations in a speedier manner.

balamber · 2014-03-02, 10:03

Roughly speaking, I need 40M (unique) relations for lpba=lpbr=29 and 80M relations for lpba=lpbr=30. (More exact estimations are out of interest now.)

My question is, How many relations I need for lpba != lpbr, i.e.
1) lpbr=29, lpba=30
2) lpbr=30, lpba=29 ?

Does the answer depend on a side of sieving, i.e.
r) rational side
a) algebraic side
b) both sides.

It seems to me, I need the maximal number in any case, i.e.

1r) 80M
1a) 80M
1b) 80M

2r) 80M
2a) 80M
2b) 80M

Is it correct? Otherwise, give your variants, please.

Thank you in advance.

axn · 2014-03-02, 11:25

Split the difference. Let's say, 60M.

chris2be8 · 2014-03-02, 16:40

I've done a few like that. 57^145-1 with LPBA=29 and LPBR=30 needed 56385091 relations after removing duplicates (53556727 relations was too few).

I don't think it would vary with which side has the larger LPB. But I've not done one with larger LPBA.

Chris

VBCurtis · 2014-03-02, 21:30

If raising both sides one bit doubles the relations needed, raising one side should scale relations by sqrt2. Chris' 56M fits this exactly.

I do not think it matters which side is up one bit, though I imagine 29/30 vs 30/29 would sieve with different efficiencies.

chris2be8 · 2014-03-04, 16:39

LPBR>LPBA is appropriate for SNFS polys with lower degree than optimal. That was a quartic at about 200 digits SNFS difficulty.

LPBA>LPBR would be appropriate if the degree is too high. Eg degree 7 or 8. But I've not done any such.

Chris

2013-12-29, 20:26	#793
EdH "Ed Hall" Dec 2009 Adirondack Mtns 111011101111₂ Posts	Yet, another query - re: vastly different sieve returns I have a machine that is returning only ~12% as many relations as others. I'm curious as to why and whether I can do anything about it. I tried it and another machine with identical poly files several times with the same results each time. I have the poly and one of the runs from each machine below: For some reason, I incorrectly believed that B²'s sievers did not run on AMD. This erroneous idea must have been from an earlier experience with one of my antiques. B²'s sievers are, indeed, returning many more relations in a speedier manner. Last fiddled with by EdH on 2013-12-29 at 20:56 Reason: Gained knowledge through experience

2014-03-02, 10:03	#794
balamber Jun 2012 spb.ru 24₈ Posts	Rels number for lpba != lpbr Roughly speaking, I need 40M (unique) relations for lpba=lpbr=29 and 80M relations for lpba=lpbr=30. (More exact estimations are out of interest now.) My question is, How many relations I need for lpba != lpbr, i.e. 1) lpbr=29, lpba=30 2) lpbr=30, lpba=29 ? Does the answer depend on a side of sieving, i.e. r) rational side a) algebraic side b) both sides. It seems to me, I need the maximal number in any case, i.e. 1r) 80M 1a) 80M 1b) 80M 2r) 80M 2a) 80M 2b) 80M Is it correct? Otherwise, give your variants, please. Thank you in advance.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Installation of GGNFS	LegionMammal978	Msieve	17	2017-01-20 19:49
Running other programs while running Prime95.	Neimanator	PrimeNet	14	2013-08-10 20:15
Error running GGNFS+msieve+factmsieve.py	D. B. Staple	Factoring	6	2011-06-12 22:23
GGNFS or something better?	Zeta-Flux	Factoring	1	2007-08-07 22:40
ggnfs	ATH	Factoring	3	2006-08-12 22:50

2014-03-02, 11:25	#795
axn Jun 2003 11724₈ Posts	Split the difference. Let's say, 60M.

2014-03-02, 16:40	#796
chris2be8 Sep 2009 2·7·149 Posts	I've done a few like that. 57^145-1 with LPBA=29 and LPBR=30 needed 56385091 relations after removing duplicates (53556727 relations was too few). I don't think it would vary with which side has the larger LPB. But I've not done one with larger LPBA. Chris

2014-03-02, 21:30	#797
VBCurtis "Curtis" Feb 2005 Riverside, CA 4864₁₀ Posts	If raising both sides one bit doubles the relations needed, raising one side should scale relations by sqrt2. Chris' 56M fits this exactly. I do not think it matters which side is up one bit, though I imagine 29/30 vs 30/29 would sieve with different efficiencies.

2014-03-04, 16:39	#798
chris2be8 Sep 2009 2×7×149 Posts	LPBR>LPBA is appropriate for SNFS polys with lower degree than optimal. That was a quartic at about 200 digits SNFS difficulty. LPBA>LPBR would be appropriate if the degree is too high. Eg degree 7 or 8. But I've not done any such. Chris