1
/* mpn_mullow_n -- multiply two n-limb nunbers and return the low n limbs
4
THIS IS (FOR NOW) AN INTERNAL FUNCTION. IT IS ONLY SAFE TO REACH THIS
5
FUNCTION THROUGH DOCUMENTED INTERFACES. IN FACT, IT IS ALMOST GUARANTEED
6
THAT IT'LL CHANGE OR DISAPPEAR IN A FUTURE GNU MP RELEASE.
8
Copyright 2004, 2005 Free Software Foundation, Inc.
10
This file is part of the GNU MP Library.
12
The GNU MP Library is free software; you can redistribute it and/or modify
13
it under the terms of the GNU Lesser General Public License as published by
14
the Free Software Foundation; either version 2.1 of the License, or (at your
15
option) any later version.
17
The GNU MP Library is distributed in the hope that it will be useful, but
18
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
19
or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
20
License for more details.
22
You should have received a copy of the GNU Lesser General Public License
23
along with the GNU MP Library; see the file COPYING.LIB. If not, write to
24
the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
25
MA 02110-1301, USA. */
31
#ifndef MULLOW_BASECASE_THRESHOLD
32
#define MULLOW_BASECASE_THRESHOLD 0 /* never use mpn_mul_basecase */
35
#ifndef MULLOW_DC_THRESHOLD
36
#define MULLOW_DC_THRESHOLD 3*MUL_KARATSUBA_THRESHOLD
39
#ifndef MULLOW_MUL_N_THRESHOLD
40
#define MULLOW_MUL_N_THRESHOLD 10*MULLOW_DC_THRESHOLD
43
/* Avoid zero allocations when MULLOW_BASECASE_THRESHOLD is 0. */
44
#define MUL_BASECASE_ALLOC \
45
(MULLOW_BASECASE_THRESHOLD_LIMIT == 0 ? 1 : 2*MULLOW_BASECASE_THRESHOLD_LIMIT)
48
FIXME: This function should accept a temporary area.
49
FIXME: Perhaps call mpn_kara_mul_n instead of mpn_mul_n?
50
THINK: If mpn_mul_basecase is always faster than mpn_mullow_basecase
51
(typically thanks to mpn_addmul_2) should we unconditionally use
53
FIXME: The recursive calls to mpn_mullow_n use sizes n/2 (one uses floor(n/2)
54
and the other ceil(n/2)). Depending on the values of the various
55
_THRESHOLDs, this may never trigger MULLOW_BASECASE_THRESHOLD.
56
Should we worry about this overhead?
60
mpn_mullow_n (mp_ptr rp, mp_srcptr xp, mp_srcptr yp, mp_size_t n)
62
if (BELOW_THRESHOLD (n, MULLOW_BASECASE_THRESHOLD))
64
/* Allocate workspace of fixed size on stack: fast! */
65
mp_limb_t ws[MUL_BASECASE_ALLOC];
66
mpn_mul_basecase (ws, xp, n, yp, n);
69
else if (BELOW_THRESHOLD (n, MULLOW_DC_THRESHOLD))
71
mpn_mullow_basecase (rp, xp, yp, n);
73
else if (BELOW_THRESHOLD (n, MULLOW_MUL_N_THRESHOLD))
75
/* Divide-and-conquer */
76
mp_size_t n2 = n >> 1; /* floor(n/2) */
77
mp_size_t n1 = n - n2; /* ceil(n/2) */
81
tp = TMP_SALLOC_LIMBS (n1);
83
/* Split as x = x1 2^(n1 GMP_NUMB_BITS) + x0,
84
y = y1 2^(n2 GMP_NUMB_BITS) + y0 */
87
mpn_mul_n (rp, xp, yp, n2);
89
rp[2 * n2] = mpn_addmul_1 (rp + n2, yp, n2, xp[n2]);
91
/* x1 * y0 * 2^(n1 GMP_NUMB_BITS) */
92
mpn_mullow_n (tp, xp + n1, yp, n2);
93
mpn_add_n (rp + n1, rp + n1, tp, n2);
95
/* x0 * y1 * 2^(n2 GMP_NUMB_BITS) */
96
mpn_mullow_n (tp, yp + n2, xp, n1);
97
mpn_add_n (rp + n2, rp + n2, tp, n1);
102
/* For really large operands, use plain mpn_mul_n but throw away upper n
107
tp = TMP_ALLOC_LIMBS (2 * n);
109
mpn_mul_n (tp, xp, yp, n);
110
MPN_COPY (rp, tp, n);
added
removed
src/gmp/mpn/generic/mullow_n.c
