Reserved for MF - Sequence 3366

[VBCurtis]

Here's my poly:

n: 49166124173619026753031982944551526151417771554402090661062654848708821920824797208853190175969799268557269709859566937412206672909073923990203898320874417048790184714368584420136913755996169900623
skew: 17728574.908
c0: -1851954754326085803755824978566355349385603620
c1: 439108109299249942599064037912565582401
c2: 37765023384211191595917223581920
c3: -1105329700560709779846731
c4: -131268139226673570
c5: 231462000
Y0: -73356500846738980870400907749624903370
Y1: 2014199892324995594119487
# MurphyE (Bf=8.590e+09,Bg=8.590e+09,area=2.147e+15) = 2.930e-08

cownoise: skew 26085083.77202 score 6.71132226e-15

Charybdis' poly is in the expected range- with first digits 49, this is closer to a c198 than a record-scoring c197 (most record polys occur when the input has first digit 1, for obvious reasons).

I'll give Ed & Sean a day or so to post any competitive polys, and then I'll queue it on my usual server/port for sieving from Q=20-100M. swellman can add it to 15e nfs@home from 100M-up (up? whatever test-sieves suggest).

[VBCurtis]

Originally Posted by charybdis

Ran 5e7-15e7 with P=6M, same parameters as Curtis. This is the best I got:
Cownoise score is 7.35059861e-15.

How close is the CADO-score of the second-best poly from your search? We ought to test-sieve your top few- maybe anything within 4% of the best one, so 5.6e-9 and up? I haven't done much testing of CADO's scoring & ranking algos, this is maybe the right size (in terms of reward vs labor) to see how much 'error' there is in the rankings.

[charybdis]

Originally Posted by VBCurtis

How close is the CADO-score of the second-best poly from your search? We ought to test-sieve your top few- maybe anything within 4% of the best one, so 5.6e-9 and up? I haven't done much testing of CADO's scoring & ranking algos, this is maybe the right size (in terms of reward vs labor) to see how much 'error' there is in the rankings.

Second best is:

n: 49166124173619026753031982944551526151417771554402090661062654848708821920824797208853190175969799268557269709859566937412206672909073923990203898320874417048790184714368584420136913755996169900623
skew: 28999185.753
c0: -825891109443158607004674603693307345947535104
c1: -458336107143097451651138666322102677926
c2: -15395815906228156284144085519903
c3: 1227648733060542278973221
c4: 2239142232086382
c5: -240267720
Y0: -52771582291447988764073188957896143899
Y1: 504323638911825938080187
# MurphyE (Bf=4.295e+09,Bg=4.295e+09,area=1.342e+16) = 5.587e-09

Cownoise says 6.99743139e-15.

Third best might not be worth testing, but here it is anyway:

n: 49166124173619026753031982944551526151417771554402090661062654848708821920824797208853190175969799268557269709859566937412206672909073923990203898320874417048790184714368584420136913755996169900623
skew: 24014871.336
c0: -1364797159130352880052679500643894828665462784
c1: -362357224006072980836162526585169670714
c2: 36967799155258551320570826109413
c3: 1576718934633143220496399
c4: -35837161535537914
c5: -269392200
Y0: -51577723996775865772212841676739840845
Y1: 74262126350696286612779
# MurphyE (Bf=4.295e+09,Bg=4.295e+09,area=1.342e+16) = 5.448e-09

This scores similar to yours: 6.74558299e-15.

[swellman]

Originally Posted by VBCurtis

I'll give Ed & Sean a day or so to post any competitive polys, and then I'll queue it on my usual server/port for sieving from Q=20-100M. swellman can add it to 15e nfs@home from 100M-up (up? whatever test-sieves suggest).

Sorry, no bandwidth. Don’t wait for me. I’ll enqueue the eventual winning poly once a breakpoint for Q between CADO and 15e is determined.

[EdH]

Good finds! I'm having troubles (switch related - maybe heat) that are crashing my CADO-NFS runs. I will probably not join the sieving in the near term. I'm finding one machine can hang the whole "farm." My best recent scores were 6.553 and 6.695. I don't expect anything better than what was found.

