Reserved for MF - Sequence 276

[Dubslow]

Looks like ryanp, or so I assume, has pushed this multiple dozen lines past the most recent thread update. Currently a C205 at index 2122. In particular, index 2121 splits as:

6212003055...94<211> = 2 · 3 · 7 · 1060027307...87<72> · 1395292684...11<139>

Where the timestamps on the left side are ~September 4th, while the primes on the right are timestamped with ~September 12 (same as the current line). I assume that has to be a lucky ECM hit, because AFAICT, not even ryanp could NFS a ~C211 in a week.

I'm not quite sure what index was the former C206 of discussion here. Several lines in the DB are timestamped with early August, which is inconsistent with ryanp's most recent post.

[RichD]

I see yoyo is partially through a t55 run. I can't imagine the resources needed should there not be an ECM hit.

[Dubslow]

Originally Posted by RichD

I see yoyo is partially through a t55 run. I can't imagine the resources needed should there not be an ECM hit.

Looks like yoyo is about 1/4th through a t65 run (t60 is complete???).

How much of the t65 should be done before moving on to poly select?

[VBCurtis]

I think the Bayesian tool floating around these parts is more accurate than the old rules of thumb, but since I don't have the tool handy I'll speculate:
0.31 * 205 = 63.5 digits, where 63 digits is really close to half a t65. So, I'd do half a t65, and then a few hundred (a thousand?) speculative curves at t70 level while doing poly select at the same time.

I think the Bayesian tool would move up to 2.9e9 curves presently to finish the rest of the fraction of t65 that's left.

Poly select should be a serious undertaking. I'm willing to donate a GPU-month at minimum, though not immediately. I'm also willing to try CADO, though not as seriously ('cause CPU time is more precious than GPU time).

Does anyone care to confirm or disagree that deg 6 is worth some effort, even though 205 is likely still best at deg 5?

[Gimarel]

The C216 of 3^766+1 was still degree 5, so I disagree.

[wombatman]

Originally Posted by Dubslow

Looks like yoyo is about 1/4th through a t65 run (t60 is complete???).

How much of the t65 should be done before moving on to poly select?

Originally Posted by VBCurtis

I think the Bayesian tool floating around these parts is more accurate than the old rules of thumb, but since I don't have the tool handy I'll speculate:
0.31 * 205 = 63.5 digits, where 63 digits is really close to half a t65. So, I'd do half a t65, and then a few hundred (a thousand?) speculative curves at t70 level while doing poly select at the same time.

I think the Bayesian tool would move up to 2.9e9 curves presently to finish the rest of the fraction of t65 that's left.

Poly select should be a serious undertaking. I'm willing to donate a GPU-month at minimum, though not immediately. I'm also willing to try CADO, though not as seriously ('cause CPU time is more precious than GPU time).

Does anyone care to confirm or disagree that deg 6 is worth some effort, even though 205 is likely still best at deg 5?

Assuming 17,500 curves at B1=850e6 (the 1/4 of a t65), the Bayesian tool wants 12500 curves with B1 = 2.9e9. FWIW.

[Dubslow]

yoyo@home has completed 22K@850M.

[RichD]

Just a timestamp placeholder, ECM is completing at yoyo@Home and will soon be ready for poly selection. 22K @ 850e6 followed by 11K @ 2900e6. In case we forget and move on to more important numbers.

[VBCurtis]

I gave CADO the ad-range 0 to 1e4 on degree 6 to see what it would turn up:

 R0: -3354915467468925449144417001544070
Tue Mar 20 18:59:33 2018  R1: 1059264301675097949551
Tue Mar 20 18:59:33 2018  A0: 5491105381231820158278184320361978075589802
Tue Mar 20 18:59:33 2018  A1: -72595406437213271949280761939448929921
Tue Mar 20 18:59:33 2018  A2: -2573347630809788149132945021820408
Tue Mar 20 18:59:33 2018  A3: -304316465030583397390963971
Tue Mar 20 18:59:33 2018  A4: 1611344929661896554242
Tue Mar 20 18:59:33 2018  A5: -28111736477664
Tue Mar 20 18:59:33 2018  A6: -1663200
Tue Mar 20 18:59:33 2018  skew 1367862.47, size 5.175e-15, alpha -9.503, combined = 1.597e-15 rroots = 6

This is only 2% better than the best deg 6 found for a c207, so it's not very good. Then again, it was just 300k thread-seconds of size-opt search.

I don't think it will be difficult to crush this with deg 5, and I got to entertain my "it's fun to play with deg 6" impulse.

[Max0526]

@VBCurtis
I'll spin the above poly just to see what falls out and let you know.

[Max0526]

Just for a baseline. And look at this alpha.

R0: -6709829520299909389910598259273599
R1: 1059264301675097949551
A0: -10749655986314933365087770314170711558554086966160
A1: 605296494786174803296166070659199958758476
A2: 199216364276574487467150687385303144
A3: -3864338940017491067769623243
A4: -753181793331327212341
A5: 8693825592816
A6: 207900
skew 21442424.05, size 5.345e-015, alpha -11.416, combined = 1.675e-015 rroots = 6