Reserved for MF - Sequence 4788

[jrk]

+150 curves @ B1=110000000, B2=1589211473866

[henryzz]

i think based on jasonp's recent posts in the msieve gpu thread that there is sufficient randomness in the selected search space that we can search the same A5 values without much loss of efficiency
i would check with him first though

[jrk]

Originally Posted by henryzz

i think based on jasonp's recent posts in the msieve gpu thread that there is sufficient randomness in the selected search space that we can search the same A5 values without much loss of efficiency
i would check with him first though

No, it isn't random. Msieve doesn't do what Jason was suggesting. If you want to run msieve, make sure you start with a higher A5 than where Tom ends. Something above 100K should be high enough.

So far I have searched through to 6K, but with a higher rational coefficient which does not overlap the normal msieve selection. I'm not entirely convinced that this is worth doing yet, though. But I have found some good polys.

[jrk]

Tom, what do you suggest is a good target score for this number?

So far my best is a 1.035e-12.

[jrk]

Once the ECM is finished it might be worth doing some poly searching with pol5 as well.

[fivemack]

Looks as if 1e-12 is a pretty reasonable score, I've got a couple above that level after a day of polsel, my best is

# norm 9.159378e-16 alpha -6.373628 e 1.028e-12
skew: 15249475.33
c0:  5865211302555522024730119243238438762464
c1:  2222587031863375855063420818065564
c2:  82026281057189973786655296
c3: -9332835943493042911
c4: -2896035406838
c5:  6960
Y0: -20007761313361394254221339534949
Y1:  350230344983122741

with (14e, 30-bit lpa/lpr, alim=rlim=50e6)

total yield: 1737, q=50001037 (0.14975 sec/rel)

I would go with the best polynomial that we've found by the time people are satisfied there's been enough ECM; not sure it's worth also doing pol51m.

This is a six-CPU-month sieving job, we'll probably get it done inside two weeks, so it's not worth delaying the start of sieving by two days even to find a polynomial that makes the sieving 10% quicker.

[Andi47]

Originally Posted by fivemack

Looks as if 1e-12 is a pretty reasonable score, I've got a couple above that level after a day of polsel, my best is

# norm 9.159378e-16 alpha -6.373628 e 1.028e-12
skew: 15249475.33
c0:  5865211302555522024730119243238438762464
c1:  2222587031863375855063420818065564
c2:  82026281057189973786655296
c3: -9332835943493042911
c4: -2896035406838
c5:  6960
Y0: -20007761313361394254221339534949
Y1:  350230344983122741

with (14e, 30-bit lpa/lpr, alim=rlim=50e6)

I guess we'll need ~100M relations?

Edit: your formula gives a "good score" as 1.276e-12.

[10metreh]

c161 is near the crossover between 14e and 15e. Which is better with the current best poly?

[Batalov]

Another 1200 @ 110e6 are done, so fire at will.

[jrk]

Originally Posted by Batalov

Another 1200 @ 110e6 are done, so fire at will.

Don't fire yet, I may have a better poly from the overnight search. I'll post again in a bit.

[jrk]

Originally Posted by jrk

Don't fire yet, I may have a better poly from the overnight search. I'll post again in a bit.

I do.

n: 22315086009543699881696673140311700365075401903410041991888847307663799610560235574994125636703013521384678857814894314475474921423859163748271799984152025524899
# norm 1.050714e-15 alpha -7.337146 e 1.121e-12
skew: 25520528.70
c0:  31410679515197091100341643412427401243420
c1:  4180601981543089138092261003493340
c2: -2347505003114131780984812301
c3: -51235139041477161474
c4:  3624790086036
c5:  7560
Y0: -19679511428703587012905042402449
Y1:  492226961856379171
rlim: 50000000
alim: 50000000
lpbr: 30
lpba: 30
mfbr: 60
mfba: 60
rlambda: 2.6
alambda: 2.6

Originally Posted by 10metreh

c161 is near the crossover between 14e and 15e. Which is better with the current best poly?

14e is faster.