msieve_obj_new and flags

[davidknippers]

Hello!

I've been using the msieve library (wonderful stuff! Thanks jasonp_sf, for your wonderful work!) to calculate aliquot sequences, and my question is as follows:

If I set no flags for msieve_obj_new (other than to disable logging to file and standard output), will the library pick the most appropriate factorization method for the size of the input (QS, ECM, NFS, etc) and produce a complete prime factorization (no composites in my factor list)?

I've been playing around with the demo application, and I'm still a bit confused as to the type of factors I get back and the factorization method used when I change the bitmask of the flag. It never behaves like I think it should...

What should the behavior of the following flag be in msieve_obj_new:

uint32 flags;
flags &= ~(MSIEVE_FLAG_USE_LOGFILE | MSIEVE_FLAG_LOG_TO_STDOUT);

Sorry if this is a bit confusing, I can rephrase my question if necessary.

Thanks!

[davidknippers]

Think I figured it out after playing with it some more. All I had to do was read the instructions.

[jasonp]

To answer your questions anyway, the default behavior is to factor any input completely, using whatever method is appropriate at any given time based on the factors that have been found so far.

That's modified by the fact that you won't run the number field sieve unless you specifically ask for it, but if you want NFS then you should be using YAFU which should call it at the optimal time. Actually YAFU's QS code and general tuning are a lot better than the versions in msieve, which honestly haven't changed in years.

Whether or not you allow a logfile, factors are always returned to the obj as a linked list of decimal strings in order of ascending size.

Glad my work has been helpful. Note that if you haven't visited the aliquot subforum you should definitely do so, as folks there already have a lot of tools for generating and processing aliquot sequences without having to write low-level factorization code.