1 #include "tommath_private.h"
2 #ifdef S_MP_MUL_COMBA_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_comba(const mp_int
*a
, const mp_int
*b
, mp_int
*c
, int digs
)
26 mp_digit
MP_ALLOC_WARRAY(W
);
36 /* grow the destination as required */
37 if ((err
= mp_grow(c
, digs
)) != MP_OKAY
) {
42 /* number of output digits to produce */
43 pa
= MP_MIN(digs
, a
->used
+ b
->used
);
47 for (ix
= 0; ix
< pa
; ix
++) {
50 /* get offsets into the two bignums */
51 ty
= MP_MIN(b
->used
-1, ix
);
54 /* this is the number of times the loop will iterate, 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
)a
->dp
[tx
+ iz
] * (mp_word
)b
->dp
[ty
- iz
];
65 W
[ix
] = (mp_digit
)_W
& MP_MASK
;
68 _W
= _W
>> (mp_word
)MP_DIGIT_BIT
;
75 for (ix
= 0; ix
< pa
; ix
++) {
76 /* now extract the previous digit [below the carry] */
80 /* clear unused digits [that existed in the old copy of c] */
81 s_mp_zero_digs(c
->dp
+ c
->used
, oldused
- c
->used
);