[QUOTE=axn;395072]When the algebraic poly shows up with large coefficient(s) and/or higher than normal degree.[/QUOTE]

[QUOTE=Batalov;395074]It is elementary, Watson.
If you [I]have[/I] to use an snfs poly of a degree higher than the "optimal degree for the input size" (which was 5 for this difficulty-190 input but we [I]have[/I] to use 6 or 4), then 1) do indeed consider the other side, 2) the more the degree is off (e.g a 230 with an obligatory quartic), the more reason to test uneven limits and/or 3LP. Homework: check why deg-4 poly works worse that deg-6.

Even more generally, any 500 core-hour job deserves a half an hour preparation. "15 minutes could save you 15% or more" on your sieving. A larger job deserves a larger still prep. The side of sieving is just one of half a dozen parameters that one can tune. If you are not doing that, then you are missing all the fun. Trial sieve frequently to get better skills -- except for boring projects (for boring snfs projects no skills are needed).[/QUOTE]

Ahh, thanks. :smile: For now though, I only want to do 10-50 core hour jobs, i.e. let YAFU do all the work.

Meanwhile, (1145924053^19-1)/1145924052 split with P80 and P84 factors.

Nobody else seems to be working from this list; does anyone mind if I just reserve all of it?

[code]
1149989833 18 C164
36529 36 C165
58789 36 C168
46733917 22 C169

84969569171 16 C173
171634701455943275693670846531537378210379003 4 C177
149107399621 16 C179

563035735138626886051 2 C182
3760067 28 C185
28409 42 C188
237250154152941828313706973109892307074156361791 4 C188
4851463 28 C188
29501 42 C188
29567 42 C188
29573 42 C188
29581 42 C188
621011411269 16 C189

39097262657 18 C191
751410597400064602523400427092397 6 C194
828277 36 C199
1394714501 22 C199[/code]

I believe most are ECMed to about t35 or t40. So once you get above C180 (2/9 * 180 = t40), it might be worth it to run the final level of ECM first. Or you can roll the dice and go directly to SNFS. Others have found a factor to avoid a large NFS job.

Thanks Batalov for the tip. I knew of a couple forms but you provided enough info to make a general statement.
Is there a tips and tricks thread this can be posted in? :smile:
(Thanks for pointing out the typo. I knew something looked funny when I was manually sorting the list but forgot to follow up on the "2" after composing the message.)

You may want to double check the list Dubslow because a couple are spoken for.

In the mean time, some heavy hitters for those so inclined.
[CODE]21767 46 C200
22063 46 C200
22469 46 C201
22679 46 C201
22739 46 C201
23159 46 C201
23687 46 C202
5438347 30 C203
24671 46 C203
24709 46 C203
25037 46 C203
18956983 28 C204
5955749 30 C204
5366319547249 16 C204
26021 46 C204
22199431 28 C206
28793 46 C206
247121218571 18 C206
3253 60 C207
349235516317 18 C208
590713 36 C208
234138894262108061 12 C209[/CODE]

36529^37-1 is done:
r1=58091338511366844622808835606358541710891753351462809858576089771 (pp65)
r2=3096426004556675371903069514609578005304057921553504091410248482067194938553235704204354047244670991 (pp100)

Of those listed by Dubslow the following is done:
36529 36 C165 (above!)

And this has been truncated in his list:
563035735138626886051 2 C182
In mwrb2000.txt it appears in 44799264616686550530996315243199966966425091434166776850060158408354539673563035735138626886051 2
See Batalov's post about it.

I'm happy for Dubslow to do the rest.

Chris

These may take a month or more, is that okay?

With the two mentioned by Chris, I have these left:

[code]1149989833 18 C164
58789 36 C168
46733917 22 C169

84969569171 16 C173
171634701455943275693670846531537378210379003 4 C177
149107399621 16 C179

3760067 28 C185
28409 42 C188
237250154152941828313706973109892307074156361791 4 C188
4851463 28 C188
29501 42 C188
29567 42 C188
29573 42 C188
29581 42 C188
621011411269 16 C189

39097262657 18 C191
751410597400064602523400427092397 6 C194
828277 36 C199
1394714501 22 C199[/code]

There wouldn't really be anything better than the obvious quartics/sextic for the three low power numbers, correct? I'll try the various things that Batalov mentioned for those. (I would hazard a guess that the degree 4 C188 would be the least "well behaved" of the three?)

For prime q>13 values, not much can be done except to pay the (p-1) surplus in snfs difficulty and then - a quintic/sextic. But for high q's, p are relatively small, so that's not too much to pay.
For a few composite q (but those are rare, q=15 and 21), there are tricks. q=11, 13 can be "halved".

P.S. I liked the earlier described tricks with "halving" (small p, q=7) cases [URL="http://www.mersenneforum.org/showthread.php?p=370596#post370596"]down to cubics[/URL] (!); that would probably slide only up to 90 (100?) digits only though. Nifty, though. ;-)

IIRC, q=17 halving doesn’t kick in until somewhere around SNFS-220. So don’t even think about that for these low numbers. The saving is created when the base is large enough, say >p12, where the traditional polynomial gets elevated to a sextic, and that larger poly requires one-bit greater LPB than the octic. (If I said all that correctly.)

[QUOTE=RichD;395131]I believe most are ECMed to about t35 or t40. [/QUOTE]

I'm not so sure about even that:
[code](1149989833^19-1)/1149989832

P40 = 3213055981208391486124934392947803251061
P33 = 851540025603923338207219104468631
P49 = 6001969163083141878388768469774790732583407142943
P42 = 753485494973649196221699984378050106899431[/code]

Lesson learned.

I ([URL="http://mersenneforum.org/showthread.php?p=395400#post395400"]sort of[/URL]) ran ECM on all the others, and reported these to the FDB. It would seem that, at least as far as the FDB knows, little or no ECM has been done.

[code]prp33 = 349157269696963342797035080620271 | 29567^43-1

prp31 = 4427846691936399890015802742591 | 171634701455943275693670846531537378210379003^5-1

prp37 = 1415174572937984635943881666247304821 | 237250154152941828313706973109892307074156361791^5-1
^ Fully factored![/code]

29567^43-1 (CF193) still has a C156, so that looks like SNFS still.

171634701455943275693670846531537378210379003^5-1, CF222, still has a C147... I think that's still SNFS?

[QUOTE=Dubslow;395401]I ([URL="http://mersenneforum.org/showthread.php?p=395400#post395400"]sort of[/URL]) ran ECM on all the others, and reported these to the FDB. It would seem that, at least as far as the FDB knows, little or no ECM has been done.[/QUOTE]

Finally found the post I was looking for earlier [URL="http://mersenneforum.org/showpost.php?p=383172&postcount=173"]here[/URL].

[QUOTE=RichD;395407]Finally found the post I was looking for earlier [URL="http://mersenneforum.org/showpost.php?p=383172&postcount=173"]here[/URL].[/QUOTE]

That is almost certainly not true of the current batch of numbers. I'm definitely glad I decided to finish ECMing to the original 0.23 I had attempted, I've already found that (84969569171^17-1)/11640830976290) splits as:
[code]
P35 = 27174722980800683612328363916964857
P39 = 368490195799838835682326713093370926377
P100 = 5381496093876794918239577461697358547135704770377871739745150208536620399018626033415340527639882809
[/code]

There's still 14 more to ECM, but it's pretty clear already that these haven't been ECMd to t40. I'd guess t30.

Reserving:
21767^47-1
22063^47-1
22679^47-1
22739^47-1
23687^47-1
29437^43-1

For ECM to T45, then SNFS if necessary.

Chris