[GiNaC-devel] Switch UniPoly to cl_UP_MI
Jens Vollinga
jensv at nikhef.nl
Tue Sep 30 12:12:56 CEST 2008
Dear Alexei,
Alexei Sheplyakov schrieb:
> You might want to try polynomial/upoly.hpp. Feel free to ask me to implement
> the operations you need, propose any reasonable API changes, etc.
for the next three/four weeks I will not change to upoly/umodpoly. I
want to try more efficient algorithms first, and there is still a
strange bug in the multivariate case that happens just occasionally and
that I need to catch.
For modular polynomials I only need *,+,- and the functions in lines
105-350 in factor.cpp like div/rem/gcd/... You are probably going to
implement them anyway, I guess.
I don't care much about the API as long as it allows for efficient and
elegant usage. Since you are using umodpoly for a modular gcd you have
the same requirements as I do, so just got ahead with it and do it as
you like. Side question: is the cln code (cl_UP_UI) going to improve in
the future or is the development stuck?
Eventually, I'll going to move factor.cpp into the ginac/polynomial
directory. Maybe I will split the modular matrix code into a separate
file as well.
BTW, do you plan to have a multivariate (modular) polynomial class? At
the moment, in factor.cpp I use the silly make_modular() function that
looks at coefficients in a ex and changes them to their ring value.
Dirty hack it is.
Regards,
Jens
More information about the GiNaC-devel
mailing list