20040525, 13:51  #1 
Apr 2004
Copenhagen, Denmark
74_{16} Posts 
Line sieving vs. lattice sieving
Hi!
I (think) I have understood how the line siever works, but I have no clue of how the lattice siever works. Could someone please explain to me (and the rest of the group ) how the lattice siever works? I have tried the CWI line siever and Jens Franke's lattice siever. His siever seems to be faster. Is this due to the implementation or is there a theoretical reason for the lattice siever to be faster?  Best regards Jes Hansen Last fiddled with by JHansen on 20040525 at 13:52 
20040525, 16:36  #2  
Bamboozled!
"πΊππ·π·π"
May 2003
Down not across
2×5,323 Posts 
Quote:
If you know how the line siever works, you know then how the norms of each polynomial which are divisible by a specific prime are equally separated along the sieving line. (Handwaving warning: I mean prime ideal or, if you're happier with this phrasing, there is one such series per root of the polynomial mod p). Right, now fix a large prime, one bigger than any in the factorbase for a particular polynomial. Call it q. Better, call it special_q because we've chosen it to be special. Now, you will again agree that norms divisible by this specialq are regularly separated in the sieving region, right? That is, they form a lattice. The clever bit of the lattice sieve is to transform the polynomials or, entirely equivalent, the coordinate system of the sieving rectangle so that these lattice points become adjacent in the transformed system. Now sieve the transformed region with the factorbase primes. You know, a priori, that all the norms are divisible by a large prime and so, after that prime is divided out, the norm will be smaller by a factor of specialq and so (handwaving again) more likely to be smooth. Smooth is good, so you are more likely to get a relation. That's the handwaving reason why the lattice sieve is likely to be faster than the line sieve. It completely glosses over some features which damage its performance. For a start, reducing the norm of one polynomial is likely, in practice, to increase the norm of the other. To some extent this can be counteracted by using specialq on the polynomial which typically has the greater norm. Another source of inefficiency is the requirement for the coordinate transformations for each specialq. It's my belief that Jens Franke's lattice siever gains most of its speed over the CWI line siever from implementational differences. It has assembly language support for various x86 systems whereas the CWI siever uses much more general purpose code. The lattice siever's use of a very fast mpqs for factoring 2large prime candidates probably outperforms the CWIsiever's use of rho and squfof, though I haven't evaluated that area in any detail. Some of the speed increase, though, probably does come from it using the lattice sieve. Again, I haven't tried to disentangle the effects and to quantify them. Down sides of the lattice sieve become apparent when you consider the postsieving phases. First off, a prime can be a specialq and also a regular large prime for a different specialq and vice versa. That is, duplicate relations are almost inevitable when using a lattice sieve. The dups have to be identified and rejected. This takes computation and storage and, in a distributed computation, comms bandwidth. It also means that the raw relations/second measure isn't quite such a good measure of efficiency as it is for the line siever. Worse, the number of duplicates increases as the number of relations grows (another view of the birthday paradox) and so the effective rate of relation production falls as the computation proceeds. Something that tends to hit the linear algebra phase is that the specialq primes act in many ways as if they were factor base primes. That is, they tend to make the matrix larger and denser compared with the relations found by a line sieve with the same factorbases. However, the argument can be turned on its head: by using a large effective factor base one can use a smaller real factorbase and so speed up the sieving without losing relations. Think about it, and you'll see that this is another way of phrasing the paragraph above beginning "The clever bit ..." Hope this helps. Paul 

