xref: /haiku/src/system/libroot/posix/glibc/arch/generic/mul.c (revision 21258e2674226d6aa732321b6f8494841895af5f)
1 /* mpn_mul -- Multiply two natural numbers.
2 
3 Copyright (C) 1991, 1993, 1994, 1996 Free Software Foundation, Inc.
4 
5 This file is part of the GNU MP Library.
6 
7 The GNU MP Library is free software; you can redistribute it and/or modify
8 it under the terms of the GNU Lesser General Public License as published by
9 the Free Software Foundation; either version 2.1 of the License, or (at your
10 option) any later version.
11 
12 The GNU MP Library is distributed in the hope that it will be useful, but
13 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
14 or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
15 License for more details.
16 
17 You should have received a copy of the GNU Lesser General Public License
18 along with the GNU MP Library; see the file COPYING.LIB.  If not, write to
19 the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
20 MA 02111-1307, USA. */
21 
22 #include "gmp.h"
23 #include "gmp-impl.h"
24 
25 /* Multiply the natural numbers u (pointed to by UP, with USIZE limbs)
26    and v (pointed to by VP, with VSIZE limbs), and store the result at
27    PRODP.  USIZE + VSIZE limbs are always stored, but if the input
28    operands are normalized.  Return the most significant limb of the
29    result.
30 
31    NOTE: The space pointed to by PRODP is overwritten before finished
32    with U and V, so overlap is an error.
33 
34    Argument constraints:
35    1. USIZE >= VSIZE.
36    2. PRODP != UP and PRODP != VP, i.e. the destination
37       must be distinct from the multiplier and the multiplicand.  */
38 
39 /* If KARATSUBA_THRESHOLD is not already defined, define it to a
40    value which is good on most machines.  */
41 #ifndef KARATSUBA_THRESHOLD
42 #define KARATSUBA_THRESHOLD 32
43 #endif
44 
45 mp_limb_t
46 #if __STDC__
47 mpn_mul (mp_ptr prodp,
48 	 mp_srcptr up, mp_size_t usize,
49 	 mp_srcptr vp, mp_size_t vsize)
50 #else
51 mpn_mul (prodp, up, usize, vp, vsize)
52      mp_ptr prodp;
53      mp_srcptr up;
54      mp_size_t usize;
55      mp_srcptr vp;
56      mp_size_t vsize;
57 #endif
58 {
59   mp_ptr prod_endp = prodp + usize + vsize - 1;
60   mp_limb_t cy;
61   mp_ptr tspace;
62   TMP_DECL (marker);
63 
64   if (vsize < KARATSUBA_THRESHOLD)
65     {
66       /* Handle simple cases with traditional multiplication.
67 
68 	 This is the most critical code of the entire function.  All
69 	 multiplies rely on this, both small and huge.  Small ones arrive
70 	 here immediately.  Huge ones arrive here as this is the base case
71 	 for Karatsuba's recursive algorithm below.  */
72       mp_size_t i;
73       mp_limb_t cy_limb;
74       mp_limb_t v_limb;
75 
76       if (vsize == 0)
77 	return 0;
78 
79       /* Multiply by the first limb in V separately, as the result can be
80 	 stored (not added) to PROD.  We also avoid a loop for zeroing.  */
81       v_limb = vp[0];
82       if (v_limb <= 1)
83 	{
84 	  if (v_limb == 1)
85 	    MPN_COPY (prodp, up, usize);
86 	  else
87 	    MPN_ZERO (prodp, usize);
88 	  cy_limb = 0;
89 	}
90       else
91 	cy_limb = mpn_mul_1 (prodp, up, usize, v_limb);
92 
93       prodp[usize] = cy_limb;
94       prodp++;
95 
96       /* For each iteration in the outer loop, multiply one limb from
97 	 U with one limb from V, and add it to PROD.  */
98       for (i = 1; i < vsize; i++)
99 	{
100 	  v_limb = vp[i];
101 	  if (v_limb <= 1)
102 	    {
103 	      cy_limb = 0;
104 	      if (v_limb == 1)
105 		cy_limb = mpn_add_n (prodp, prodp, up, usize);
106 	    }
107 	  else
108 	    cy_limb = mpn_addmul_1 (prodp, up, usize, v_limb);
109 
110 	  prodp[usize] = cy_limb;
111 	  prodp++;
112 	}
113       return cy_limb;
114     }
115 
116   TMP_MARK (marker);
117 
118   tspace = (mp_ptr) TMP_ALLOC (2 * vsize * BYTES_PER_MP_LIMB);
119   MPN_MUL_N_RECURSE (prodp, up, vp, vsize, tspace);
120 
121   prodp += vsize;
122   up += vsize;
123   usize -= vsize;
124   if (usize >= vsize)
125     {
126       mp_ptr tp = (mp_ptr) TMP_ALLOC (2 * vsize * BYTES_PER_MP_LIMB);
127       do
128 	{
129 	  MPN_MUL_N_RECURSE (tp, up, vp, vsize, tspace);
130 	  cy = mpn_add_n (prodp, prodp, tp, vsize);
131 	  mpn_add_1 (prodp + vsize, tp + vsize, vsize, cy);
132 	  prodp += vsize;
133 	  up += vsize;
134 	  usize -= vsize;
135 	}
136       while (usize >= vsize);
137     }
138 
139   /* True: usize < vsize.  */
140 
141   /* Make life simple: Recurse.  */
142 
143   if (usize != 0)
144     {
145       mpn_mul (tspace, vp, vsize, up, usize);
146       cy = mpn_add_n (prodp, prodp, tspace, vsize);
147       mpn_add_1 (prodp + vsize, tspace + vsize, usize, cy);
148     }
149 
150   TMP_FREE (marker);
151   return *prod_endp;
152 }
153