3 * $Id: mp-arith.c,v 1.3 1999/12/11 10:57:43 mdw Exp $
5 * Basic arithmetic on multiprecision integers
7 * (c) 1999 Straylight/Edgeware
10 /*----- Licensing notice --------------------------------------------------*
12 * This file is part of Catacomb.
14 * Catacomb is free software; you can redistribute it and/or modify
15 * it under the terms of the GNU Library General Public License as
16 * published by the Free Software Foundation; either version 2 of the
17 * License, or (at your option) any later version.
19 * Catacomb is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22 * GNU Library General Public License for more details.
24 * You should have received a copy of the GNU Library General Public
25 * License along with Catacomb; if not, write to the Free
26 * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
30 /*----- Revision history --------------------------------------------------*
32 * $Log: mp-arith.c,v $
33 * Revision 1.3 1999/12/11 10:57:43 mdw
34 * Karatsuba squaring algorithm.
36 * Revision 1.2 1999/12/10 23:18:39 mdw
37 * Change interface for suggested destinations.
39 * Revision 1.1 1999/11/17 18:02:16 mdw
40 * New multiprecision integer arithmetic suite.
44 /*----- Header files ------------------------------------------------------*/
48 /*----- Macros ------------------------------------------------------------*/
50 #define MAX(x, y) ((x) >= (y) ? (x) : (y))
52 /*----- Main code ---------------------------------------------------------*/
56 * Arguments: @mp *a@ = source
58 * Returns: Result, @a@ converted to two's complement notation.
61 mp
*mp_2c(mp
*d
, mp
*a
)
66 MP_MODIFY(d
, MP_LEN(a
));
67 mpx_2c(d
->v
, d
->vl
, a
->v
, a
->vl
);
68 d
->f
= a
->f
& MP_BURN
;
75 * Arguments: @mp *d@ = destination
78 * Returns: Result, @a@ converted to the native signed-magnitude
82 mp
*mp_sm(mp
*d
, mp
*a
)
84 if (!MP_LEN(a
) || a
->vl
[-1] < MPW_MAX
/ 2)
87 MP_MODIFY(d
, MP_LEN(a
));
88 mpx_2c(d
->v
, d
->vl
, a
->v
, a
->vl
);
89 d
->f
= (a
->f
& (MP_BURN
| MP_NEG
)) ^ MP_NEG
;
96 * Arguments: @mp *d@ = destination
98 * @size_t n@ = number of bits to move
100 * Returns: Result, @a@ shifted left by @n@.
103 mp
*mp_lsl(mp
*d
, mp
*a
, size_t n
)
105 MP_MODIFY(d
, MP_LEN(a
) + (n
+ MPW_BITS
- 1) / MPW_BITS
);
106 mpx_lsl(d
->v
, d
->vl
, a
->v
, a
->vl
, n
);
107 d
->f
= a
->f
& (MP_NEG
| MP_BURN
);
112 /* --- @mp_lsr@ --- *
114 * Arguments: @mp *d@ = destination
116 * @size_t n@ = number of bits to move
118 * Returns: Result, @a@ shifted left by @n@.
121 mp
*mp_lsr(mp
*d
, mp
*a
, size_t n
)
123 MP_MODIFY(d
, MP_LEN(a
));
124 mpx_lsr(d
->v
, d
->vl
, a
->v
, a
->vl
, n
);
125 d
->f
= a
->f
& (MP_NEG
| MP_BURN
);
130 /* --- @mp_cmp@ --- *
132 * Arguments: @const mp *a, *b@ = two numbers
134 * Returns: Less than, equal to or greater than zero, according to
135 * whether @a@ is less than, equal to or greater than @b@.
138 int mp_cmp(const mp
*a
, const mp
*b
)
140 if (!((a
->f
^ b
->f
) & MP_NEG
))
141 return (mpx_ucmp(a
->v
, a
->vl
, b
->v
, b
->vl
));
142 else if (a
->f
& MP_NEG
)
148 /* --- @mp_add@ --- *
150 * Arguments: @mp *d@ = destination
151 * @mp *a, *b@ = sources
153 * Returns: Result, @a@ added to @b@.
