kaidokert / modmath-rs

99%

push

github

Add wide_montgomery_mul_acc for fused inner products (#19)

* Add wide_montgomery_mul_acc for fused inner products

Adds a wide multiply-accumulate primitive that produces a*b as a
double-width pair and folds it into a caller-held (lo, hi) accumulator,
returning the new accumulator without reduction. Pair with wide_redc
once at the end to get a fused inner product / dot product / NTT
matrix-vector multiply that pays one REDC per output coefficient
instead of one per multiply.

279 of 279 new or added lines in 2 files covered. (100.0%)

4892 of 4964 relevant lines covered (98.55%)

1999.08 hits per line