20081206, 20:54  #1 
Nov 2008
2·3^{3}·43 Posts 
Msieve NFS minimum size
Why does msieve's NFS require the number to be at least 97 digits? SNFS is much faster than QS for lower levels than that. Even just postprocessing will switch to the quadratic sieve instead.

20081207, 02:26  #2 
Tribal Bullet
Oct 2004
3^{3}·131 Posts 
Reports from users (admittedly old now) put the crossover point between msieve's QS and the GGNFS tools at 94100 digits. Admittedly this sucks if you are completing a <97 digit job started by GGNFS. I can make the limit smaller, but I've already had a case of a public project try to use msieve for 20digit NFS for instructional purposes.
I'll put in logic that uses NFS for postprocessing even if the input is smaller than the bound. Last fiddled with by jasonp on 20081207 at 02:28 
20081207, 09:34  #3  
Nov 2008
2·3^{3}·43 Posts 
Quote:
Just out of interest, how much slower is NFS than QS at 20 digits? Quote:
BTW, I assume the crossover point you mentioned is for GNFS numbers. It's SNFS numbers that I posted this thread about. Last fiddled with by 10metreh on 20081207 at 09:37 

20081207, 14:32  #4 
Tribal Bullet
Oct 2004
DD1_{16} Posts 
Yes, the crossover point assumes GNFS. A 100digit SNFS job with a correctly sized polynomial is around 10x faster than using QS.
At the 20digit size, the NFS code is all overhead. The 40digit sample test in the GGNFS source takes something like 30 seconds, whereas QS finishes in milliseconds. 
20081207, 17:47  #5  
Nov 2008
2×3^{3}×43 Posts 
Quote:
Will you update in the next version? 

20081214, 11:23  #6 
Nov 2008
2×3^{3}×43 Posts 
My question has still not had an answer...

20081214, 15:53  #7 
Tribal Bullet
Oct 2004
3^{3}·131 Posts 
The answer is yes, this change is trivial to make.
The next version could be out several months from now, though, which is why I wasn't in a hurry to answer. 
20081214, 16:23  #8  
Nov 2008
2×3^{3}×43 Posts 
Quote:
However, it's such an easy change to make that you should easily be able to do it now and then wait those few months for version 1.41. 

20081214, 17:43  #9 
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
2·29·101 Posts 
i have attached a modified version of msieve 1.36 that doesnt reject lower number sizes which i compiled a while back on my athlon 64

20081214, 17:48  #10 
Nov 2008
2×3^{3}×43 Posts 
Could we have a similar one for 1.39? The poly selection has changed, if I want to do a C96 GNFS, for example (that will probably crop up in my aliquot sequence).
Ther is also a cygwin dll needed. Last fiddled with by 10metreh on 20081214 at 17:50 
20081214, 18:00  #11  
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
2·29·101 Posts 
Quote:
i will have a go at compiling it i havent compiled anything like that on my newish pc so it might not be set up properly yet i cant do anything about needing a couple of cygwin dlls to run it you should be able to download them easily though 

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Estimating minimum relations  bchaffin  Factoring  24  20120324 18:37 
What is minimum RAM needed for 6 Core CPU for P1?  odin  Hardware  15  20100418 14:22 
Minimum/desired CPU specs for ECM factoring  Kaboom  PrimeNet  10  20090417 14:58 
Minimum primes of various forms database?  jasong  Information & Answers  1  20071101 01:58 
Minimum delay between server connections  vaughan  ElevenSmooth  5  20050908 17:17 