mersenneforum.org Reserved for MF - Sequence 3366
 Register FAQ Search Today's Posts Mark Forums Read

 2021-08-24, 08:30 #485 kruoli     "Oliver" Sep 2017 Porta Westfalica, DE 647 Posts Next on is a C201. I finished around one and a quarter t55 on it. t60 would take me more than two weeks, so that's heavier than I can lift in a reasonable amount of time, it seems. How high do we want to go? YAFU said t62? Any help is appreciated!
2021-08-25, 00:43   #486
swellman

Jun 2012

3,203 Posts

Quote:
 Originally Posted by kruoli Next on is a C201. I finished around one and a quarter t55 on it. t60 would take me more than two weeks, so that's heavier than I can lift in a reasonable amount of time, it seems. How high do we want to go? YAFU said t62? Any help is appreciated!
I use 0.31*GNFS difficulty to get the ECM test metric, which is t62 in this case. After this week, I can contribute some t60 cycles if you still need them then.

 2021-08-25, 01:58 #487 VBCurtis     "Curtis" Feb 2005 Riverside, CA 116008 Posts My vote (not that you asked for it) is half a t60, then a t60 worth of curves at the t65 level. C200ish jobs are a whole lot easier and faster than they used to be, so we need less ECM than we used to need.
2021-08-26, 16:04   #488
kruoli

"Oliver"
Sep 2017
Porta Westfalica, DE

647 Posts

Quote:
 Originally Posted by swellman I can contribute some t60 cycles if you still need them then.
I'd really appreciate that! On Sunday, I'll likely have finished around 15k-18k curves @ B1=260M and 400-500 @ B1=850M from another machine.

After doing ECM and assuming no factors, is this a job for team sieving or for some of the NFS@home queues? In the first case, I'd volunteer for hosting a CADO server and to do the LA in the end. I might need some help with the initial setup since I never ran a CADO server with computers all over the world, only locally. But I am going too far into the future right now, I guess. There is still a good chance that we hit something with ECM until then.

@VBCurtis: I always appreciate input, especially because I have a lot to learn in this field. I guess "t60 worth of curves" refers to the amount of 260M curves?

Last fiddled with by kruoli on 2021-08-26 at 16:04 Reason: Misspelling names is a felony.

 2021-08-26, 16:44 #489 VBCurtis     "Curtis" Feb 2005 Riverside, CA 499210 Posts If you invoke ecm with the -v flag, you'll get a table of curve counts for each t-value in multiples of 5 digits. So, ecm -v 85e7 will tell you how many curves at B1=850M is worth a t60. Doing so shows me 15391 curves, so my advice was 15k curves at 850M after half a normal t60. Edit: C201 is simple for the 16-small queue, but if you'd like the experience of hosting a team CADO sieve I'm willing to run clients and advise on management. Should be ~30 core-years of sieving. You'll need 64GB ram for the postprocessing, or plenty of experience using msieve postprocessing after CADO sieving. Last fiddled with by VBCurtis on 2021-08-26 at 16:46
2021-08-26, 18:39   #490
charybdis

Apr 2020

17·29 Posts

Depending on the progress of the 4788 poly search, I may be able to run some curves at 850M next week.

Quote:
 Originally Posted by VBCurtis You'll need 64GB ram for the postprocessing, or plenty of experience using msieve postprocessing after CADO sieving.
Come on, it's not too hard to convert the poly format and cat all the relations into a msieve.dat.gz file
Msieve postprocessing should fit within 32GB RAM.

 2021-08-26, 20:14 #491 kruoli     "Oliver" Sep 2017 Porta Westfalica, DE 647 Posts Thanks for the explanation, I now understand what you meant originally. The RAM reqirements can be met. How much efficiency could be gained by running msieve for LA instead of CADO?
2021-08-26, 20:44   #492
VBCurtis

"Curtis"
Feb 2005
Riverside, CA

27×3×13 Posts

Quote:
 Originally Posted by charybdis Come on, it's not too hard to convert the poly format and cat all the relations into a msieve.dat.gz file Msieve postprocessing should fit within 32GB RAM.
Once is "plenty"! :) (sorry for my bad word choice originally)

For the same-sized matrix, msieve LA will run in maybe 50-60% of the time CADO runs. But, CADO filtering makes smaller matrices- for the C207 team sieve, CADO created a matrix around 61M dimensions while msieve on the same data set was 73M dimensions. So, that ate up about half the speed advantage (in very rough terms- I think msieve was 11 weeks while CADO would have been 15ish weeks).

2021-08-26, 22:10   #493
swellman

Jun 2012

3,203 Posts

Quote:
 Originally Posted by kruoli I'd really appreciate that! On Sunday, I'll likely have finished around 15k-18k curves @ B1=260M and 400-500 @ B1=850M from another machine.
Ok I’ll start some ECM at B1=260M this weekend. I can do about 2k curves per day.

Quote:
 Originally Posted by charybdis Depending on the progress of the 4788 poly search, I may be able to run some curves at 850M next week.
Glad you can play. Personally I don’t care who finds a factor or at what level, but if you start running ECM at 850M early next week just be aware that we will still be running ECM at the t60 level. Race to the finish!

 2021-08-27, 07:50 #494 bur     Aug 2020 79*6581e-4;3*2539e-3 2×199 Posts While msieve is faster at LA, I often had the problem that cado could build a matrix while msieve couldn't and required more relations. Depending on the circumstances that might make it faster to use cado for everything, especially if using just a few cores for sieving.
 2021-08-29, 00:54 #495 swellman     Jun 2012 3,203 Posts I’ll start ECM with B1=260M shortly. I’ll shoot for 10000 curves, then we can regroup later in the week and see where we are. Using Code: 131059381160969890473184300494609882864431978456309837392557442177052630795870868382351497139096300267026719895523089453234540027837272339241326353564822104741922921032481193771448768282517397405699723

 Similar Threads Thread Thread Starter Forum Replies Last Post sweety439 And now for something completely different 17 2017-06-13 03:49 RichD Aliquot Sequences 36 2013-11-29 07:03 petrw1 Lone Mersenne Hunters 82 2010-01-11 01:57 roger Puzzles 16 2006-10-18 19:52 Citrix Puzzles 5 2005-09-14 23:33

All times are UTC. The time now is 01:19.

Mon Oct 18 01:19:50 UTC 2021 up 86 days, 19:48, 0 users, load averages: 1.54, 1.41, 1.40