Poly select and CADO sieving for 2,1165+ C217

[swellman]

Originally Posted by VBCurtis

I neglected to list adrange: this should be 2x or 4x incr, since poly select runs 2-threaded. That is, for incr = 420, using adrange = 840 gives one c6 value to each thread, while using adrange = 1680 gives two values to each thread per workunit. I've usually used 4x incr (=1680 here), figuring it reduces idle thread time a bit.

Also, ropteffort must be set. The root-opt phase is tiny compared to the main size-opt search; a value like 35 or 40 "should" be effectively "try as hard as you can". Number of polys to root-opt (nrkeep, I think, in param-speak) can be quite small for your 0.5M range, like 30 to 60. I think CADO defaults to like 200 for an entire search of this size, while I used 36 for each of my two runs.

Thanks for the helpful advice. I added ropteffort = 35 though the default was 0.2(?), and all the rest you mentioned in your previous two posts. For my fist try, I cut the admax to 2.1e6, just to make sure I didn’t miss anything.

Did you use default qmin? I think it’s 4e7 (or 4e8) by default, but I left it alone for this run.

[VBCurtis]

qmin doesn't affect poly select, since it merely tells the siever what Q-value to begin at. I think I changed it to 10M anyway. I did set sieve params to values I might choose for a test-sieve, so your CADO score will be different from mine (but cownoise fixes that for us).

If you add tasks.sieve.run=false to the command line, CADO will exit after poly select without building factor base/starting to sieve.

[swellman]

Originally Posted by VBCurtis

qmin doesn't affect poly select, since it merely tells the siever what Q-value to begin at. I think I changed it to 10M anyway. I did set sieve params to values I might choose for a test-sieve, so your CADO score will be different from mine (but cownoise fixes that for us).

If you add tasks.sieve.run=false to the command line, CADO will exit after poly select without building factor base/starting to sieve.

Thanks about qmin. And yes, I used tasks.sieve.run=false in my command invocation. Also used P = 12M in the parameters.

-break-

Do you have any feel for CADO’s deg = 5 performance? IIRC, one gets better deg = 5 results with msieve-GPU correct? I’m estimating a minimum search range of 35M < c5 < 55M for msieve-GPU but that’s based on extrapolating my results to date so take that estimate with a grain of salt.

I have yet to run any msieve-GPU poly search for this c217 though I hope to start early in 2020 after finishing my current HCN efforts.

[VBCurtis]

I think they're competitive in degree 5; I won't talk folks out of using msieve-gpu, but when my CUDA install broke last spring I didn't bother fixing it and just shifted to using CADO for all my personal poly-select needs. If I had a GPU faster than a 750ti, I imagine I'd still be putting effort into using msieve.

[Gimarel]

Originally Posted by VBCurtis

Behold:

skew: 2382635.076
c0: 3987095627700723668556412616355906815127705992
c1: 10621479627429548899812810235604051573030
c2: -4682349569391313556729607478081235
c3: -18433207794474459950952563135
c4: 1015293418294354295928
c5: 1005061270114440
c6: -39916800
Y0: -110058966911042321282835020413351605
Y1: 14771789483157557668102843
# MurphyE (Bf=3.436e+10,Bg=1.718e+10,area=2.147e+16) = 1.201e-08

Cownoise: skew 2581491.68923 score 6.25337175e-16

Msieve found a small improvement:

# norm 1.602538e-15 alpha -10.543517 e 6.454054e-16 rroots 6
skew: 3226967.46
c0: -2543005241608251659326321959460452765214183842
c1: -33139987534593547344167381165622666435255
c2: 15215551779833751980190681203957330
c3: 17257164061453212524534319695
c4: -1977227853181703834928
c5: -958120310818440
c6: 39916800
Y0: -110056071714162569817319860596637820
Y1: 14771789483157557668102843

[swellman]

My best result:

n: 3213773553731676734673029767649076184026081109505643819433123522398392216242493939130914684819990019996494667995485530318906995221039607546449005638410444687839629323036119079997039490112821634482150496640217083568211
skew: 135395.269
c0: -731803863642148184907152920695855624679056
c1: 7082389535680844516273731959418297544
c2: -215202423296858113904589199612942
c3: 314236139762316178602510253
c4: 37515382402362600715141
c5: -3515549390483760
c6: 1519015680
Y0: -108149603432644423078656532479187581
Y1: 18837068487969546705833869
# MurphyE (Bf=3.436e+10,Bg=1.718e+10,area=8.590e+17) = 1.751e-09
# f(x) = 1519015680*x^6-3515549390483760*x^5+37515382402362600715141*x^4+314236139762316178602510253*x^3-215202423296858113904589199612942*x^2+7082389535680844516273731959418297544*x-731803863642148184907152920695855624679056
# g(x) = 18837068487969546705833869*x-108149603432644423078656532479187581

Cownoise says 257289.42671 3.99079450e-16 so not a high scoring poly but at least I’m getting CADO to produce something.

Will continue my search to admax = 2.5e6.

ETA: Now running ad = 2.1 to 2.5e6 with P=14M.

[Dylan14]

I'll do a deg 6 poly search for this on CADO, with admin = 5e6, admax = 5.5e6 and P = 16e6 after I finish 47771_188.

[Dylan14]

I just realized I also have to finish an msieve matrix for NFS@Home as well, so I'm cancelling my run.

[VBCurtis]

Taking 3e6 to 5e6 for CADO deg 6 with P=12M and other params as listed above.

[swellman]

admin = 2.1 and admax = 2.5e6 with P=14M generated a disappointing result:

n: 3213773553731676734673029767649076184026081109505643819433123522398392216242493939130914684819990019996494667995485530318906995221039607546449005638410444687839629323036119079997039490112821634482150496640217083568211
skew: 559386.705
c0: 783269053626373507790443787773958091992171250
c1: -7541006308544676935957788588704394135191
c2: 7450178083390142379828586195307731
c3: 67582585756453219302354745929
c4: -93930118344551034304789
c5: -113304614208890370
c6: 14630616000
Y0: -104993607429839019815662076500592188
Y1: 18938768306573546433782113
# MurphyE (Bf=3.436e+10,Bg=1.718e+10,area=8.590e+17) = 1.918e-09
# f(x) = 14630616000*x^6-113304614208890370*x^5-93930118344551034304789*x^4+67582585756453219302354745929*x^3+7450178083390142379828586195307731*x^2-7541006308544676935957788588704394135191*x+783269053626373507790443787773958091992171250
# g(x) = 18938768306573546433782113*x-104993607429839019815662076500592188

Cownoise says skew of 575076.46750 gives a score of 4.58159900e-16. Nice low skew at least.

Hitting 2.5e6-3e6 (deg 6) next, again with P=14M and other parameters as above.

[VBCurtis]

3e6 to 5e6 on CADO deg 6 complete. Best poly is underwhelming:

skew: 1079681.594
c0: -50457770615217022869877235860673265892222752
c1: -265092486914500987328363191756903620546
c2: -1272306689762274825524465213573345
c3: -1158546181084523419055466666
c4: 1745276163212225808374
c5: 109670409952434
c6: 43201620
Y0: -93528774949199039661242124580223849
Y1: 7310614514345840354266259
# MurphyE (Bf=3.436e+10,Bg=1.718e+10,area=2.147e+16) = 1.051e-08

cownoise: skew 1266113.52169 score 4.89683793e-16

Next I'll take 10M to 30M with incr=4620 (rather than 420 used for the smaller ad-values).