xref: /netbsd-src/external/lgpl3/gmp/dist/mpn/generic/dive_1.c (revision 72c7faa4dbb41dbb0238d6b4a109da0d4b236dd4)
1 /* mpn_divexact_1 -- mpn by limb exact division.
2 
3    THE FUNCTIONS IN THIS FILE ARE FOR INTERNAL USE ONLY.  THEY'RE ALMOST
4    CERTAIN TO BE SUBJECT TO INCOMPATIBLE CHANGES OR DISAPPEAR COMPLETELY IN
5    FUTURE GNU MP RELEASES.
6 
7 Copyright 2000-2003, 2005, 2013 Free Software Foundation, Inc.
8 
9 This file is part of the GNU MP Library.
10 
11 The GNU MP Library is free software; you can redistribute it and/or modify
12 it under the terms of either:
13 
14   * the GNU Lesser General Public License as published by the Free
15     Software Foundation; either version 3 of the License, or (at your
16     option) any later version.
17 
18 or
19 
20   * the GNU General Public License as published by the Free Software
21     Foundation; either version 2 of the License, or (at your option) any
22     later version.
23 
24 or both in parallel, as here.
25 
26 The GNU MP Library is distributed in the hope that it will be useful, but
27 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
28 or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
29 for more details.
30 
31 You should have received copies of the GNU General Public License and the
32 GNU Lesser General Public License along with the GNU MP Library.  If not,
33 see https://www.gnu.org/licenses/.  */
34 
35 #include "gmp-impl.h"
36 #include "longlong.h"
37 
38 
39 
40 /* Divide a={src,size} by d=divisor and store the quotient in q={dst,size}.
41    q will only be correct if d divides a exactly.
42 
43    A separate loop is used for shift==0 because n<<GMP_LIMB_BITS doesn't
44    give zero on all CPUs (for instance it doesn't on the x86s).  This
45    separate loop might run faster too, helping odd divisors.
46 
47    Possibilities:
48 
49    mpn_divexact_1c could be created, accepting and returning c.  This would
50    let a long calculation be done piece by piece.  Currently there's no
51    particular need for that, and not returning c means that a final umul can
52    be skipped.
53 
54    Another use for returning c would be letting the caller know whether the
55    division was in fact exact.  It would work just to return the carry bit
56    "c=(l>s)" and let the caller do a final umul if interested.
57 
58    When the divisor is even, the factors of two could be handled with a
59    separate mpn_rshift, instead of shifting on the fly.  That might be
60    faster on some CPUs and would mean just the shift==0 style loop would be
61    needed.
62 
63    If n<<GMP_LIMB_BITS gives zero on a particular CPU then the separate
64    shift==0 loop is unnecessary, and could be eliminated if there's no great
65    speed difference.
66 
67    It's not clear whether "/" is the best way to handle size==1.  Alpha gcc
68    2.95 for instance has a poor "/" and might prefer the modular method.
69    Perhaps a tuned parameter should control this.
70 
71    If src[size-1] < divisor then dst[size-1] will be zero, and one divide
72    step could be skipped.  A test at last step for s<divisor (or ls in the
73    even case) might be a good way to do that.  But if this code is often
74    used with small divisors then it might not be worth bothering  */
75 
76 void
mpn_divexact_1(mp_ptr dst,mp_srcptr src,mp_size_t size,mp_limb_t divisor)77 mpn_divexact_1 (mp_ptr dst, mp_srcptr src, mp_size_t size, mp_limb_t divisor)
78 {
79   mp_size_t  i;
80   mp_limb_t  c, h, l, ls, s, s_next, inverse, dummy;
81   unsigned   shift;
82 
83   ASSERT (size >= 1);
84   ASSERT (divisor != 0);
85   ASSERT (MPN_SAME_OR_SEPARATE_P (dst, src, size));
86   ASSERT_MPN (src, size);
87   ASSERT_LIMB (divisor);
88 
89   if ((divisor & 1) == 0)
90     {
91       count_trailing_zeros (shift, divisor);
92       divisor >>= shift;
93     }
94   else
95     shift = 0;
96 
97   binvert_limb (inverse, divisor);
98   divisor <<= GMP_NAIL_BITS;
99 
100   if (shift != 0)
101     {
102       c = 0;
103 
104       s = src[0];
105 
106       for (i = 1; i < size; i++)
107 	{
108 	  s_next = src[i];
109 	  ls = ((s >> shift) | (s_next << (GMP_NUMB_BITS-shift))) & GMP_NUMB_MASK;
110 	  s = s_next;
111 
112 	  SUBC_LIMB (c, l, ls, c);
113 
114 	  l = (l * inverse) & GMP_NUMB_MASK;
115 	  dst[i - 1] = l;
116 
117 	  umul_ppmm (h, dummy, l, divisor);
118 	  c += h;
119 	}
120 
121       ls = s >> shift;
122       l = ls - c;
123       l = (l * inverse) & GMP_NUMB_MASK;
124       dst[size - 1] = l;
125     }
126   else
127     {
128       s = src[0];
129 
130       l = (s * inverse) & GMP_NUMB_MASK;
131       dst[0] = l;
132       c = 0;
133 
134       for (i = 1; i < size; i++)
135 	{
136 	  umul_ppmm (h, dummy, l, divisor);
137 	  c += h;
138 
139 	  s = src[i];
140 	  SUBC_LIMB (c, l, s, c);
141 
142 	  l = (l * inverse) & GMP_NUMB_MASK;
143 	  dst[i] = l;
144 	}
145     }
146 }
147