[EdH]

Originally Posted by VBCurtis

How close is the CADO-score of the second-best poly from your search? We ought to test-sieve your top few- maybe anything within 4% of the best one, so 5.6e-9 and up? I haven't done much testing of CADO's scoring & ranking algos, this is maybe the right size (in terms of reward vs labor) to see how much 'error' there is in the rankings.

I'm confused, but my CADO-NFS scores are above 7e-09. In case you want them:

n: 49166124173619026753031982944551526151417771554402090661062654848708821920824797208853190175969799268557269709859566937412206672909073923990203898320874417048790184714368584420136913755996169900623
skew: 25821086.753
c0: -11743468923861695132221057378748091237266929053
c1: 822195130408646047596119796224044539797
c2: 171274131595837393197610465587927
c3: -1460057612578687511406813
c4: -124782639869557370
c5: 1020419400
Y0: -47464778048999806071817527973227339466
Y1: 655097728475803123967
# MurphyE (Bf=8.590e+09,Bg=4.295e+09,area=2.684e+16) = 7.590e-9
# f(x) = 1020419400*x^5-124782639869557370*x^4-1460057612578687511406813*x^3+171274131595837393197610465587927*x^2+822195130408646047596119796224044539797*x-11743468923861695132221057378748091237266929053
# g(x) = 655097728475803123967*x-47464778048999806071817527973227339466

cownoise 6.553

n: 49166124173619026753031982944551526151417771554402090661062654848708821920824797208853190175969799268557269709859566937412206672909073923990203898320874417048790184714368584420136913755996169900623
skew: 17278493.78
c0: 1912123293028593031857952411951842469496728494
c1: 364715500413933850521068886544044195217
c2: 24250640943297613801121770438218
c3: 2542815473406655921784837
c4: -188210066225793606
c5: -1276589160
Y0: -47070976904761820131596894873426724724
Y1: 9742264672163558165951
# MurphyE (Bf=8.590e+09,Bg=4.295e+09,area=2.684e+16) = 7.695e-09
# f(x) = -1276589160*x^5-188210066225793606*x^4+2542815473406655921784837*x^3+24250640943297613801121770438218*x^2+364715500413933850521068886544044195217*x+1912123293028593031857952411951842469496728494
# g(x) = 9742264672163558165951*x-47070976904761820131596894873426724724

cownoise 6.695

[charybdis]

Originally Posted by EdH

I'm confused, but my CADO-NFS scores are above 7e-09. In case you want them:

From the CADO c90.params file:

    # Note: now I is used also in the polynomial selection (for the computation
    # of Murphy's E value) thus we define it at the "tasks" level, not at the
    # "tasks.sieve" level.

So the CADO poly score depends on the sieve parameters you choose.

[VBCurtis]

If anyone cares, the usual server & port is running Charybdis' best scoring poly with A=31, 4 threads per process default.

I'll get his second-best fired up shortly on a different port and see how ETAs compare in the morning (lazy man's test-sieve). 20 threads on each, on a 20-core machine.

[EdH]

Originally Posted by charybdis

From the CADO c90.params file:

    # Note: now I is used also in the polynomial selection (for the computation
    # of Murphy's E value) thus we define it at the "tasks" level, not at the
    # "tasks.sieve" level.

So the CADO poly score depends on the sieve parameters you choose.

Does it depend on more than just the siever? The wording seems to say siever. Was my I=15 different from yours, or was it the other values that gave me larger CADO scores than your polys with higher cownoise scores?

[charybdis]

Originally Posted by EdH

Does it depend on more than just the siever? The wording seems to say siever. Was my I=15 different from yours, or was it the other values that gave me larger CADO scores than your polys with higher cownoise scores?

I think it might depend on lpb0 and lpb1 too? They correspond to Bf and Bg in the line

# MurphyE (Bf=4.295e+09,Bg=4.295e+09,area=1.342e+16) = 5.807e-09

[VBCurtis]

Yep! My records suggest it's large-prime-size choice and siever area, but lim's might go into the calculation of Bf and Bg too.