Work on a BASH Scipt to Find Polynomials Using the Best of Msieve and CADO-NFS

[EdH]

Originally Posted by Gimarel

I use a starting value relative to the msieve computed stage2_norm of the poly. To do this I modified msieve .
I don't know how to do this without a modified msieve in an efficiant way. My modification to msieve:
. . .
Additionally I run msieve again on the best polys after rootsieve with a constant 1.0327 for 100 times.

I don't understand this last statement. Can you provide a more specific example?

[Gimarel]

Originally Posted by EdH

Update: I now have added the Msieve rootopt for the three best scores from each sopt operation. The rootopt is done with stage2_norm going in steps from 1.0 through 9.9e<from original>. I plan to have a step value variable in case 1 is too fine. From there, the best of each process (thread) will be collected, and eventually, the top (or two or three) of those will be displayed.

I missed that post, sorry.
I think, that the "from original" is way off.

Originally Posted by EdH

I don't understand this last statement. Can you provide a more specific example?

Suppose that msieve calculates the stage2 norm of a poly to 1e25. Then I run rootsieve five times with an increase of 50%, i.e. 1.5e25, 2.25e25, 3.375e25, 5.0625e25 and 7.59375e25. For sizeopt hits that have a good rootsieve result, I repeat that step with a smaller increase of only 3.27% but for a hundred times (1.0327e25, 1.06646929e25, ...).

A way to determine the stage2 norm without patching msieve may be to start rootsieve with a really low stage2 norm and a really low target evalue (like 1e-20) and increase the stage2 norm until rootsieve finds polys. Then you can continue as above. But I guess this procedure will be rather slow.

[Gimarel]

One step of the spinning script is to invert the signs of the algebraic side. I run the rootsieve for both, the original poly and the poly with inverted signs on the algebraic side. Sometimes this produces better results.

And another part of the spinning is to run the rootsieve again on the best results of the rootsieve. I also do this additionally with inverted signs on the algebraic side.

[EdH]

Thanks Gimarel! This will help me try some better things. I have been looking at adding the inverted polys. I've also considered adding the whole spin to the script, maybe combining everything, since much of it is already there.

[EdH]

More questions:

- As I understand, the values entered for the norms are max values, to keep. In the polynomials found, there is a norm value (.dat.p listing). Are these the same norm?

- It would seem a polynomial with the largest norm is chosen by Msieve root opt as best. Do I have something backwards?

- Should I be looking for larger or smaller norms, and are these associated with the exp_E in CADO-NFS?

I've been through the Msieve docs a number of times, but if I should be reading (an)other document(s) for this info, please let me know.

[Gimarel]

Originally Posted by EdH

More questions:

- As I understand, the values entered for the norms are max values, to keep. In the polynomials found, there is a norm value (.dat.p listing). Are these the same norm?

No, these are different.

- It would seem a polynomial with the largest norm is chosen by Msieve root opt as best. Do I have something backwards?

Sometimes a poly with a lower norm has a better score, but most of the time this is correct.

- Should I be looking for larger or smaller norms, and are these associated with the exp_E in CADO-NFS?

I've been through the Msieve docs a number of times, but if I should be reading (an)other document(s) for this info, please let me know.

I don't know how the exp_E in CADO is computed.

[EdH]

Thanks! More study needed, of course. My interest has waned considerably, since, in all my runs, I haven't been finding anything even close to what others have been posting.