I confirm Fredrik diagnostic sage: type(M1) <class 'sage.matrix.matrix_generic_dense.Matrix_generic_dense'>
It is a pity that flint fmpz_mod_mat and mpn_mod_mat are not interfaced (and used) in Sagemath! On Sat, 28 Mar 2026 at 15:02, Fredrik Johansson <[email protected]> wrote: > > On Saturday, March 28, 2026 at 11:53:40 AM UTC+1 Georgi Guninski wrote: > > I believed that matrix multiplication and squaring to be significantly > faster over GF(p) instead over ZZ. > > > Numerical evidence supports the opposite, is it a bug or reality? > > Session: > > def randmat(N,K): return Matrix(K,N,N,[K.random_element() for _ in > range(N^2)]) > N=200 > Kq=GF(next_prime(2^200)) > M1=randmat(N,Kq) > %time MsZZ=M1.change_ring(ZZ)^2 #Wall time: 188 ms > %time Msqu=M1^2 #Wall time: 7.75 s > > > I would expect matrix multiplication to be around as fast over GF(p) and ZZ > if the entries are the same. Large-p GF(p) matrices being a lot slower in > Sage is surely because they use a generic matrix implementation rather than a > specialized one. > > With current FLINT on my machine squaring a 200x200 matrix with 200-bit > entries costs as follows: > > 0.032 seconds over ZZ with fmpz_mat > 0.035 seconds over GF(p) with fmpz_mod_mat > 0.029 seconds over GF(p) with mpn_mod_mat > > Fredrik > > > -- > You received this message because you are subscribed to the Google Groups > "sage-devel" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion visit > https://groups.google.com/d/msgid/sage-devel/5bd108cd-ca0c-4b45-8193-dc1f51b010e2n%40googlegroups.com. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/sage-devel/CAGEwAA%3DwQdBEon1mY6ERxSkfV69-uj05wBX%3Dpq3RY%3DHU382ZLw%40mail.gmail.com.
