mersenneforum.org Planning & Coordination for 2,2330L
 Register FAQ Search Today's Posts Mark Forums Read

 2019-04-20, 16:19 #56 VBCurtis     "Curtis" Feb 2005 Riverside, CA 10001010000102 Posts I'm not so sure about that- for one, it appears Vebis did a *lot* of work to find this poly; second, when we tested polys for XYYX 138_123 C206, the deg 5 sieved about 20% faster than deg 6 of the same score. That is, a rough conversion last time around was to add 20% to the score of deg 5 when comparing to deg 6. That means a 1.60 in deg 5 should be as strong a candidate as Max's spun-up version of Vebis' stellar poly. We expect to be able to find 1.6's if we put in the time Vebis did.
 2019-04-20, 16:27 #57 vebis     Oct 2015 22·17 Posts I started the search for a 5th degree poly. It will be done tomorrow evening.
 2019-04-20, 19:00 #58 RichD     Sep 2008 Kansas 32×349 Posts I'll start back up at 32M to try to get a lower skew. (deg-5)
 2019-04-21, 01:38 #59 VBCurtis     "Curtis" Feb 2005 Riverside, CA 10001010000102 Posts I've fired up CADO to test the deg 6 poly from Vebis as spun by Max. After an hour or so, I have sieved 2 Q=1k workunits on 4 hyperthreads of a busy Ivy Bridge. Stats: Yield is 6.4 at Q=200M ETA suggests 55 core-years of sieving (I chose Q=200M because it's close to the midpoint of expected Q-range of 20M to 450M, so the time estimate is the best a single-point test-sieve can be). Initial LP and MFB are 34/35 and 66/99; I can change these by exiting CADO and editing the snapshot file, which I plan to do daily while trying to minimize ETA. I chose 2.6G raw relations as target; this should be ample over-sieving to build a reasonable matrix. I left lim's and ncurves at default for c210; 260M/550M and 26/21. I am pleased to report memory use from "top" as 13.3GB virtual, 12GB resident. So, these tasks should fit into Carlos' planned virtual-linux-on-windows on his 16GB machine. I test-sieved a bit on 16f GGNFS siever too; with lim's at 250/500M, sec/rel was around 0.350 with the same LP/MFB listed above, while sec/rel was slower with MFBR 67 or MFBA 98 or 100. Yield on 16f was a bit higher, though I only tested Q-range of 500. 55 core-years happens to match our present pledge level nicely, so we have a legit hope to get this sieved in 6 months! I hope that some quiet observers of this thread will contribute a few workunits too, leading us to less than 6 wall-clock months to sieve. :) Last fiddled with by VBCurtis on 2019-04-21 at 01:40 Reason: changed thread to core for ETA estimate
 2019-04-21, 08:12 #60 pinhodecarlos     "Carlos Pinho" Oct 2011 Milton Keynes, UK 474810 Posts Any experience on CADO server side in terms of how many client connections it can handle. Was wondering what would happen if you start being hit by more than 300-400 cores.
2019-04-21, 15:06   #61
R.D. Silverman

Nov 2003

26·113 Posts
Other "small" composites

Quote:
 Originally Posted by RichD I'll start back up at 32M to try to get a lower skew. (deg-5)
Perhaps the forum might want to do a poly search for the C210, C211, and C217?
If the forum provides polynomials perhaps Greg might queue them for lasievef?

2019-04-21, 15:57   #62
VBCurtis

"Curtis"
Feb 2005
Riverside, CA

2×472 Posts

Quote:
 Originally Posted by pinhodecarlos Any experience on CADO server side in terms of how many client connections it can handle. Was wondering what would happen if you start being hit by more than 300-400 cores.
No experience, of course; that's not something a normal person could/would test. However, I can do some estimating: A single 4-threaded workunit is producing ~5000 relations in half an hour of wall-clock time at Q=200M. So, 400 cores would produce 100 such workunits in 30 minutes, or 3 per minute; 500k relations would be produced per 30 minutes. That's substantially slower than running a typical C110 locally, so the server would keep up without any issue.

