bignum.c 12.4 KB
Newer Older
Paul Eggert's avatar
Paul Eggert committed
1 2
/* Big numbers for Emacs.

Paul Eggert's avatar
Paul Eggert committed
3
Copyright 2018-2020 Free Software Foundation, Inc.
Paul Eggert's avatar
Paul Eggert committed
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

This file is part of GNU Emacs.

GNU Emacs is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or (at
your option) any later version.

GNU Emacs 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 General Public License for more details.

You should have received a copy of the GNU General Public License
along with GNU Emacs.  If not, see <https://www.gnu.org/licenses/>.  */

#include <config.h>

#include "bignum.h"

#include "lisp.h"

Paul Eggert's avatar
Paul Eggert committed
26
#include <math.h>
27 28
#include <stdlib.h>

29 30 31 32 33
/* mpz global temporaries.  Making them global saves the trouble of
   properly using mpz_init and mpz_clear on temporaries even when
   storage is exhausted.  Admittedly this is not ideal.  An mpz value
   in a temporary is made permanent by mpz_swapping it with a bignum's
   value.  Although typically at most two temporaries are needed,
34 35
   rounddiv_q and rounding_driver both need four and time_arith needs
   five.  */
36

37
mpz_t mpz[5];
38

Paul Eggert's avatar
Paul Eggert committed
39 40 41 42 43 44 45 46 47 48 49 50
static void *
xrealloc_for_gmp (void *ptr, size_t ignore, size_t size)
{
  return xrealloc (ptr, size);
}

static void
xfree_for_gmp (void *ptr, size_t ignore)
{
  xfree (ptr);
}

51
void
Paul Eggert's avatar
Paul Eggert committed
52
init_bignum (void)
53
{
Paul Eggert's avatar
Paul Eggert committed
54 55 56 57
  eassert (mp_bits_per_limb == GMP_NUMB_BITS);
  integer_width = 1 << 16;
  mp_set_memory_functions (xmalloc, xrealloc_for_gmp, xfree_for_gmp);

58 59 60 61
  for (int i = 0; i < ARRAYELTS (mpz); i++)
    mpz_init (mpz[i]);
}

Paul Eggert's avatar
Paul Eggert committed
62 63 64 65
/* Return the value of the Lisp bignum N, as a double.  */
double
bignum_to_double (Lisp_Object n)
{
66
  return mpz_get_d_rounded (*xbignum_val (n));
Paul Eggert's avatar
Paul Eggert committed
67 68
}

Paul Eggert's avatar
Paul Eggert committed
69 70
/* Return D, converted to a Lisp integer.  Discard any fraction.
   Signal an error if D cannot be converted.  */
Paul Eggert's avatar
Paul Eggert committed
71
Lisp_Object
72
double_to_integer (double d)
Paul Eggert's avatar
Paul Eggert committed
73
{
Paul Eggert's avatar
Paul Eggert committed
74 75
  if (!isfinite (d))
    overflow_error ();
76 77
  mpz_set_d (mpz[0], d);
  return make_integer_mpz ();
Paul Eggert's avatar
Paul Eggert committed
78 79
}

80 81
/* Return a Lisp integer equal to mpz[0], which has BITS bits and which
   must not be in fixnum range.  Set mpz[0] to a junk value.  */
Paul Eggert's avatar
Paul Eggert committed
82
static Lisp_Object
83
make_bignum_bits (size_t bits)
Paul Eggert's avatar
Paul Eggert committed
84 85
{
  /* The documentation says integer-width should be nonnegative, so
86 87 88 89 90
     comparing it to BITS works even though BITS is unsigned.  Treat
     integer-width as if it were at least twice the machine integer width,
     so that timefns.c can safely use bignums for double-precision
     timestamps.  */
  if (integer_width < bits && 2 * max (INTMAX_WIDTH, UINTMAX_WIDTH) < bits)
91
    overflow_error ();
Paul Eggert's avatar
Paul Eggert committed
92

