about summary refs log tree commit diff
path: root/stdlib/mul_n.c
blob: f0a9a304dd07c435fc1fd2d5649a457aa74c7ae9 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
/* mpn_mul_n -- Multiply two natural numbers of length n.

Copyright (C) 1991-2013 Free Software Foundation, Inc.

This file is part of the GNU MP Library.

The GNU MP Library is free software; you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation; either version 2.1 of the License, or (at your
option) any later version.

The GNU MP Library is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
License for more details.

You should have received a copy of the GNU Lesser General Public License
along with the GNU MP Library; see the file COPYING.LIB.  If not, see
<http://www.gnu.org/licenses/>.  */

#include <gmp.h>
#include "gmp-impl.h"

/* Multiply the natural numbers u (pointed to by UP) and v (pointed to by VP),
   both with SIZE limbs, and store the result at PRODP.  2 * SIZE limbs are
   always stored.  Return the most significant limb.

   Argument constraints:
   1. PRODP != UP and PRODP != VP, i.e. the destination
      must be distinct from the multiplier and the multiplicand.  */

/* If KARATSUBA_THRESHOLD is not already defined, define it to a
   value which is good on most machines.  */
#ifndef KARATSUBA_THRESHOLD
#define KARATSUBA_THRESHOLD 32
#endif

/* The code can't handle KARATSUBA_THRESHOLD smaller than 2.  */
#if KARATSUBA_THRESHOLD < 2
#undef KARATSUBA_THRESHOLD
#define KARATSUBA_THRESHOLD 2
#endif

/* Handle simple cases with traditional multiplication.

   This is the most critical code of multiplication.  All multiplies rely
   on this, both small and huge.  Small ones arrive here immediately.  Huge
   ones arrive here as this is the base case for Karatsuba's recursive
   algorithm below.  */

void
#if __STDC__
impn_mul_n_basecase (mp_ptr prodp, mp_srcptr up, mp_srcptr vp, mp_size_t size)
#else
impn_mul_n_basecase (prodp, up, vp, size)
     mp_ptr prodp;
     mp_srcptr up;
     mp_srcptr vp;
     mp_size_t size;
#endif
{
  mp_size_t i;
  mp_limb_t cy_limb;
  mp_limb_t v_limb;

  /* Multiply by the first limb in V separately, as the result can be
     stored (not added) to PROD.  We also avoid a loop for zeroing.  */
  v_limb = vp[0];
  if (v_limb <= 1)
    {
      if (v_limb == 1)
	MPN_COPY (prodp, up, size);
      else
	MPN_ZERO (prodp, size);
      cy_limb = 0;
    }
  else
    cy_limb = mpn_mul_1 (prodp, up, size, v_limb);

  prodp[size] = cy_limb;
  prodp++;

  /* For each iteration in the outer loop, multiply one limb from
     U with one limb from V, and add it to PROD.  */
  for (i = 1; i < size; i++)
    {
      v_limb = vp[i];
      if (v_limb <= 1)
	{
	  cy_limb = 0;
	  if (v_limb == 1)
	    cy_limb = mpn_add_n (prodp, prodp, up, size);
	}
      else
	cy_limb = mpn_addmul_1 (prodp, up, size, v_limb);

      prodp[size] = cy_limb;
      prodp++;
    }
}

void
#if __STDC__
impn_mul_n (mp_ptr prodp,
	     mp_srcptr up, mp_srcptr vp, mp_size_t size, mp_ptr tspace)
#else
impn_mul_n (prodp, up, vp, size, tspace)
     mp_ptr prodp;
     mp_srcptr up;
     mp_srcptr vp;
     mp_size_t size;
     mp_ptr tspace;
#endif
{
  if ((size & 1) != 0)
    {
      /* The size is odd, the code code below doesn't handle that.
	 Multiply the least significant (size - 1) limbs with a recursive
	 call, and handle the most significant limb of S1 and S2
	 separately.  */
      /* A slightly faster way to do this would be to make the Karatsuba
	 code below behave as if the size were even, and let it check for
	 odd size in the end.  I.e., in essence move this code to the end.
	 Doing so would save us a recursive call, and potentially make the
	 stack grow a lot less.  */

      mp_size_t esize = size - 1;	/* even size */
      mp_limb_t cy_limb;

      MPN_MUL_N_RECURSE (prodp, up, vp, esize, tspace);
      cy_limb = mpn_addmul_1 (prodp + esize, up, esize, vp[esize]);
      prodp[esize + esize] = cy_limb;
      cy_limb = mpn_addmul_1 (prodp + esize, vp, size, up[esize]);

      prodp[esize + size] = cy_limb;
    }
  else
    {
      /* Anatolij Alekseevich Karatsuba's divide-and-conquer algorithm.

