Polynomial Request Thread

[XYYXF]

C173_139_59

34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721

Thanks in advance :)

[Dubslow]

I'll run C173_139_59 through CADO's full polyselect phase. ETA ~2 days.

[wombatman]

Is there a good way to sort the optimized roots output by msieve to the .p file? My current run looks to have a couple of good polynomials, but I'm hoping I won't have to go digging through the .p file to find polys that aren't the "best" by e-score.

[chris2be8]

Originally Posted by wombatman

Is there a good way to sort the optimized roots output by msieve to the .p file?

On Linux:

 grep alpha msieve.dat.p | sort -g -k 7 | tail

will get the best 10 scores (or tail -20 for the best 20 etc). Then

 grep -A 9 8.349e-09 msieve.dat.p

for each score (8.349e-09 above) will get the poly for that score (replace -A 9 with -A 10 if you are creating sextics).

Chris

[wombatman]

Thanks very much!

[wombatman]

For the C165:

Best:

expecting poly E from 7.15e-013 to > 8.23e-013
R0: -18501216759147940907566438431336
R1: 310889683329527
A0: -244538593936315670244006535150142290375
A1: -104961390854368280791726556309896
A2: 249303157764270728264644015
A3: -234959794554141918452
A4: 33303098790468
A5: 48725040
skew 1834679.80, size 6.584e-016, alpha -8.423, combined = 8.956e-013 rroots = 3

4 others that also fell in the expected E range:

# norm 6.387581e-016 alpha -7.945035 e 7.787e-013 rroots 3
skew: 19694169.03
c0: -94150349733298768589216580366479523599864
c1: 6932485644850450503947410493206181
c2: 620411484096805424417627253
c3: -171277681769170053341
c4: 160397020136
c5: 188700
Y0: -56181890915510328202545396702213
Y1: 1437139919714719

# norm 6.095134e-016 alpha -6.833837 e 7.503e-013 rroots 5
skew: 8965292.60
c0: 304072707986022432422564443598606812251
c1: 1755431229943558060522571273475981
c2: -122990835297584071621920747
c3: -166304690853238806181
c4: 1540044047636
c5: 188700
Y0: -56181890913408848797843753123678
Y1: 1437139919714719

# norm 5.678878e-016 alpha -7.719373 e 7.183e-013 rroots 5
skew: 2344641.41
c0: 232141936819381380873307868100123002299
c1: 1549369138542319412996710966340618
c2: -503401322792822515009686837
c3: -969604934067811874312
c4: 101801290691084
c5: 41081040
Y0: -19143548648846981506165015026546
Y1: 943859578994221

# norm 5.643528e-016 alpha -7.001131 e 7.167e-013 rroots 5
skew: 15383886.76
c0: 29265308953953929596949889774837373022016
c1: 13579754341672337664716174805076986
c2: 45487258929410961727758198
c3: -168160227689950826461
c4: 1223268640136
c5: 188700
Y0: -56181890913891361340188371454333
Y1: 1437139919714719

[swellman]

Originally Posted by Dubslow

I'll run C173_139_59 through CADO's full polyselect phase. ETA ~2 days.

Is it worth also running this number through gpu msieve and comparing results? No doubt CADO will produce a very usable poly, just wondering where the msieve vs CADO line is in terms of poly efficiency, i.e. at what GNFS level is CADO "better" than msieve?

[VBCurtis]

Yes, it's worth doing with msieve also. I haven't seen CADO produce a degree-5 poly better than msieve yet, given similar computrons of effort for each. CADO's are comparable, but often score 5-10% lower; say, similar to a 3rd-best msieve poly from a typical run.

[swellman]

Can someone with gpu msieve run a poly search for C173_139_59? I do not currently have the capability or I would help. Thank you in advance.

[Dubslow]

Originally Posted by VBCurtis

Yes, it's worth doing with msieve also. I haven't seen CADO produce a degree-5 poly better than msieve yet, given similar computrons of effort for each. CADO's are comparable, but often score 5-10% lower; say, similar to a 3rd-best msieve poly from a typical run.

Perhaps, but you're comparing GPU-msieve to CPU-cado. At least on the most recent one I did with wombatman, cado was competitive and possibly marginally better than msieve, and I'm only using a HT SandyBridge quad core (I have no idea what GPU wombatman is using, though I do assume it's a GPU).

[wombatman]

I'm using a GTX 980Ti. I think VBCurtis is arguing that msieve's polynomials test-sieve at a better rate than CADO's do, even if CADO's murphy score is higher or similar.