Tommy Hofmann
Tommy Hofmann
0.1 is a literal which can not be distinguished from 3602879701896397//36028797018963968. This value can be represented with a ball of radius 0 if the precision is bigger than 53, which...
Acceptable for me.
I see. For `gfp_fmpz_mat` everything should work? We could adjust the test code in Nemo to use residue field instead of residue ring. Just to move forward.
I wrote this comment here after I read and merged the other PR for `fmpz_mod_mat`. I don't think we have `gfp_fmpz_mat` but we should, similar to `nmod_mat` and `gfp_mat`.
@fredrik-johansson Here is a Nemo version for the besselj case: ```julia julia> for p in 100:10:250 CC = AcbField(p); a = CC(2.0); b = CC("57.9668555791947111700112064539202300037282162") @show p, besselj(a, b) end...
I don't think we need universal cutoffs, but it would be good to have the algorithms with better asymptotic complexity. But we don't we to duplicate everything for polynomials and...
No, no one is suggesting this.The suggestion is to have, for example, a mullow implementation working with arrays that can be used by both polynomials and series. This also has...
So what should canonical_unit of a m x n matrix return?
I did not define it for matrices. It is already there as canoncial_unit(M[1, 1]). I just wanted to know how to do it properly for matrices while I fix it...
OK. This is more complicated than I anticipated. Let us ignore the canonical_unit of a matrix for a moment. When writing the generic code Hermite normal forms and stuff like...