	 Split U in two pieces, U1 and U0, such that
	 U = U0 + U1*(B**n),
	 and V in V1 and V0, such that
	 V = V0 + V1*(B**n).

	 UV is then computed recursively using the identity

		2n   n          n                     n
	 UV = (B  + B )U V  +  B (U -U )(V -V )  +  (B + 1)U V
			1 1        1  0   0  1              0 0

	 Where B = 2**BITS_PER_MP_LIMB.  */

      mp_size_t hsize = size >> 1;
      mp_limb_t cy;
      int negflg;

      /*** Product H.	 ________________  ________________
			|_____U1 x V1____||____U0 x V0_____|  */
      /* Put result in upper part of PROD and pass low part of TSPACE
	 as new TSPACE.  */
      MPN_MUL_N_RECURSE (prodp + size, up + hsize, vp + hsize, hsize, tspace);

      /*** Product M.	 ________________
			|_(U1-U0)(V0-V1)_|  */
      if (mpn_cmp (up + hsize, up, hsize) >= 0)
	{
	  mpn_sub_n (prodp, up + hsize, up, hsize);
	  negflg = 0;
	}
      else
	{
	  mpn_sub_n (prodp, up, up + hsize, hsize);
	  negflg = 1;
	}
      if (mpn_cmp (vp + hsize, vp, hsize) >= 0)
	{
	  mpn_sub_n (prodp + hsize, vp + hsize, vp, hsize);
	  negflg ^= 1;
	}
      else
	{
	  mpn_sub_n (prodp + hsize, vp, vp + hsize, hsize);
	  /* No change of NEGFLG.  */
	}
      /* Read temporary operands from low part of PROD.
	 Put result in low part of TSPACE using upper part of TSPACE
	 as new TSPACE.  */
      MPN_MUL_N_RECURSE (tspace, prodp, prodp + hsize, hsize, tspace + size);

      /*** Add/copy product H.  */
      MPN_COPY (prodp + hsize, prodp + size, hsize);
      cy = mpn_add_n (prodp + size, prodp + size, prodp + size + hsize, hsize);

      /*** Add product M (if NEGFLG M is a negative number).  */
      if (negflg)
	cy -= mpn_sub_n (prodp + hsize, prodp + hsize, tspace, size);
      else
	cy += mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);

      /*** Product L.	 ________________  ________________
			|________________||____U0 x V0_____|  */
      /* Read temporary operands from low part of PROD.
	 Put result in low part of TSPACE using upper part of TSPACE
	 as new TSPACE.  */
      MPN_MUL_N_RECURSE (tspace, up, vp, hsize, tspace + size);

      /*** Add/copy Product L (twice).  */

      cy += mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);
      if (cy)
	mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy);

      MPN_COPY (prodp, tspace, hsize);
      cy = mpn_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize);
      if (cy)
	mpn_add_1 (prodp + size, prodp + size, size, 1);
    }
}

void
#if __STDC__
impn_sqr_n_basecase (mp_ptr prodp, mp_srcptr up, mp_size_t size)
#else
impn_sqr_n_basecase (prodp, up, size)
     mp_ptr prodp;
     mp_srcptr up;
     mp_size_t size;
#endif
{
  mp_size_t i;
  mp_limb_t cy_limb;
  mp_limb_t v_limb;

  /* Multiply by the first limb in V separately, as the result can be
     stored (not added) to PROD.  We also avoid a loop for zeroing.  */
  v_limb = up[0];
  if (v_limb <= 1)
    {
      if (v_limb == 1)
	MPN_COPY (prodp, up, size);
      else
	MPN_ZERO (prodp, size);
      cy_limb = 0;
    }
  else
    cy_limb = mpn_mul_1 (prodp, up, size, v_limb);

  prodp[size] = cy_limb;
  prodp++;