156 mp
*mp_add(mp
*d
, mp
*a
, mp
*b
)
158 MP_MODIFY(d
, MAX(MP_LEN(a
), MP_LEN(b
)) + 1);
159 if (!((a
->f
^ b
->f
) & MP_NEG
))
160 mpx_uadd(d
->v
, d
->vl
, a
->v
, a
->vl
, b
->v
, b
->vl
);
162 if (MPX_UCMP(a
->v
, a
->vl
, <, b
->v
, b
->vl
)) {
163 mp
*t
= a
; a
= b
; b
= t
;
165 mpx_usub(d
->v
, d
->vl
, a
->v
, a
->vl
, b
->v
, b
->vl
);
167 d
->f
= ((a
->f
| b
->f
) & MP_BURN
) | (a
->f
& MP_NEG
);
172 /* --- @mp_sub@ --- *
174 * Arguments: @mp *d@ = destination
175 * @mp *a, *b@ = sources
177 * Returns: Result, @b@ subtracted from @a@.
180 mp
*mp_sub(mp
*d
, mp
*a
, mp
*b
)
183 MP_MODIFY(d
, MAX(MP_LEN(a
), MP_LEN(b
)) + 1);
184 if ((a
->f
^ b
->f
) & MP_NEG
)
185 mpx_uadd(d
->v
, d
->vl
, a
->v
, a
->vl
, b
->v
, b
->vl
);
187 if (MPX_UCMP(a
->v
, a
->vl
, <, b
->v
, b
->vl
)) {
188 mp
*t
= a
; a
= b
; b
= t
;
191 mpx_usub(d
->v
, d
->vl
, a
->v
, a
->vl
, b
->v
, b
->vl
);
193 d
->f
= ((a
->f
| b
->f
) & MP_BURN
) | ((a
->f
^ sgn
) & MP_NEG
);
198 /* --- @mp_mul@ --- *
200 * Arguments: @mp *d@ = destination
201 * @mp *a, *b@ = sources
203 * Returns: Result, @a@ multiplied by @b@.
206 mp
*mp_mul(mp
*d
, mp
*a
, mp
*b
)
211 MP_MODIFY(d
, MP_LEN(a
) + MP_LEN(b
));
212 if (MP_LEN(a
) <= KARATSUBA_CUTOFF
|| MP_LEN(b
) <= KARATSUBA_CUTOFF
)
213 mpx_umul(d
->v
, d
->vl
, a
->v
, a
->vl
, b
->v
, b
->vl
);
215 size_t m
= MAX(MP_LEN(a
), MP_LEN(b
)) * 2 + KARATSUBA_SLOP
;
219 mpx_kmul(d
->v
, d
->vl
, a
->v
, a
->vl
, b
->v
, b
->vl
, s
, s
+ m
);
223 d
->f
= ((a
->f
| b
->f
) & MP_BURN
) | ((a
->f
^ b
->f
) & MP_NEG
);
230 /* --- @mp_sqr@ --- *
232 * Arguments: @mp *d@ = destination
235 * Returns: Result, @a@ squared.
238 mp
*mp_sqr(mp
*d
, mp
*a
)
240 size_t m
= MP_LEN(a
);
244 if (m
> KARATSUBA_CUTOFF
) {
246 m
= 2 * (m
+ 1) + 32;
248 mpx_ksqr(d
->v
, d
->vl
, a
->v
, a
->vl
, s
, s
+ m
);
251 mpx_usqr(d
->v
, d
->vl
, a
->v
, a
->vl
);
252 d
->f
= a
->f
& MP_BURN
;
258 /* --- @mp_div@ --- *
260 * Arguments: @mp **qq, **rr@ = destination, quotient and remainder
261 * @mp *a, *b@ = sources
263 * Use: Calculates the quotient and remainder when @a@ is divided by
264 * @b@. The destinations @*qq@ and @*rr@ must be distinct.
265 * Either of @qq@ or @rr@ may be null to indicate that the
266 * result is irrelevant. (Discarding both results is silly.)
267 * There is a performance advantage if @a == *rr@.
269 * The behaviour when @a@ and @b@ have the same sign is
270 * straightforward. When the signs differ, this implementation
271 * chooses @r@ to have the same sign as @b@, rather than the
272 * more normal choice that the remainder has the same sign as
273 * the dividend. This makes modular arithmetic a little more
277 void mp_div(mp
**qq
, mp
**rr
, mp
*a
, mp
*b
)
279 mp
*r
= rr ?
*rr
: MP_NEW
;
280 mp
*q
= qq ?
*qq
: MP_NEW
;
283 /* --- Set up some temporary workspace --- */
286 size_t rq
= MP_LEN(b
) + 1;
291 /* --- Set the remainder up right --- *
293 * Just in case the divisor is larger, be able to cope with this. It's not
294 * important in @mpx_udiv@, but it is here because of the sign correction.