20040526, 15:30  #3  
Nov 2003
2^{2}·5·373 Posts 
Quote:
of the sublattice by finding a basis (which is easy) then using LLL or some other method to QUICKLY find a reduced basis. The quality of the reduction is important. See below. Quote:
lattice isn't small enough, the decreased norm is outweighed by a bigger increase (bigger than the decrease) in the other norm. Quote:
Quote:
precision) would be faster. In any event my code takes ~ 7% of the total run time in squfof. Doubling the speed would only yield ~3% improvement in total run time. Pollard Rho is quite a bit slower than squfof, but my code succeeds with squfof better than 95% of the time. Only if squfof fails do I use Rho. QS looks better for the 3 large prime variation. Here is siever data (mine) for 2,653+. The sieve length is [13M, 13M] per bvalue. Total values sieved is 13.6 x 10^9 (26 x 2^20 x 500) Siever built on Mar 18 2004 12:47:42 Finished processing the range 704500 to 704999 < b's In 2296.507117 elapsed seconds This is approximately 1920 Million Arithmetic Operations/sec < by estimated count of arithmetic ops only < times are msec > Total sieve time = 1419237.100308 (odd b sieving plus all subroutines) Total even sieve time = 858458.400465 ("" even b's plus subroutines) Total resieve time = 74504.261681 (time to factor by resieving odd b) Total even resieve time = 62589.224396 ("" even b) Total trial int time = 12307.307793 (time for trial division; linear poly) Total trial alg time = 208669.855863 ("" sextic poly) Total alg scan time = 29042.052749 (time to scan for successes) Total alg squfof time = 161006.544098 (time running squfof on sextic) Total int squfof time = 9264.036061 ("" linear) This last line is the actual time spent JUST sieving odd b's. Even b's take about 55% of the odd ones. Total asieve, isieve = 561935.884146 599546.527601 Quote:
the two methods because toward the end the lattice siever generates quite a lot of duplicates. One advantage the lattice siever has is the following. The yield rate for the line siever decreases over time because the norms get bigger as the sieve region moves away from the origin. The lattice siever brings the sieve region "back to the origin" when specialq's are changed. This might be its biggest advantage (if there is one) Quote:
I have considered a version where the specialq's lie outside the factor base, but backofenvelope shows that the "seesaw" effect means that for such primes, the increase in one polynomial more than offsets the decrease in the other. Of course a better implementation might alleviate this. My lattice siever was quite crude. I never spent much time on it. Last fiddled with by Wacky on 20040527 at 11:45 Reason: Fix quotes 

20040527, 11:21  #4 
Mar 2004
Ukraine, Kiev
2×23 Posts 
use [ quote="xilman" ] to make quotes (without spaces)
and [ /quote ] to close quotation (also w/o spaces) Last fiddled with by Death on 20040527 at 11:23 
20040528, 14:03  #5  
Nov 2003
2^{2}×5×373 Posts 
Quote:
Now. Would anyone like to discuss the lattice sieve? Or is it too specialized a topic? 

20040528, 15:11  #6  
Bamboozled!
"πΊππ·π·π"
May 2003
Down not across
2×5,323 Posts 
Quote:
Paul 

20100609, 15:23  #7 
Sep 2009
2^{2}·3·13^{2} Posts 
As a follow on from this I've read about classical sieving and line sieving, are they two names for the same thing? If not what is the difference?
Thanks in advance. Chris K 
20100609, 17:19  #8 
Nov 2003
2^{2}×5×373 Posts 

20100609, 17:36  #9 
Tribal Bullet
Oct 2004
3^{3}×131 Posts 
Boy you guys reach waaaay back. I have sometimes seen sieving near the roots of the algebraic polynomial in the (a,b) plane referred to as 'line sieving'. This was terminology that Chris Monico favored but I've never seen anyone else use it.

20100609, 19:25  #10 
Bamboozled!
"πΊππ·π·π"
May 2003
Down not across
10100110010110_{2} Posts 

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
I'm getting an error when yafu wants to start lattice sieving  Hailstone  YAFU  30  20180523 19:33 
Lattice Sieving Parameters  paul0  Factoring  6  20151120 21:12 
Lattice Sieving  where do I start?  paul0  Factoring  3  20150309 13:54 
A question on lattice sieving  joral  Factoring  5  20080403 08:01 
Initialization for lattice sieving  jasonp  Factoring  16  20060112 22:53 