  /* For each iteration in the outer loop, multiply one limb from
     U with one limb from V, and add it to PROD.  */
  for (i = 1; i < size; i++)
    {
      v_limb = up[i];
      if (v_limb <= 1)
	{
	  cy_limb = 0;
	  if (v_limb == 1)
	    cy_limb = mpn_add_n (prodp, prodp, up, size);
	}
      else
	cy_limb = mpn_addmul_1 (prodp, up, size, v_limb);

      prodp[size] = cy_limb;
      prodp++;
    }
}

void
#if __STDC__
impn_sqr_n (mp_ptr prodp,
	     mp_srcptr up, mp_size_t size, mp_ptr tspace)
#else
impn_sqr_n (prodp, up, size, tspace)
     mp_ptr prodp;
     mp_srcptr up;
     mp_size_t size;
     mp_ptr tspace;
#endif
{
  if ((size & 1) != 0)
    {
      /* The size is odd, the code code below doesn't handle that.
	 Multiply the least significant (size - 1) limbs with a recursive
	 call, and handle the most significant limb of S1 and S2
	 separately.  */
      /* A slightly faster way to do this would be to make the Karatsuba
	 code below behave as if the size were even, and let it check for
	 odd size in the end.  I.e., in essence move this code to the end.
	 Doing so would save us a recursive call, and potentially make the
	 stack grow a lot less.  */

      mp_size_t esize = size - 1;	/* even size */
      mp_limb_t cy_limb;

      MPN_SQR_N_RECURSE (prodp, up, esize, tspace);
      cy_limb = mpn_addmul_1 (prodp + esize, up, esize, up[esize]);
      prodp[esize + esize] = cy_limb;
      cy_limb = mpn_addmul_1 (prodp + esize, up, size, up[esize]);

      prodp[esize + size] = cy_limb;
    }
  else
    {
      mp_size_t hsize = size >> 1;
      mp_limb_t cy;

      /*** Product H.	 ________________  ________________
			|_____U1 x U1____||____U0 x U0_____|  */
      /* Put result in upper part of PROD and pass low part of TSPACE
	 as new TSPACE.  */
      MPN_SQR_N_RECURSE (prodp + size, up + hsize, hsize, tspace);

      /*** Product M.	 ________________
			|_(U1-U0)(U0-U1)_|  */
      if (mpn_cmp (up + hsize, up, hsize) >= 0)
	{
	  mpn_sub_n (prodp, up + hsize, up, hsize);
	}
      else
	{
	  mpn_sub_n (prodp, up, up + hsize, hsize);
	}

      /* Read temporary operands from low part of PROD.
	 Put result in low part of TSPACE using upper part of TSPACE
	 as new TSPACE.  */
      MPN_SQR_N_RECURSE (tspace, prodp, hsize, tspace + size);

      /*** Add/copy product H.  */
      MPN_COPY (prodp + hsize, prodp + size, hsize);
      cy = mpn_add_n (prodp + size, prodp + size, prodp + size + hsize, hsize);

      /*** Add product M (if NEGFLG M is a negative number).  */
      cy -= mpn_sub_n (prodp + hsize, prodp + hsize, tspace, size);

      /*** Product L.	 ________________  ________________
			|________________||____U0 x U0_____|  */
      /* Read temporary operands from low part of PROD.
	 Put result in low part of TSPACE using upper part of TSPACE
	 as new TSPACE.  */
      MPN_SQR_N_RECURSE (tspace, up, hsize, tspace + size);

      /*** Add/copy Product L (twice).  */

      cy += mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);
      if (cy)
	mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy);

      MPN_COPY (prodp, tspace, hsize);
      cy = mpn_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize);
      if (cy)
	mpn_add_1 (prodp + size, prodp + size, size, 1);
    }
}

/* This should be made into an inline function in gmp.h.  */
void
#if __STDC__
mpn_mul_n (mp_ptr prodp, mp_srcptr up, mp_srcptr vp, mp_size_t size)
#else
mpn_mul_n (prodp, up, vp, size)
     mp_ptr prodp;
     mp_srcptr up;
     mp_srcptr vp;
     mp_size_t size;
#endif
{
  TMP_DECL (marker);
  TMP_MARK (marker);
  if (up == vp)
    {
      if (size < KARATSUBA_THRESHOLD)
	{
	  impn_sqr_n_basecase (prodp, up, size);
	}
      else
	{
	  mp_ptr tspace;
	  tspace = (mp_ptr) TMP_ALLOC (2 * size * BYTES_PER_MP_LIMB);
	  impn_sqr_n (prodp, up, size, tspace);
	}
    }
  else
    {
      if (size < KARATSUBA_THRESHOLD)
	{
	  impn_mul_n_basecase (prodp, up, vp, size);
	}
      else
	{
	  mp_ptr tspace;
	  tspace = (mp_ptr) TMP_ALLOC (2 * size * BYTES_PER_MP_LIMB);
	  impn_mul_n (prodp, up, vp, size, tspace);
	}
    }
  TMP_FREE (marker);
}