mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Aliquot Sequences (https://www.mersenneforum.org/forumdisplay.php?f=90)

 smh 2009-03-22 14:40

Or just print the lucky curve if a factor is found

 mklasson 2009-03-22 14:50

[QUOTE=smh;166265]Or just print the lucky curve if a factor is found[/QUOTE]

Arrr, and here I thought I could get away cheap... :smile: But yeah, it would be cleaner. I'll think about it.

 10metreh 2009-03-22 15:37

[quote=mklasson;166259]On what do you base this and what would you prefer instead? I'm not saying you're wrong, it's just that if you want to spend about 1/4 of the qs time on ecm that's roughly the amount of ecm you'll have to do. Again, I'm not sure if 1/4 time is the best figure but it seems decent enough and I have yet to find some solid justification for any specific amount. Please show me the light of reason if you're hoarding it.[/quote]

The most I would possibly run on a C93 is t30. t30 + 73 @ 35 digits is too much. On my incredibly slow computer, a t30 on a C93 takes 51 minutes. 73 curves at 1e6 takes another 35 minutes. That's 86 minutes. And that's a lot.

 mklasson 2009-03-22 16:07

[QUOTE=10metreh;166268]The most I would possibly run on a C93 is t30. t30 + 73 @ 35 digits is too much. On my incredibly slow computer, a t30 on a C93 takes 51 minutes. 73 curves at 1e6 takes another 35 minutes. That's 86 minutes. And that's a lot.[/QUOTE]

But do you have any reasoning, math, or experimental data that supports your conclusion of "too much"? Like I said, I'm aiming for roughly a fourth of the qs time spent ecming. Can you explain [b]why[/b] that is too much? I'd be very happy to hear it, but just saying that you think it's "too much" and "a lot" doesn't really help me much I'm afraid. :mellow:

 10metreh 2009-03-22 16:13

[quote=mklasson;166270]But do you have any reasoning, math, or experimental data that supports your conclusion of "too much"? Like I said, I'm aiming for roughly a fourth of the qs time spent ecming. Can you explain [B]why[/B] that is too much? I'd be very happy to hear it, but just saying that you think it's "too much" and "a lot" doesn't really help me much I'm afraid. :mellow:[/quote]

Sorry. BTW, do you know what a more common term for "a fourth" is? [COLOR=white]A quarter.[/COLOR]

Feature request for aliqueit: store the known factors of the iteration being worked on in a file and immediately check the file for them as soon as you resume, so you don't have to go searching for the factors in aliqueit.log when you want to resume.

 mklasson 2009-03-22 16:34

[QUOTE=10metreh;166271]Sorry. BTW, do you know what a more common term for "a fourth" is? [COLOR=white]A quarter.[/COLOR][/quote]

:lol: uh, yes? Yes, yes, I do! If you look carefully you'll see I have in fact already used the magic word in this very thread! :shock: Was there a particular reason you said that? Am I missing out on a joke or something here?

[QUOTE=10metreh;166271]Feature request for aliqueit: store the known factors of the iteration being worked on in a file and immediately check the file for them as soon as you resume, so you don't have to go searching for the factors in aliqueit.log when you want to resume.[/QUOTE]

You rarely have a big enough factor in the middle of an iteration that it's worth the bother saving and restoring it though. But sure, it would be neat. I'll think about it.

 10metreh 2009-03-22 16:47

[quote=mklasson;166274]You rarely have a big enough factor in the middle of an iteration that it's worth the bother saving and restoring it though. But sure, it would be neat. I'll think about it.[/quote]

The other thing involves the -e option. If you use -e, you have to put the factor(s) in the command line if they are larger than the trial factoring cutoff. For instance, with my current iteration of 130396, there was a C99 which had a P7 factor, leaving a C93 which I am sieving at this very moment. However, when I started aliqueit with "aliqueit 130396 -e", it went off into a GNFS on the C99 without finding the P7. So I had to run "aliqueit 130396 -f 3373717 -e" instead, after finding the factor in aliqueit.log buried under full t20, t25 and t30 runs.

 mklasson 2009-03-22 16:50

[QUOTE=mklasson;166274]You rarely have a big enough factor in the middle of an iteration that it's worth the bother saving and restoring it though. But sure, it would be neat. I'll think about it.[/QUOTE]

Hmm... except if you're restarting with "-e" and skipping ecm altogether of course. It would be a great shame to run gnfs on a c120 instead of a c110...

 10metreh 2009-03-22 16:52

[quote=mklasson;166279]Hmm... except if you're restarting with "-e" and skipping ecm altogether of course. It would be a great shame to run gnfs on a c120 instead of a c110...[/quote]

...when a P10 is lurking within. Of course, msieve would find the factor during the run, so it wouldn't be a GNFS at all if you used msieve for the poly search, and wouldn't be a complete GNFS unless you used factLat.pl.

P.S. Look at my post count!

 mklasson 2009-03-22 17:04

No joke then? :cry:

[QUOTE=10metreh;166277]...buried under full t20, t25 and t30 runs.[/QUOTE]

Very subtle. :lol:

Fortunately you'll at least have much less dirt to paw your way through in the next version. I'll see about doing something nicer.

 10metreh 2009-03-22 17:12

[quote=mklasson;166281]No joke then? :cry:[/quote]

Yes joke, actually (I apologise about my deliberately poor English). I just forgot!

And I'm moving out of beastly territory now. Change to location field needed!

All times are UTC. The time now is 09:15.