Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * rational fractions
4 *
5 * Copyright (C) 2009 emlix GmbH, Oskar Schirmer <oskar@scara.com>
6 *
7 * helper functions when coping with rational numbers
8 */
9
10#include <linux/rational.h>
11#include <linux/compiler.h>
12#include <linux/export.h>
13
14/*
15 * calculate best rational approximation for a given fraction
16 * taking into account restricted register size, e.g. to find
17 * appropriate values for a pll with 5 bit denominator and
18 * 8 bit numerator register fields, trying to set up with a
19 * frequency ratio of 3.1415, one would say:
20 *
21 * rational_best_approximation(31415, 10000,
22 * (1 << 8) - 1, (1 << 5) - 1, &n, &d);
23 *
24 * you may look at given_numerator as a fixed point number,
25 * with the fractional part size described in given_denominator.
26 *
27 * for theoretical background, see:
28 * http://en.wikipedia.org/wiki/Continued_fraction
29 */
30
31void rational_best_approximation(
32 unsigned long given_numerator, unsigned long given_denominator,
33 unsigned long max_numerator, unsigned long max_denominator,
34 unsigned long *best_numerator, unsigned long *best_denominator)
35{
36 unsigned long n, d, n0, d0, n1, d1;
37 n = given_numerator;
38 d = given_denominator;
39 n0 = d1 = 0;
40 n1 = d0 = 1;
41 for (;;) {
42 unsigned long t, a;
43 if ((n1 > max_numerator) || (d1 > max_denominator)) {
44 n1 = n0;
45 d1 = d0;
46 break;
47 }
48 if (d == 0)
49 break;
50 t = d;
51 a = n / d;
52 d = n % d;
53 n = t;
54 t = n0 + a * n1;
55 n0 = n1;
56 n1 = t;
57 t = d0 + a * d1;
58 d0 = d1;
59 d1 = t;
60 }
61 *best_numerator = n1;
62 *best_denominator = d1;
63}
64
65EXPORT_SYMBOL(rational_best_approximation);
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * rational fractions
4 *
5 * Copyright (C) 2009 emlix GmbH, Oskar Schirmer <oskar@scara.com>
6 * Copyright (C) 2019 Trent Piepho <tpiepho@gmail.com>
7 *
8 * helper functions when coping with rational numbers
9 */
10
11#include <linux/rational.h>
12#include <linux/compiler.h>
13#include <linux/export.h>
14#include <linux/kernel.h>
15
16/*
17 * calculate best rational approximation for a given fraction
18 * taking into account restricted register size, e.g. to find
19 * appropriate values for a pll with 5 bit denominator and
20 * 8 bit numerator register fields, trying to set up with a
21 * frequency ratio of 3.1415, one would say:
22 *
23 * rational_best_approximation(31415, 10000,
24 * (1 << 8) - 1, (1 << 5) - 1, &n, &d);
25 *
26 * you may look at given_numerator as a fixed point number,
27 * with the fractional part size described in given_denominator.
28 *
29 * for theoretical background, see:
30 * https://en.wikipedia.org/wiki/Continued_fraction
31 */
32
33void rational_best_approximation(
34 unsigned long given_numerator, unsigned long given_denominator,
35 unsigned long max_numerator, unsigned long max_denominator,
36 unsigned long *best_numerator, unsigned long *best_denominator)
37{
38 /* n/d is the starting rational, which is continually
39 * decreased each iteration using the Euclidean algorithm.
40 *
41 * dp is the value of d from the prior iteration.
42 *
43 * n2/d2, n1/d1, and n0/d0 are our successively more accurate
44 * approximations of the rational. They are, respectively,
45 * the current, previous, and two prior iterations of it.
46 *
47 * a is current term of the continued fraction.
48 */
49 unsigned long n, d, n0, d0, n1, d1, n2, d2;
50 n = given_numerator;
51 d = given_denominator;
52 n0 = d1 = 0;
53 n1 = d0 = 1;
54
55 for (;;) {
56 unsigned long dp, a;
57
58 if (d == 0)
59 break;
60 /* Find next term in continued fraction, 'a', via
61 * Euclidean algorithm.
62 */
63 dp = d;
64 a = n / d;
65 d = n % d;
66 n = dp;
67
68 /* Calculate the current rational approximation (aka
69 * convergent), n2/d2, using the term just found and
70 * the two prior approximations.
71 */
72 n2 = n0 + a * n1;
73 d2 = d0 + a * d1;
74
75 /* If the current convergent exceeds the maxes, then
76 * return either the previous convergent or the
77 * largest semi-convergent, the final term of which is
78 * found below as 't'.
79 */
80 if ((n2 > max_numerator) || (d2 > max_denominator)) {
81 unsigned long t = min((max_numerator - n0) / n1,
82 (max_denominator - d0) / d1);
83
84 /* This tests if the semi-convergent is closer
85 * than the previous convergent.
86 */
87 if (2u * t > a || (2u * t == a && d0 * dp > d1 * d)) {
88 n1 = n0 + t * n1;
89 d1 = d0 + t * d1;
90 }
91 break;
92 }
93 n0 = n1;
94 n1 = n2;
95 d0 = d1;
96 d1 = d2;
97 }
98 *best_numerator = n1;
99 *best_denominator = d1;
100}
101
102EXPORT_SYMBOL(rational_best_approximation);