BOINC NFS sieving - NFS@Home

[pinhodecarlos]

Originally Posted by jyb

Yes, looks like you're right. Hmm, usually Carlos warns us when one of these things is coming.

, I knew but other concerns/priorities since I’m on A/L.

[jyb]

Originally Posted by pinhodecarlos

, I knew but other concerns/priorities since I’m on A/L.

No need to be sorry. It wasn't a criticism; you're under no obligations here. I just found myself missing your customary warnings and suggestions.

[pinhodecarlos]

Originally Posted by jyb

No need to be sorry. It wasn't a criticism; you're under no obligations here. I just found myself missing your customary warnings and suggestions.

Just be aware a project can be chosen more than once this year. Already happened twice with Rosetta@Home.

Dates are here (https://formula-boinc.org/sprint.py?...ang=&year=2021), a project is announced 24h before start but by its voting from the forum you can have an idea which one could be.

[swellman]

Originally Posted by frmky

I increased the wu generation limits so there's plenty of work available now.

The system seems to have reached equilibrium. But the remaining jobs currently sieving won’t last long. 15e will be empty by morning.

Does anyone have composites ready for factoring? If so please post them!

[debrouxl]

Several ideas from near-repdigit:
* in the 15e range, https://stdkmd.net/nrr/c.cgi?q=14449_260 is at t55 for SNFS difficulty 261, so it wouldn't be too bad of an ECM miss. https://stdkmd.net/nrr/c.cgi?q=19997_274 and https://stdkmd.net/nrr/c.cgi?q=79999_277 are at t55 as well, so with higher risks of ECM miss, and closer to the end of the optimal 15e difficulty range;
* https://stdkmd.net/nrr/c.cgi?q=71111_329 is supposedly at t65 and has a polynomial. Factoring this 208-digit number would blast the current GNFS factorization record for the near-repdigit project, currently at 183 digits. It's way beyond the lesser difficulties I'm used to for 14e and 15e to a lesser extent, so I have no clue whether this polynomial is good and which siever is best for such a number

[chris2be8]

I'll test sieve a few Brent tables numbers, to be sieved on either 15e or 15e_small as needed to keep the queues full.

[swellman]

Originally Posted by debrouxl

* https://stdkmd.net/nrr/c.cgi?q=71111_329 is supposedly at t65 and has a polynomial. Factoring this 208-digit number would blast the current GNFS factorization record for the near-repdigit project, currently at 183 digits. It's way beyond the lesser difficulties I'm used to for 14e and 15e to a lesser extent, so I have no clue whether this polynomial is good and which siever is best for such a number

Running with this idea, the poly listed (with a current record e-score) is given below:

n: 1122306776491337588607322631818708778200214577213206237089370253284687138834200770294167044440649977604444005064220843458726284245767449140856119245849745806395315477933380280425456889166505800707705642211253
skew: 326211947.89
type: gnfs
lss: 0
c0: 192558034193459742319046371398586259699707875364425
c1: 5387479895042599888816938205129089277819185
c2: -1098937059524264818729815697407187
c3: -104048030268009541344497393
c4: -47857156754642446
c5: 421635720
Y0: -4842112079991293167491670985200158992968
Y1: 4692246580297096789
# Murphy_E = 1.439e-15, selected by Erik Branger
# Polynomial selection took 4 months on a GTX760
# selected mechanically
rlim: 1180000000
alim: 1180000000
lpbr: 32
lpba: 32
mfbr: 69
mfba: 69
rlambda: 2.8
alambda: 2.8

BUT, can this poly be spun or otherwise improved? And is a G208 too difficult for 16f_small?

[RichD]

Originally Posted by swellman

The system seems to have reached equilibrium. But the remaining jobs currently sieving won’t last long. 15e will be empty by morning.

Does anyone have composites ready for factoring? If so please post them!

I've posted 4733^71-1 late last night...

[jyb]

Originally Posted by swellman

The system seems to have reached equilibrium. But the remaining jobs currently sieving won’t last long. 15e will be empty by morning.

Does anyone have composites ready for factoring? If so please post them!

Keep in mind that the sprint only goes for 3 days, and the first day is more than half over. The equilibrium you mentioned appears to have settled in on somewhere around 30,000 wu's per day per queue (e-small, e, f-small). Unless that number increases substantially, then:

- The numbers already in e-small should just about be enough to satisfy the demand.
- One more number in e should be sufficient to satisfy demand there.
- We won't even come close to finishing the next f-small number.

Of course, it would be nice to have a margin for error. But I don't think we have to scour far and wide for candidates. Just two or three of a reasonable size should do it.

[swellman]

Originally Posted by RichD

I've posted 4733^71-1 late last night...

Indeed and I’ve just now enqueued it. Sorry I had missed it!

[swellman]

Originally Posted by jyb

Keep in mind that the sprint only goes for 3 days, and the first day is more than half over. The equilibrium you mentioned appears to have settled in on somewhere around 30,000 wu's per day per queue (e-small, e, f-small). Unless that number increases substantially, then:

- The numbers already in e-small should just about be enough to satisfy the demand.
- One more number in e should be sufficient to satisfy demand there.
- We won't even come close to finishing the next f-small number.

Of course, it would be nice to have a margin for error. But I don't think we have to scour far and wide for candidates. Just two or three of a reasonable size should do it.

Agreed. If this challenge only lasts a few more days then the surge will fade quickly. And now that RichD’s OPN is enqueued in 15e that queue may be good to go.

I just like the idea of a G208 but it may be too difficult for 16f_small. Perhaps a team sieve? It is a record poly found over 4 years ago but never sieved.