I'm shopping for a 1TB SSD presently to handle the job data. I am CADOing a C186 right now that should head into filtering in 3 weeks. Once I see what memory use is for postprocessing that number, I can decide whether to upgrade to 96GB or 128GB memory (64GB now) and have a more-educated guess for how much swap to allocate on the SSD.

2019-04-21, 18:08   #63
vebis

Oct 2015

22·17 Posts

Code:
n: 334377437706404684733884220732190564550147039308686365251932207592010868660163280944362994773042289837865752943086095333357861608005044927356022538834108845220872215597190097852981295940487497385774178731621
skew: 146825723.142
c0: 8927615796337795909706360434154979487944027812870
c1: 249426878913680820299934039511380732178383
c2: -1269993042201960025915734295984162
c3: -44702903780233907372762163
c4: 43380588291919832
c5: 199837440
Y0: -6688519829695165238165832445322722085046
Y1: 3276459567865397949373
# MurphyE (Bf=3.436e+10,Bg=1.718e+10,area=5.469e+16) = 1.58e-08
# found by revision 9436c2ff8
# f(x) = 199837440*x^5+43380588291919832*x^4-44702903780233907372762163*x^3-1269993042201960025915734295984162*x^2+249426878913680820299934039511380732178383*x+8927615796337
795909706360434154979487944027812870
# g(x) = 3276459567865397949373*x-6688519829695165238165832445322722085046
My best 5th degree poly. Default c210.poly of cado-nfs, but with degree 5 as parameters for the search. You can find the other polys in the attachment.
Attached Files
 c207-deg5.tgz (563.6 KB, 27 views)

Last fiddled with by VBCurtis on 2019-04-21 at 19:33 Reason: VBCurtis: Changed quote tages to code tags again

 2019-04-21, 19:31 #64 VBCurtis     "Curtis" Feb 2005 Riverside, CA 105028 Posts The cownoise.com skew-maker evaluates this poly: skew: 86201388.01399 E-score: 1.35974390e-13 I'll test this poly on 16f today, but I'll wait for possible spin from Max before I turn CADO to it.
2019-04-22, 03:02   #65
axn

Jun 2003

4,723 Posts

Quote:
 Originally Posted by VBCurtis The cownoise.com skew-maker evaluates this poly: skew: 86201388.01399 E-score: 1.35974390e-13
Which poly? The degree-5 gets evaluated as
skew: 175281707.18300
E: 1.51187373e-15

1.36e-13 couldn't possibly be a legit value - it is about two orders of magnitude better than anything posted here.

2019-04-22, 04:43   #66
VBCurtis

"Curtis"
Feb 2005
Riverside, CA

105028 Posts

Quote:
 Originally Posted by axn Which poly? The degree-5 gets evaluated as skew: 175281707.18300 E: 1.51187373e-15 1.36e-13 couldn't possibly be a legit value - it is about two orders of magnitude better than anything posted here.
Strange that I got such a different value from cownoise for the deg 5; I'll try it again. {time passes}

My result now matches yours. I cannot figure out what I miscopied to generate the other value; I'm not even sure that the -13 was a typo versus the result of some copy/paste getting truncated.

I test-sieved, but with the skew from CADO rather than from my post (error #2 today!); I'll do some more tests presently.

Last fiddled with by VBCurtis on 2019-04-22 at 04:44

 Similar Threads Thread Thread Starter Forum Replies Last Post ixfd64 Lone Mersenne Hunters 53 2020-10-30 19:39 henryzz Cunningham Tables 16 2010-08-07 05:08 gd_barnes No Prime Left Behind 2 2008-02-16 03:28 gd_barnes Conjectures 'R Us 32 2008-01-22 03:09 wblipp ElevenSmooth 2 2004-02-19 05:25

All times are UTC. The time now is 08:29.

Sat Oct 31 08:29:17 UTC 2020 up 51 days, 5:40, 2 users, load averages: 1.26, 1.40, 1.61