Fast Breeding (guru management)

[wblipp]

Originally Posted by debrouxl

The 15e grid needs something more substantial soon, though

This has had ECM to 2/9 the SNFS size by yoyo@home.

251101831^29-1

[fivemack]

OK, I've queued up 251101831^29-1. It would be nice if somebody could do a bit more polynomial selection on C177_148_94, but if the queue drains again I would be inclined to queue up VBCurtis's preliminary polynomial. I'm afraid my systems are all committed to other things for at least the next week.

[VBCurtis]

I've reached 10 days on the C195 from 4788 without improving on Rich's poly (in fact, I found no poly over 1e-14, so not within 5%). I'll move to the C177, since there's an immediate need.

[debrouxl]

After test sieving on 14e, I suggest targeting two XYYXF numbers at 15e:

* for C226_118_109, snfspoly produces a sextic with decent-looking coefficients:

n: 2188631741508591403562986651632391146379880884734645955005812183277583487899017905891245273673542274647836721631476604265405152233693016534178740852390420954793799826044483718604333023557056283828855655156707029152010737589839
deg: 6
c6: 118
c5: 0
c4: 0
c3: 0
c2: 0
c1: 0
c0: 141158161
Y1: 514166125484246966737602341678133311989
Y0: -19673250936660415417029531820024397824
type: snfs
skew: 10.3031652179546
rlim: 134217727
alim: 134217727
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

but it sieves pretty badly, well below 1 rel/q and nearly 0.6 s/rel at q=rlim/2=alim/2, on this test computer. Using 32-bit LPs on 14e could help, but I think it's better to switch gears.

* likewise, C213_119_103 sieves badly, below 1 rel/q and ~0.5 s/rel at q=rlim/2=alim/2 here:

n: 375891859168290888618515533403195981147472163511562214121431333874627860313286561493948705757702810931173654456742886255080215519359875888847846457536080441599565518163441566286826930121390202964824411348611499983
deg: 6
c6: 119
c5: 0
c4: 0
c3: 0
c2: 0
c1: 0
c0: 11592740743
Y1: 175350605307710078811389641512788920567
Y0: -192441327313530246357280390753883639
type: snfs
skew: 21.450620542434
rlim: 134217727
alim: 134217727
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

Same cause, same treatment ?

For reference: despite c6 and c0 being much higher, the sextic for C204_119_99

n: 479873652024788694867787610438213252055477100315679358935651957440860083358098397722970076875195892606226009323886570230485431162602984527677762711480271946889349965985508220148978699023308352816407973617
deg: 6
c6: 1685159
c5: 0
c4: 0
c3: 0
c2: 0
c1: 0
c0: 13045131
Y1: 247850587150676021481575351680101875697
Y0: -1617154011038069297120003283646081
type: snfs
skew: 1.40648334377312
rlim: 134217727
alim: 134217727
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

sieves a bit above 1 rel/q and around 0.3 s/rel at q=rlim/2=alim/2, so 14e can deal with it, and I queued it there. The quintic has c5: 395307 and c0: 200533921, so probably not better.

[fivemack]

Trial-sieving debrouxl's two polynomials with 15e/32LP, will enqueue tomorrow once I've figured out plausible parameters.

[fivemack]

debrouxl's polynomials both enqueued (A-side yields at 15e/32LP ~3.0 for 119_103 and ~2.4 for 118_109). Aiming for 400M relations, which I think is quite strong oversieving for SNFS of that small a difficulty.

[debrouxl]

I think I've queued all numbers suitable for 14e which were posted here. The 14e part of the grid is not in danger of immediate starvation (that is, unless a sizable team decides to run for stats without caring about project management, as usual), but will be in several days

[XYYXF]

C206_119_97 and C235_119_101 survived 9k+ at 110M.

[swellman]

C197_118_105 has survived 8000+ curves @B1=11e7 with no factors found.

[jyb]

Originally Posted by debrouxl

I think I've queued all numbers suitable for 14e which were posted here. The 14e part of the grid is not in danger of immediate starvation (that is, unless a sizable team decides to run for stats without caring about project management, as usual), but will be in several days

Should you want them, the following Homogeneous Cunninghams are all around difficulty 250 and have all had 10,000 curves at B1 = 11e7.

11+4,239
11+5,239
11+7,239
11+8,239

[debrouxl]

I have just queued C206_119_97.

C235_119_101 is borderline for 14e. The sextic's coefficients are fantastically large (c6: 10510100501, c0: 23863536599), so I haven't even test-sieved that, and I went with the quintic of more reasonable coefficients

n: 2589310456899832928933301076778669578879120122369630480355405997172253313910033528160466408537532221128911504403237712622758278562537906325833003635581785560347127735770567490072987128252367045976109835671185970738159965132020094283971
deg: 5
c5: 119
c0: 104060401
Y1: 12571630183484301672314008717756984377273532301
Y0: -324294234694341316421188266002423799213601
type: snfs
skew: 15.4293527015567
rlim: 134217727
alim: 134217727
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

but that is below 1 rel/q and around 0.4 s/rel on this computer. That's better than the two XYYXF tasks recently steered at 15e instead, and 14e/32 could probably do it. What do other grid sheepherders think ?

I'll preprocess C197_118_105 and the 4 HCN later.