|
Packit |
5c3484 |
/* mpn_divexact_1 -- mpn by limb exact division.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
THE FUNCTIONS IN THIS FILE ARE FOR INTERNAL USE ONLY. THEY'RE ALMOST
|
|
Packit |
5c3484 |
CERTAIN TO BE SUBJECT TO INCOMPATIBLE CHANGES OR DISAPPEAR COMPLETELY IN
|
|
Packit |
5c3484 |
FUTURE GNU MP RELEASES.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
Copyright 2000-2003, 2005, 2013 Free Software Foundation, Inc.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
This file is part of the GNU MP Library.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
The GNU MP Library is free software; you can redistribute it and/or modify
|
|
Packit |
5c3484 |
it under the terms of either:
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
* the GNU Lesser General Public License as published by the Free
|
|
Packit |
5c3484 |
Software Foundation; either version 3 of the License, or (at your
|
|
Packit |
5c3484 |
option) any later version.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
or
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
* the GNU General Public License as published by the Free Software
|
|
Packit |
5c3484 |
Foundation; either version 2 of the License, or (at your option) any
|
|
Packit |
5c3484 |
later version.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
or both in parallel, as here.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
The GNU MP Library is distributed in the hope that it will be useful, but
|
|
Packit |
5c3484 |
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
|
|
Packit |
5c3484 |
or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
|
Packit |
5c3484 |
for more details.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
You should have received copies of the GNU General Public License and the
|
|
Packit |
5c3484 |
GNU Lesser General Public License along with the GNU MP Library. If not,
|
|
Packit |
5c3484 |
see https://www.gnu.org/licenses/. */
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
#include "gmp.h"
|
|
Packit |
5c3484 |
#include "gmp-impl.h"
|
|
Packit |
5c3484 |
#include "longlong.h"
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
/* Divide a={src,size} by d=divisor and store the quotient in q={dst,size}.
|
|
Packit |
5c3484 |
q will only be correct if d divides a exactly.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
A separate loop is used for shift==0 because n<
|
|
Packit |
5c3484 |
give zero on all CPUs (for instance it doesn't on the x86s). This
|
|
Packit |
5c3484 |
separate loop might run faster too, helping odd divisors.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
Possibilities:
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
mpn_divexact_1c could be created, accepting and returning c. This would
|
|
Packit |
5c3484 |
let a long calculation be done piece by piece. Currently there's no
|
|
Packit |
5c3484 |
particular need for that, and not returning c means that a final umul can
|
|
Packit |
5c3484 |
be skipped.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
Another use for returning c would be letting the caller know whether the
|
|
Packit |
5c3484 |
division was in fact exact. It would work just to return the carry bit
|
|
Packit |
5c3484 |
"c=(l>s)" and let the caller do a final umul if interested.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
When the divisor is even, the factors of two could be handled with a
|
|
Packit |
5c3484 |
separate mpn_rshift, instead of shifting on the fly. That might be
|
|
Packit |
5c3484 |
faster on some CPUs and would mean just the shift==0 style loop would be
|
|
Packit |
5c3484 |
needed.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
If n<
|
|
Packit |
5c3484 |
shift==0 loop is unnecessary, and could be eliminated if there's no great
|
|
Packit |
5c3484 |
speed difference.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
It's not clear whether "/" is the best way to handle size==1. Alpha gcc
|
|
Packit |
5c3484 |
2.95 for instance has a poor "/" and might prefer the modular method.
|
|
Packit |
5c3484 |
Perhaps a tuned parameter should control this.
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
If src[size-1] < divisor then dst[size-1] will be zero, and one divide
|
|
Packit |
5c3484 |
step could be skipped. A test at last step for s
|
|
Packit |
5c3484 |
even case) might be a good way to do that. But if this code is often
|
|
Packit |
5c3484 |
used with small divisors then it might not be worth bothering */
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
void
|
|
Packit |
5c3484 |
mpn_divexact_1 (mp_ptr dst, mp_srcptr src, mp_size_t size, mp_limb_t divisor)
|
|
Packit |
5c3484 |
{
|
|
Packit |
5c3484 |
mp_size_t i;
|
|
Packit |
5c3484 |
mp_limb_t c, h, l, ls, s, s_next, inverse, dummy;
|
|
Packit |
5c3484 |
unsigned shift;
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
ASSERT (size >= 1);
|
|
Packit |
5c3484 |
ASSERT (divisor != 0);
|
|
Packit |
5c3484 |
ASSERT (MPN_SAME_OR_SEPARATE_P (dst, src, size));
|
|
Packit |
5c3484 |
ASSERT_MPN (src, size);
|
|
Packit |
5c3484 |
ASSERT_LIMB (divisor);
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
if ((divisor & 1) == 0)
|
|
Packit |
5c3484 |
{
|
|
Packit |
5c3484 |
count_trailing_zeros (shift, divisor);
|
|
Packit |
5c3484 |
divisor >>= shift;
|
|
Packit |
5c3484 |
}
|
|
Packit |
5c3484 |
else
|
|
Packit |
5c3484 |
shift = 0;
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
binvert_limb (inverse, divisor);
|
|
Packit |
5c3484 |
divisor <<= GMP_NAIL_BITS;
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
if (shift != 0)
|
|
Packit |
5c3484 |
{
|
|
Packit |
5c3484 |
c = 0;
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
s = src[0];
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
for (i = 1; i < size; i++)
|
|
Packit |
5c3484 |
{
|
|
Packit |
5c3484 |
s_next = src[i];
|
|
Packit |
5c3484 |
ls = ((s >> shift) | (s_next << (GMP_NUMB_BITS-shift))) & GMP_NUMB_MASK;
|
|
Packit |
5c3484 |
s = s_next;
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
SUBC_LIMB (c, l, ls, c);
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
l = (l * inverse) & GMP_NUMB_MASK;
|
|
Packit |
5c3484 |
dst[i - 1] = l;
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
umul_ppmm (h, dummy, l, divisor);
|
|
Packit |
5c3484 |
c += h;
|
|
Packit |
5c3484 |
}
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
ls = s >> shift;
|
|
Packit |
5c3484 |
l = ls - c;
|
|
Packit |
5c3484 |
l = (l * inverse) & GMP_NUMB_MASK;
|
|
Packit |
5c3484 |
dst[size - 1] = l;
|
|
Packit |
5c3484 |
}
|
|
Packit |
5c3484 |
else
|
|
Packit |
5c3484 |
{
|
|
Packit |
5c3484 |
s = src[0];
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
l = (s * inverse) & GMP_NUMB_MASK;
|
|
Packit |
5c3484 |
dst[0] = l;
|
|
Packit |
5c3484 |
c = 0;
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
for (i = 1; i < size; i++)
|
|
Packit |
5c3484 |
{
|
|
Packit |
5c3484 |
umul_ppmm (h, dummy, l, divisor);
|
|
Packit |
5c3484 |
c += h;
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
s = src[i];
|
|
Packit |
5c3484 |
SUBC_LIMB (c, l, s, c);
|
|
Packit |
5c3484 |
|
|
Packit |
5c3484 |
l = (l * inverse) & GMP_NUMB_MASK;
|
|
Packit |
5c3484 |
dst[i] = l;
|
|
Packit |
5c3484 |
}
|
|
Packit |
5c3484 |
}
|
|
Packit |
5c3484 |
}
|