20120725, 01:51  #23  
"Ben"
Feb 2007
3,617 Posts 
Quote:
Quote:
Another reason is momentum. At least for me, it was easier (and more fun) to start designing/writing a simple QS routine and build/optimize from there then to try to be immediately productive with a vastly complex thing like msieve. That's why we have yafu today and not a slightly faster msieve. Shallow vs. steep learning curve. This may or not make sense with something as complex as a lattice sieve when something like CADO exists. 

20120725, 02:07  #24  
Basketry That Evening!
"Bunslow the Bold"
Jun 2011
40<A<43 89<O<88
3·29·83 Posts 
Quote:
And btw, what about (the unbanned, apparently) RDS's lattice siever? How optimized/extensible is that? Last fiddled with by Dubslow on 20120725 at 02:09 Reason: s/confuses/puzzles (also needed more smileys) 

20120725, 02:30  #25 
"Ben"
Feb 2007
3,617 Posts 
If he's unbanned then he can make his own comments, but I don't believe it is any more ready for kilobit jobs then ggnfs. It is probably more easily extensible because it is far easier to read, but it is also starting on unequal footing because it uses a slower methodology to construct the reduced lattice basis.

20120725, 03:59  #26 
Bemusing Prompter
"Danny"
Dec 2002
California
9A2_{16} Posts 
But as CADONFS is licensed under the GPL, there is nothing to prevent someone from forking the project.

20120725, 04:01  #27 
Basketry That Evening!
"Bunslow the Bold"
Jun 2011
40<A<43 89<O<88
3·29·83 Posts 

20120725, 06:10  #28  
Sep 2009
1111010010_{2} Posts 
Quote:
It worked so well for RSALS, NFS@Home, yoyo@home and yafu, all of which are highermemory than polynomial selection would be (at least, when yafu is processing C110 and higher). Should stage 1 and stage 2 of polynomial selection be separated ? I mean, collecting all stage 1 hits, then keeping several hundreds / thousands / dozens of thousands of the best hits, and run the full stage 2 (root sieve, etc.) on these hits separately. This way, if the polynomial selection algorithms / parameters are tuned as a result of postprocessing the first dozens of coreyears, we could boil the best stage 1 hits multiple times. Last fiddled with by debrouxl on 20120725 at 06:16 

20120725, 11:39  #29 
Tribal Bullet
Oct 2004
5·709 Posts 
The best way to split poly selection is to have one group of computers run stage 1 plus the size optimization part of stage 2, which is a lot slower for degree 6 but still pretty quick in absolute terms and 100x faster than the root sieve. The code could be modified to save the top 10 or so optimized polynomials from each node, which get returned to the server and spotchecked. With RSA768 only about 1 in 100000 stage 1 hits were good enough to run the root sieve in stage 2, and for a much larger input the survival rate will be exponentially smaller.

20120725, 19:09  #30 
Aug 2002
8,461 Posts 
We believe that the coordination of a task such as this will be even more interesting, and noteworthy, because everybody is not in the same place. It will require some very good organization.
BTW, if "Operation Kibibit" succeeds, we will commission another extraordinary run of limited edition buttons. 
20120725, 20:38  #31 
6809 > 6502
"""""""""""""""""""
Aug 2003
101×103 Posts
24456_{8} Posts 

20120725, 21:01  #32  
Basketry That Evening!
"Bunslow the Bold"
Jun 2011
40<A<43 89<O<88
1C35_{16} Posts 
Quote:
Quote:
jasonp, how long do you think it would take to modify the poly select code? More importantly, we need to figure out where to get a siever. 1) Start from scratch 2) Modify/extend CADO's siever 3) Optimize/extend RDS's siever I can't comment on such things though. Last fiddled with by Dubslow on 20120725 at 21:02 Reason: dropping pic 

20120725, 21:09  #33  
"Carlos Pinho"
Oct 2011
Milton Keynes, UK
2·11·229 Posts 
Quote:


Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Where should I start?  christian_  Information & Answers  9  20160122 19:28 
Where to start  Jellyfish420  Homework Help  46  20130206 13:51 
How to start?  Thomas11  Lone Mersenne Hunters  29  20081221 13:47 
how to start with P1?  ValerieVonck  Marin's Mersennearies  8  20060429 22:21 
How to start?  OmbooHankvald  Factoring  15  20050903 13:42 