BOINC NFS sieving - NFS@Home

[frmky]

Originally Posted by VBCurtis

I agree that the data suggests a dead heat for sieving time on 34/34 vs 33/34, so we should go with the smaller LP bound for disk space and expected matrix difficulty reasons. Maybe two digits bigger would call for 34/34.

If 34/34 doesn't save sieving time, then let's use 33/34. I'm not so worried about disk space or matrix difficulty, but the smaller bounds will be a bit faster in sqrt. And yes, I can do the LA.

[swellman]

Alright I will enqueue this job on 16f_small as a 33/34 later this week.

And thank you Greg for agreeing to post process it, much appreciated.

[pinhodecarlos]

Hogmanay Hootenanny 2022 - challenge

https://www.boincstats.com/stats/cha...team/chat/1098
Name Hogmanay Hootenanny 2022
Status Upcoming
Project NFS@Home
Issued by The Scottish Boinc Team
Start time 2021-12-31 12:00 UTC
End time 2022-01-07 12:00 UTC
Late entrants allowed? Yes
Number of teams participating » 9
Number of users participating 0

[swellman]

Carlos - when should we have new work in place for this new BOINC challenge? By 20 Dec?

I believe we are good with 16_small as there’s enough work loaded in that queue to carry us for a few months. There are only 9 teams participating though I realize a single team can bring thousands of cores to the party.

15e and 15e_small are the queues where we should stockpile. 14d jobs are like throwing paper on a bonfire - poof and they’re gone.

[pinhodecarlos]

At least 7 days before plus grace period, maybe 8 days before.
Still premature the number of teams but don’t expect huge participation.

[chris2be8]

I may have finished the Brent tables by then, there are only 8 more numbers to do. But if there are any left I'll add them to the queue.

[frmky]

I added beta Linux x86_64 binaries that support octics. I'll watch it for the next day or so, and if it looks good I'll take them out of beta. If you want to try them now, enable running test applications in your project preferences. https://escatter11.fullerton.edu/nfs...subset=project

[pinhodecarlos]

Have an issue. Had a laptop charger failure and only after almost two week I’ve replaced it and now I cannot get 16eV5 wus only the 16e small ones. Is this related to the fact that my computer “ Fraction of time BOINC is running” is close to 40% and it needs to raise up to a certain threshold, to be a more trustful computer. I’ve also update the project for the update link.

[frmky]

The mysteries of the BOINC scheduler baffle me. I see you have that computer set to the home venue, which is set to prefer 16e V5, so that all looks right. You can either try detaching and reattaching the project, or just wait it out a bit and see if it fixes itself.

[pinhodecarlos]

Independently of the way I’ve setup the wus type for all profiles I fail to get more than one 16eV5 wu. My hope is the only one sitting on my queue will trigger something on the serve side. Will wait a few more days then I will try detaching and reattaching to project. Thank you.

[charybdis]

Originally Posted by frmky

I added beta Linux x86_64 binaries that support octics. I'll watch it for the next day or so, and if it looks good I'll take them out of beta. If you want to try them now, enable running test applications in your project preferences. https://escatter11.fullerton.edu/nfs...subset=project

If you want something to try this out on, the octic for 2,1180+ (diff. 285) sieves ~3 times faster than the quartic. Hard to do an exact comparison because yield drops off so much faster for the octic, but it should be very easily within reach. There's also 2,2350M (diff. 283) and 2,2390L (diff. 288) from the original base-2 tables.

Edit: these should of course be done with lpb 33/34, 33/35 or 34/35 rather than 34/34.