comparison tomsfastmath/src/divide/fp_div.c @ 643:a362b62d38b2 dropbear-tfm

Add tomsfastmath from git rev bfa4582842bc3bab42e4be4aed5703437049502a with Makefile.in renamed
author Matt Johnston <matt@ucc.asn.au>
date Wed, 23 Nov 2011 18:10:20 +0700
parents
children
comparison
equal deleted inserted replaced
642:33fd2f3499d2 643:a362b62d38b2
1 /* TomsFastMath, a fast ISO C bignum library.
2 *
3 * This project is meant to fill in where LibTomMath
4 * falls short. That is speed ;-)
5 *
6 * This project is public domain and free for all purposes.
7 *
8 * Tom St Denis, [email protected]
9 */
10 #include <tfm.h>
11
12 /* a/b => cb + d == a */
13 int fp_div(fp_int *a, fp_int *b, fp_int *c, fp_int *d)
14 {
15 fp_int q, x, y, t1, t2;
16 int n, t, i, norm, neg;
17
18 /* is divisor zero ? */
19 if (fp_iszero (b) == 1) {
20 return FP_VAL;
21 }
22
23 /* if a < b then q=0, r = a */
24 if (fp_cmp_mag (a, b) == FP_LT) {
25 if (d != NULL) {
26 fp_copy (a, d);
27 }
28 if (c != NULL) {
29 fp_zero (c);
30 }
31 return FP_OKAY;
32 }
33
34 fp_init(&q);
35 q.used = a->used + 2;
36
37 fp_init(&t1);
38 fp_init(&t2);
39 fp_init_copy(&x, a);
40 fp_init_copy(&y, b);
41
42 /* fix the sign */
43 neg = (a->sign == b->sign) ? FP_ZPOS : FP_NEG;
44 x.sign = y.sign = FP_ZPOS;
45
46 /* normalize both x and y, ensure that y >= b/2, [b == 2**DIGIT_BIT] */
47 norm = fp_count_bits(&y) % DIGIT_BIT;
48 if (norm < (int)(DIGIT_BIT-1)) {
49 norm = (DIGIT_BIT-1) - norm;
50 fp_mul_2d (&x, norm, &x);
51 fp_mul_2d (&y, norm, &y);
52 } else {
53 norm = 0;
54 }
55
56 /* note hac does 0 based, so if used==5 then its 0,1,2,3,4, e.g. use 4 */
57 n = x.used - 1;
58 t = y.used - 1;
59
60 /* while (x >= y*b**n-t) do { q[n-t] += 1; x -= y*b**{n-t} } */
61 fp_lshd (&y, n - t); /* y = y*b**{n-t} */
62
63 while (fp_cmp (&x, &y) != FP_LT) {
64 ++(q.dp[n - t]);
65 fp_sub (&x, &y, &x);
66 }
67
68 /* reset y by shifting it back down */
69 fp_rshd (&y, n - t);
70
71 /* step 3. for i from n down to (t + 1) */
72 for (i = n; i >= (t + 1); i--) {
73 if (i > x.used) {
74 continue;
75 }
76
77 /* step 3.1 if xi == yt then set q{i-t-1} to b-1,
78 * otherwise set q{i-t-1} to (xi*b + x{i-1})/yt */
79 if (x.dp[i] == y.dp[t]) {
80 q.dp[i - t - 1] = ((((fp_word)1) << DIGIT_BIT) - 1);
81 } else {
82 fp_word tmp;
83 tmp = ((fp_word) x.dp[i]) << ((fp_word) DIGIT_BIT);
84 tmp |= ((fp_word) x.dp[i - 1]);
85 tmp /= ((fp_word) y.dp[t]);
86 q.dp[i - t - 1] = (fp_digit) (tmp);
87 }
88
89 /* while (q{i-t-1} * (yt * b + y{t-1})) >
90 xi * b**2 + xi-1 * b + xi-2
91
92 do q{i-t-1} -= 1;
93 */
94 q.dp[i - t - 1] = (q.dp[i - t - 1] + 1);
95 do {
96 q.dp[i - t - 1] = (q.dp[i - t - 1] - 1);
97
98 /* find left hand */
99 fp_zero (&t1);
100 t1.dp[0] = (t - 1 < 0) ? 0 : y.dp[t - 1];
101 t1.dp[1] = y.dp[t];
102 t1.used = 2;
103 fp_mul_d (&t1, q.dp[i - t - 1], &t1);
104
105 /* find right hand */
106 t2.dp[0] = (i - 2 < 0) ? 0 : x.dp[i - 2];
107 t2.dp[1] = (i - 1 < 0) ? 0 : x.dp[i - 1];
108 t2.dp[2] = x.dp[i];
109 t2.used = 3;
110 } while (fp_cmp_mag(&t1, &t2) == FP_GT);
111
112 /* step 3.3 x = x - q{i-t-1} * y * b**{i-t-1} */
113 fp_mul_d (&y, q.dp[i - t - 1], &t1);
114 fp_lshd (&t1, i - t - 1);
115 fp_sub (&x, &t1, &x);
116
117 /* if x < 0 then { x = x + y*b**{i-t-1}; q{i-t-1} -= 1; } */
118 if (x.sign == FP_NEG) {
119 fp_copy (&y, &t1);
120 fp_lshd (&t1, i - t - 1);
121 fp_add (&x, &t1, &x);
122 q.dp[i - t - 1] = q.dp[i - t - 1] - 1;
123 }
124 }
125
126 /* now q is the quotient and x is the remainder
127 * [which we have to normalize]
128 */
129
130 /* get sign before writing to c */
131 x.sign = x.used == 0 ? FP_ZPOS : a->sign;
132
133 if (c != NULL) {
134 fp_clamp (&q);
135 fp_copy (&q, c);
136 c->sign = neg;
137 }
138
139 if (d != NULL) {
140 fp_div_2d (&x, norm, &x, NULL);
141
142 /* the following is a kludge, essentially we were seeing the right remainder but
143 with excess digits that should have been zero
144 */
145 for (i = b->used; i < x.used; i++) {
146 x.dp[i] = 0;
147 }
148 fp_clamp(&x);
149 fp_copy (&x, d);
150 }
151
152 return FP_OKAY;
153 }
154
155 /* $Source$ */
156 /* $Revision$ */
157 /* $Date$ */