1/* mpn_redc_2.  Set rp[] <- up[]/R^n mod mp[].  Clobber up[].
2   mp[] is n limbs; up[] is 2n limbs.
3
4   THIS IS AN INTERNAL FUNCTION WITH A MUTABLE INTERFACE.  IT IS ONLY
5   SAFE TO REACH THIS FUNCTION THROUGH DOCUMENTED INTERFACES.
6
7Copyright (C) 2000-2002, 2004, 2008, 2012 Free Software Foundation, Inc.
8
9This file is part of the GNU MP Library.
10
11The GNU MP Library is free software; you can redistribute it and/or modify
12it 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
18or
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
24or both in parallel, as here.
25
26The GNU MP Library is distributed in the hope that it will be useful, but
27WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
28or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
29for more details.
30
31You should have received copies of the GNU General Public License and the
32GNU Lesser General Public License along with the GNU MP Library.  If not,
33see https://www.gnu.org/licenses/.  */
34
35#include "gmp-impl.h"
36#include "longlong.h"
37
38
39#if GMP_NAIL_BITS != 0
40you lose
41#endif
42
43/* For testing purposes, define our own mpn_addmul_2 if there is none already
44   available.  */
45#ifndef HAVE_NATIVE_mpn_addmul_2
46#undef mpn_addmul_2
47static mp_limb_t
48mpn_addmul_2 (mp_ptr rp, mp_srcptr up, mp_size_t n, mp_srcptr vp)
49{
50  rp[n] = mpn_addmul_1 (rp, up, n, vp[0]);
51  return mpn_addmul_1 (rp + 1, up, n, vp[1]);
52}
53#endif
54
55#if defined (__GNUC__) && ! defined (NO_ASM) \
56  && defined (__ia64) && W_TYPE_SIZE == 64
57#define umul2low(ph, pl, uh, ul, vh, vl) \
58  do {									\
59    mp_limb_t _ph, _pl;							\
60    __asm__ ("xma.hu %0 = %3, %5, f0\n\t"				\
61	     "xma.l %1 = %3, %5, f0\n\t"				\
62	     ";;\n\t"							\
63	     "xma.l %0 = %3, %4, %0\n\t"				\
64	     ";;\n\t"							\
65	     "xma.l %0 = %2, %5, %0"					\
66	     : "=&f" (ph), "=&f" (pl)					\
67	     : "f" (uh), "f" (ul), "f" (vh), "f" (vl));			\
68  } while (0)
69#endif
70
71#ifndef umul2low
72#define umul2low(ph, pl, uh, ul, vh, vl) \
73  do {									\
74    mp_limb_t _ph, _pl;							\
75    umul_ppmm (_ph, _pl, ul, vl);					\
76    (ph) = _ph + (ul) * (vh) + (uh) * (vl);				\
77    (pl) = _pl;								\
78  } while (0)
79#endif
80
81mp_limb_t
82mpn_redc_2 (mp_ptr rp, mp_ptr up, mp_srcptr mp, mp_size_t n, mp_srcptr mip)
83{
84  mp_limb_t q[2];
85  mp_size_t j;
86  mp_limb_t upn;
87  mp_limb_t cy;
88
89  ASSERT (n > 0);
90  ASSERT_MPN (up, 2*n);
91
92  if ((n & 1) != 0)
93    {
94      up[0] = mpn_addmul_1 (up, mp, n, (up[0] * mip[0]) & GMP_NUMB_MASK);
95      up++;
96    }
97
98  for (j = n - 2; j >= 0; j -= 2)
99    {
100      umul2low (q[1], q[0], mip[1], mip[0], up[1], up[0]);
101      upn = up[n];		/* mpn_addmul_2 overwrites this */
102      up[1] = mpn_addmul_2 (up, mp, n, q);
103      up[0] = up[n];
104      up[n] = upn;
105      up += 2;
106    }
107
108  cy = mpn_add_n (rp, up, up - n, n);
109  return cy;
110}
111