2^877-1 polynomial selection

[axn]

Originally Posted by Andi47

Pol51m0b printed boatloads of lines containing numbers with a "!" to screen (is this something like msieve's "integrator failed" error?) and only a fistful of lines to the .m file, so I guess that I've lost a significant amount of polys to this error. .

Don't worry. That's just the way it works. Question is, how to suppress the screen output?

[smh]

200 done. Nothing worth mentioning

[frmky]

500-600 is done. Here's the best:

BEGIN POLY #skewness 547461.22 norm 5.80e+24 alpha -6.05 Murphy_E 9.20e-14
X5 52101320040
X4 -44407918935106581
X3 -79015216368833586747403
X2 11164378525237099289579014018
X1 5527464674743398811310201276866547
X0 -673974942315034055091970689174411290421
Y1 1493967901901591351
Y0 -2155946141079123383394238690321070
M 11500375157769035299295133381748440806665134708448152341342997737069574639205081055883409910814286223252649$
END POLY

I'll take
frmky 225-300

[fivemack]

Originally Posted by jasonp

I haven't had the time to add Tom's work on poly selection parameters for big jobs, and my experience is that using msieve to help for a job this big would be a waste of effort. On the bright side, degree-4 and -6 poly selection is coming along nicely.

I was quite pleased with the polynomials that I persuaded msieve-1.39 to give me for an 180-digit test example. However, when I try tweaking msieve-1.41 to take the three bounds on the command line rather than from a hard-wired table, and run on M877 with b1=10^27.5, b2=10^27, I just get

error: rational leading coefficient is too large

messages

[jasonp]

Stage 1 of the poly selector cannot generate polynomials when the leading rational coefficient exceeds 2^64; you'll either have to use a larger leading algebraic coefficient or reduce the stage 1 norm. The 109!+1 job was very nearly at the limit of what the current code can handle.

Degree 6 might make this work :)

[fivemack]

frmky's 9.20e-14 polynomial comes close to answering Silverman's question of the second post: it produces more relations per second (though not by very many) than the obvious SNFS polynomial does with the same parameters (picked as optimised for SNFS). I don't know if the extra speed yet compensates for the time taken by the polynomial selection, timing for NFS is not an exact science (oversieving ...) and for distributed NFS even less so.

[frmky]

225-300 produced nothing better. I'll now take
frmky 113-200

[fivemack]

BEGIN POLY #skewness 143041.66 norm 2.94e+23 alpha -3.75 Murphy_E 9.23e-14
X5 21926606100
X4 -31451126262494076
X3 -2259400467300850427219
X2 106567670885434817955395000
X1 20941526259992061474399809393706
X0 -125552384749484687222699803637623825
Y1 2020030177733371421
Y0 -2563382232192727759573274378156942

Given how fast this is going, you might want to change the pol51opt parameters to

-n 2e26 -N 3e23

or similar, which makes that stage a bit slower but might potentially produce more polynomials.

[frmky]

113-200 is done. Nothing of interest found. I'll take
frmky 20-112 excluding 28, 47, 56

[Andi47]

I give it a try with the new msieve 1.42 beta and reserve a range of 1 to 1000.

[Andi47]

Originally Posted by Andi47

I give it a try with the new msieve 1.42 beta and reserve a range of 1 to 1000.

No polynomial found, extending the search up to 6000.