1 #include "tommath_private.h"
2 #ifdef BN_S_MP_MUL_DIGS_FAST_C
3 /* LibTomMath, multiple-precision integer library -- Tom St Denis */
4 /* SPDX-License-Identifier: Unlicense */
6 /* Fast (comba) multiplier
8 * This is the fast column-array [comba] multiplier. It is
9 * designed to compute the columns of the product first
10 * then handle the carries afterwards. This has the effect
11 * of making the nested loops that compute the columns very
12 * simple and schedulable on super-scalar processors.
14 * This has been modified to produce a variable number of
15 * digits of output so if say only a half-product is required
16 * you don't have to compute the upper half (a feature
17 * required for fast Barrett reduction).
19 * Based on Algorithm 14.12 on pp.595 of HAC.
22 mp_err s_mp_mul_digs_fast(const mp_int *a, const mp_int *b, mp_int *c, int digs)
24 int olduse, pa, ix, iz;
26 mp_digit W[MP_WARRAY];
29 /* grow the destination as required */
30 if (c->alloc < digs) {
31 if ((err = mp_grow(c, digs)) != MP_OKAY) {
36 /* number of output digits to produce */
37 pa = MP_MIN(digs, a->used + b->used);
41 for (ix = 0; ix < pa; ix++) {
44 mp_digit *tmpx, *tmpy;
46 /* get offsets into the two bignums */
47 ty = MP_MIN(b->used-1, ix);
50 /* setup temp aliases */
54 /* this is the number of times the loop will iterrate, essentially
55 while (tx++ < a->used && ty-- >= 0) { ... }
57 iy = MP_MIN(a->used-tx, ty+1);
60 for (iz = 0; iz < iy; ++iz) {
61 _W += (mp_word)*tmpx++ * (mp_word)*tmpy--;
66 W[ix] = (mp_digit)_W & MP_MASK;
69 _W = _W >> (mp_word)MP_DIGIT_BIT;
79 for (ix = 0; ix < pa; ix++) {
80 /* now extract the previous digit [below the carry] */
84 /* clear unused digits [that existed in the old copy of c] */
85 MP_ZERO_DIGITS(tmpc, olduse - ix);