nmod_mat_mul_classical

The dot products use the parameters of _nmod_vec_dot_params for length $k$ (Dot products modulo a word-size integer). The code takes one of three forms.

• If the unreduced sums fit in one word and all dimensions exceed 10, $B$ is transposed and several of its entries are packed into each word, as many as fit with $\lceil \log_2 (k (n-1)^2 + 1) \rceil$ bits each, so that one word product computes several dot products at once.
• If $k < 20$, or if $m < 20$ and neither $m \ge 4, k \ge 256$ nor $m \ge 2, k \ge 1000$, the dot products read the columns of $B$ in place with NMOD_VEC_DOT.
• Otherwise $B$ is transposed into a temporary matrix (NMOD_MAT_MUL_TRANSPOSE_CUTOFF is 20) and the dot products use _nmod_vec_dot on contiguous rows.

@misc{flint-doc-nmod-mat-nmod-mat-mul-classical,
  author       = {{The FLINT team}},
  title        = {{nmod\_mat\_mul\_classical (nmod\_mat.h)}},
  howpublished = {FLINT: Fast Library for Number Theory, documentation, version 3.7.0-dev},
  year         = {2026},
  url          = {https://flintlib.org/doc/nmod_mat/nmod_mat_mul_classical.html},
  note         = {Generated 2026-11-02}
}