[QUOTE=Dubslow;319902]How do you choose the params for GNFS polys?
[/QUOTE]
Lionel chooses the params for lasieved tasks. I'm not sure how...

[QUOTE=frmky;320097]Sure there are. In Windows, lasievef supports 16e V4 workunits. In Linux, lasievef uses the V5 binary.[/QUOTE]

I didn't know that. What it used in the past at NFS@Home?

In team sievings for Aliquot 4788 on this forum, I had noticed that some people posted polynomials where rlim & alim were 2^([lpba or lpbr]-4)-1, at least for 29-bit LPs and 30-bit LPs tasks. That's a bit higher than what factMsieve.pl tends to generate.
For a 31-bit LPs task, which 2_1584P (in ElevenSmooth notation) is because 30-bit LPs just don't cut it with 14e on a GNFS difficulty 168 task, this corresponds to rlim=alim=2^27-1 :smile:

I know that having too small a factor base hurts; conversely, having a factor base a bit larger than necessary doesn't seem to hurt yield or speed (at least, when sieving near rlim/2), though it clearly raises the time for building the fb, and AFAICS, it also raises the memory consumption (but it should remain lower than the consumption of 15e, which has a wider sieving area).

I'm all ears about ways to improve my polynomials, though :smile:

The V5 app could be sped up a bit if someone worked out the optimal amount of ecm and p-1 to do on each number before factoring with the qs routine. Currently no ecm or p-1 is done. BTW if ever need it we should be able to boincify 16f which sieves with composite q. This might be helpful if we run out of useful prime special q. It runs at about half the speed per relation but with higher yield. I haven't tested duplication rate with composite special q. I couldn't get the -R resume feature to work with it as that relies on prime q. 16g is also available although I couldn't work out how it worked. It is also slower.

[QUOTE=pinhodecarlos;320110]I didn't know that. What it used in the past at NFS@Home?[/QUOTE]

Damm, posted that at 3:43 am in the morning. Forget the question with grammar error.

863_83_minus1
 

[CODE]prp98 factor: 32968935720300560969344075865670759358503953157211261045249172290998451663657771694296812904708207
prp120 factor: 312685375967612103907819391528358697186649672662208181378156603351734738890605204228546900791446767907105556551283853443[/CODE]

677_95_minus1
 

[CODE]prp99 factor: 457226377236875494318218827815257413307389302178535344025368250642888329021072353153319514487859431
prp103 factor: 7253999621599379774542671416227215652175483263628442856864454044668807159620786834540016356629446692441[/CODE]

701_95_minus1
 

[CODE]prp83 factor: 10166112043795248896526895028504375101209927409953951017566750551469670864598061561
prp121 factor: 4007643265716711128336352634159033229181316018184315987787894190152169438153778256508984193436005706859411437738518784551[/CODE]

10 teams joined so far the "NFS@Home December to Remember!" challenge issued by California State U. Fullerton Physics Dept.

[B]Start time[/B] 2012-12-11 00:00 UTC

[B]End time[/B] 2012-12-18 00:00 UTC

[URL]http://boincstats.com/en/stats/challenge/team/chat/283[/URL]

Maybe it is better to create more 16e and 16e V5 wu's.
One hour to start.

Carlos Pinho

797_95_minus1
 

[CODE]prp71 factor: 22757933820887707325922752200910310540875987249248032150091835626373761
prp121 factor: 1602531850382600666503544888564812919590889297167321390543270930723958545888802623384261716775711441535831115078491627161[/CODE]

Reserve odd perfect numbers
 

Can I please reserve 941_67_minus1 and 947_67_minus1. Thanks