298 size_t rq
= MP_LEN(a
) + 2;
306 MP_ENSURE(r
, MP_LEN(r
) + 2);
309 MP_MODIFY(r
, MP_LEN(a
) + 2);
310 memcpy(r
->v
, a
->v
, MPWS(MP_LEN(a
)));
311 memset(r
->v
+ MP_LEN(a
), 0, MPWS(2));
315 /* --- Fix up the quotient too --- */
317 MP_MODIFY(q
, MP_LEN(a
));
319 /* --- Perform the calculation --- */
321 mpx_udiv(q
->v
, q
->vl
, r
->v
, r
->vl
, b
->v
, b
->vl
, sv
, svl
);
323 /* --- Sort out the sign of the results --- *
325 * If the signs of the arguments differ, and the remainder is nonzero, I
326 * must add one to the absolute value of the quotient and subtract the
327 * remainder from @b@.
330 q
->f
= ((a
->f
| b
->f
) & MP_BURN
) | ((a
->f
^ b
->f
) & MP_NEG
);
333 for (v
= r
->v
; v
< r
->vl
; v
++) {
335 MPX_UADDN(q
->v
, q
->vl
, 1);
336 mpx_usub(r
->v
, r
->vl
, b
->v
, b
->vl
, r
->v
, r
->vl
);
342 r
->f
= ((a
->f
| b
->f
) & MP_BURN
) | (b
->f
& MP_NEG
);
344 /* --- Store the return values --- */
365 /*----- Test rig ----------------------------------------------------------*/
369 static int verify(const char *op
, mp
*expect
, mp
*result
, mp
*a
, mp
*b
)
371 if (MP_CMP(expect
, !=, result
)) {
372 fprintf(stderr
, "\n*** %s failed", op
);
373 fputs("\n*** a = ", stderr
); mp_writefile(a
, stderr
, 10);
374 fputs("\n*** b = ", stderr
); mp_writefile(b
, stderr
, 10);
375 fputs("\n*** result = ", stderr
); mp_writefile(result
, stderr
, 10);
376 fputs("\n*** expect = ", stderr
); mp_writefile(expect
, stderr
, 10);
383 #define RIG(name, op) \
384 static int t##name(dstr *v) \
386 mp *a = *(mp **)v[0].buf; \
387 mpw n = *(int *)v[1].buf; \
389 mp *r = *(mp **)v[2].buf; \
390 mp *c = op(MP_NEW, a, n); \
392 mp_build(&b, &n, &n + 1); \
393 ok = verify(#name, r, c, a, &b); \
394 mp_drop(a); mp_drop(c); mp_drop(r); \
395 assert(mparena_count(MPARENA_GLOBAL) == 0); \
404 #define RIG(name, op) \
405 static int t##name(dstr *v) \
407 mp *a = *(mp **)v[0].buf; \
408 mp *b = *(mp **)v[1].buf; \
409 mp *r = *(mp **)v[2].buf; \
410 mp *c = op(MP_NEW, a, b); \
411 int ok = verify(#name, r, c, a, b); \
412 mp_drop(a); mp_drop(b); mp_drop(c); mp_drop(r); \
413 assert(mparena_count(MPARENA_GLOBAL) == 0); \
423 static int tdiv(dstr
*v
)
425 mp
*a
= *(mp
**)v
[0].buf
;
426 mp
*b
= *(mp
**)v
[1].buf
;
427 mp
*q
= *(mp
**)v
[2].buf
;
428 mp
*r
= *(mp
**)v
[3].buf
;
429 mp
*c
= MP_NEW
, *d
= MP_NEW
;
431 mp_div(&c
, &d
, a
, b
);
432 ok
&= verify("div(quotient)", q
, c
, a
, b
);
433 ok
&= verify("div(remainder)", r
, d
, a
, b
);
434 mp_drop(a
); mp_drop(b
); mp_drop(c
); mp_drop(d
); mp_drop(r
); mp_drop(q
);
435 assert(mparena_count(MPARENA_GLOBAL
) == 0);
439 static test_chunk tests
[] = {
440 { "lsl", tlsl
, { &type_mp
, &type_mp
, &type_mp
, 0 } },
441 { "lsr", tlsr
, { &type_mp
, &type_mp
, &type_mp
, 0 } },
442 { "add", tadd
, { &type_mp
, &type_mp
, &type_mp
, 0 } },
443 { "sub", tsub
, { &type_mp
, &type_mp
, &type_mp
, 0 } },
444 { "mul", tmul
, { &type_mp
, &type_mp
, &type_mp
, 0 } },
445 { "div", tdiv
, { &type_mp
, &type_mp
, &type_mp
, &type_mp
, 0 } },
449 int main(int argc
, char *argv
[])
452 test_run(argc
, argv
, tests
, SRCDIR
"/tests/mp");
458 /*----- That's all, folks -------------------------------------------------*/