48-bit large primes! - Page 3

Andi47 · 2010-06-01, 17:58

Quote:

Originally Posted by fivemack

I don't think it makes a difference for current projects, because for 2801^79-1 we have alim=2^27, and 2.6 * 27 is already greater than mfba; for the aliquot job alim=2^26, and 2.6*26 is also large enough.

Whilst in this case alim=2^20, and 2.6*20 is substantially less than 66.

Hmmm...

Lambdas 2.6

Code:

C:\F_Fun\Wissenschaft\Mathe\GGNFS\BIN>gnfs-lasieve4I15e -r 2801_79-.poly -o 2801_79-_test2.out -f 19900000 -c 1000
 Warning:  lowering FB_bound to 19899999.
total yield: 2106, q=19901029 (0.34393 sec/rel)

Lambdas 3.2

Code:

C:\F_Fun\Wissenschaft\Mathe\GGNFS\BIN>gnfs-lasieve4I15e -r 2801_79-b.poly -o 2801_79-_test3.out -f 19900000 -c 1000
 Warning:  lowering FB_bound to 19899999.
total yield: 2128, q=19901029 (0.36133 sec/rel)

So there is a slight difference, but I think it's not worth taking higher lambdas.

jrk · 2010-06-01, 19:09

Quote:

Originally Posted by fivemack

I had thought that the cutoff was 2^(lpba*lambda), not alim^lambda !

This is strange. The ggnfs siever appears to use a cutoff of fblim^lambda, if I understood the code correctly, but the cado nfs siever (as it exists in the most recent svn) indeed does use 2^(lpb * lambda).

Batalov · 2010-06-01, 19:14

A couple of weeks ago, I checked the lambda efficiency for a large project: starting from 4.0 down until relations start getting lost significantly and sec/rel gets worse rather than better. Unsurprisingly, I found that the optimum is indeed somewhere at 2.5 and 2.6, so these settings are both good.
(The bottom is flat and to make any conclusions a lot of sieving is needed, but I got a not-so-well founded suspicion that 16e may run slightly better with 2.5, while 15e with 2.6. But the difference is tiny, so both values are just fine - and better than 2.4 or 2.7, 2.8. Caveat: this most probably depends on a CPU family. YMMV.)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Where do I send my PRP primes with large k?	Trilo	Riesel Prime Search	3	2013-08-20 00:32
lots of large primes	Peter Hackman	Factoring	2	2008-08-15 14:26
NFS with 5 and 6 large primes	jasonp	Factoring	4	2007-12-04 18:32
Why only three large primes	fivemack	Factoring	18	2007-05-10 12:14
What is the use of these large primes	Prime Monster	Lounge	34	2004-06-10 18:12

2010-06-01, 19:14	#25
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 19·23² Posts	A couple of weeks ago, I checked the lambda efficiency for a large project: starting from 4.0 down until relations start getting lost significantly and sec/rel gets worse rather than better. Unsurprisingly, I found that the optimum is indeed somewhere at 2.5 and 2.6, so these settings are both good. (The bottom is flat and to make any conclusions a lot of sieving is needed, but I got a not-so-well founded suspicion that 16e may run slightly better with 2.5, while 15e with 2.6. But the difference is tiny, so both values are just fine - and better than 2.4 or 2.7, 2.8. Caveat: this most probably depends on a CPU family. YMMV.)