Factoring database

[Syd]

Originally Posted by 10metreh

The text format:
1. Could you make it write the full numbers instead of "P86" and the like?
2. Sometimes the factors are written backwards!
3. A big bug: when there is a power of 2 (or another number), the text format will only give "2", instead of "2^2".

Now it looks a little bit more like the .ELF files

[schickel]

Originally Posted by Syd

Hi Frank,

I've added a simple compact text output, is it what you expected?

That's perfect. Awesome job on this.

[mdettweiler]

Workers 13-16 appear to be "offline" according to the worker status page--and if memory serves they've been like that since last night. I've seen this happen before but it was always corrected in somewhat short order; Syd, do you know what's going on here?

[mdettweiler]

Small feature request for the workers: could we have a button for doing P-1 and 3 runs of P+1 all in one step? None of those steps take very long individually, so most of the time spent doing them is the human effort involved in waiting for the one job to finish, then starting the next one, etc.

[smh]

The P-1 and P+1 bound might also be a bit larger than 1M

[Syd]

Originally Posted by mdettweiler

Workers 13-16 appear to be "offline" according to the worker status page--and if memory serves they've been like that since last night. I've seen this happen before but it was always corrected in somewhat short order; Syd, do you know what's going on here?

They were (and are now) offline because I need the CPU time to do some sieving jobs on sequence 162126. Everything is working fine.

The P-1 and P+1 bound might also be a bit larger than 1M

could we have a button for doing P-1 and 3 runs of P+1 all in one step?

Going to change this two.

[smh]

What about a limit on the "ECM very high limits". I think it doesn't make sense to run 4000 curves with B1=1M

[henryzz]

Originally Posted by smh

What about a limit on the "ECM very high limits". I think it doesn't make sense to run 4000 curves with B1=1M

yes a limit of 1000 curves would be good

when doing a sequence search would it be possible in the show pane to have options to show only numbers with the largest composite cofactor below x(a textbox input?) or/and have an option to sort numbers by the size of the remaining cofactor with smallest at the top?

[mdettweiler]

Originally Posted by smh

What about a limit on the "ECM very high limits". I think it doesn't make sense to run 4000 curves with B1=1M

But what if someone wanted to run a number up to the full 40 digit level (9000 curves according to the worker status page)? In that case a limit of 4000 curves (or smaller) would end up being more of an annoying limitation than anything.

Now, possibly a limit of 9000 curves would work better; that would ensure that runaway jobs don't steal too much of Syd's personal time on the worker machines. If users really, really wanted to run more, of course, they still could, but then again, most of the time a very high limits job will be stopped long before 9000 curves anyway due to other jobs coming in.

[smh]

Originally Posted by mdettweiler

But what if someone wanted to run a number up to the full 40 digit level (9000 curves according to the worker status page)? In that case a limit of 4000 curves (or smaller) would end up being more of an annoying limitation than anything.

B1=1M is not optimal for finding 40 digit factors. You should increase to B1=3M.

A small example running a curve with B1=1M and B1=3M on a c106 using th e-v output of GMP-ECM.

B1=1M

Expected number of curves to find a factor of n digits:
20      25      30      35      40      45      50
5       20      118     904     8613    97057   1270662

Expected time to find a factor of n digits:
20      25      30      35      40      45      50
30.64s  2.21m   13.18m  1.69h   16.09h  7.55d   98.89d

B1=3M

Expected number of curves to find a factor of n digits
20      25      30      35      40      45      50
3       11      54      322     2350    20265   199745
Expected time to find a factor of n digits:

20      25      30      35      40      45      50
58.99s  3.41m   16.07m  1.60h   11.67h  4.19d   41.33d

[henryzz]

Originally Posted by smh

B1=1M is not optimal for finding 40 digit factors. You should increase to B1=3M.

A small example running a curve with B1=1M and B1=3M on a c106 using th e-v output of GMP-ECM.

B1=1M

Expected number of curves to find a factor of n digits:
20      25      30      35      40      45      50
5       20      118     904     8613    97057   1270662

Expected time to find a factor of n digits:
20      25      30      35      40      45      50
30.64s  2.21m   13.18m  1.69h   16.09h  7.55d   98.89d

B1=3M

Expected number of curves to find a factor of n digits
20      25      30      35      40      45      50
3       11      54      322     2350    20265   199745
Expected time to find a factor of n digits:

20      25      30      35      40      45      50
58.99s  3.41m   16.07m  1.60h   11.67h  4.19d   41.33d

note that the time for 35 digits is also lower with a bound of 3M