93 94
  struct Lisp_Bignum *b = ALLOCATE_PLAIN_PSEUDOVECTOR (struct Lisp_Bignum,
						       PVEC_BIGNUM);
95 96
  mpz_init (b->value);
  mpz_swap (b->value, mpz[0]);
Paul Eggert's avatar
Paul Eggert committed
97 98 99
  return make_lisp_ptr (b, Lisp_Vectorlike);
}

100 101
/* Return a Lisp integer equal to mpz[0], which must not be in fixnum range.
   Set mpz[0] to a junk value.  */
Paul Eggert's avatar
Paul Eggert committed
102
static Lisp_Object
103
make_bignum (void)
Paul Eggert's avatar
Paul Eggert committed
104
{
105
  return make_bignum_bits (mpz_sizeinbase (mpz[0], 2));
Paul Eggert's avatar
Paul Eggert committed
106 107 108 109 110 111 112
}

/* Return a Lisp integer equal to N, which must not be in fixnum range.  */
Lisp_Object
make_bigint (intmax_t n)
{
  eassert (FIXNUM_OVERFLOW_P (n));
113 114
  mpz_set_intmax (mpz[0], n);
  return make_bignum ();
Paul Eggert's avatar
Paul Eggert committed
115
}
116 117 118 119
Lisp_Object
make_biguint (uintmax_t n)
{
  eassert (FIXNUM_OVERFLOW_P (n));
120 121
  mpz_set_uintmax (mpz[0], n);
  return make_bignum ();
122
}
Paul Eggert's avatar
Paul Eggert committed
123

124 125 126 127 128 129 130 131 132 133
/* Return a Lisp integer equal to -N, which must not be in fixnum range.  */
Lisp_Object
make_neg_biguint (uintmax_t n)
{
  eassert (-MOST_NEGATIVE_FIXNUM < n);
  mpz_set_uintmax (mpz[0], n);
  mpz_neg (mpz[0], mpz[0]);
  return make_bignum ();
}

134 135
/* Return a Lisp integer with value taken from mpz[0].
   Set mpz[0] to a junk value.  */
Paul Eggert's avatar
Paul Eggert committed
136
Lisp_Object
137
make_integer_mpz (void)
Paul Eggert's avatar
Paul Eggert committed
138
{
139
  size_t bits = mpz_sizeinbase (mpz[0], 2);
Paul Eggert's avatar
Paul Eggert committed
140 141 142 143 144 145 146 147

  if (bits <= FIXNUM_BITS)
    {
      EMACS_INT v = 0;
      int i = 0, shift = 0;

      do
	{
148
	  EMACS_INT limb = mpz_getlimbn (mpz[0], i++);
Paul Eggert's avatar
Paul Eggert committed
149 150 151 152 153
	  v += limb << shift;
	  shift += GMP_NUMB_BITS;
	}
      while (shift < bits);

154
      if (mpz_sgn (mpz[0]) < 0)
Paul Eggert's avatar
Paul Eggert committed
155 156 157 158 159 160
	v = -v;

      if (!FIXNUM_OVERFLOW_P (v))
	return make_fixnum (v);
    }

161
  return make_bignum_bits (bits);
Paul Eggert's avatar
Paul Eggert committed
162 163
}

164
/* Set RESULT to V.  This code is for when intmax_t is wider than long.  */
Paul Eggert's avatar
Paul Eggert committed
165 166 167
void
mpz_set_intmax_slow (mpz_t result, intmax_t v)
{
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187
  int maxlimbs = (INTMAX_WIDTH + GMP_NUMB_BITS - 1) / GMP_NUMB_BITS;
  mp_limb_t *limb = mpz_limbs_write (result, maxlimbs);
  int n = 0;
  uintmax_t u = v;
  bool negative = v < 0;
  if (negative)
    {
      uintmax_t two = 2;
      u = -u & ((two << (UINTMAX_WIDTH - 1)) - 1);
    }

  do
    {
      limb[n++] = u;
      u = GMP_NUMB_BITS < UINTMAX_WIDTH ? u >> GMP_NUMB_BITS : 0;
    }
  while (u != 0);

  mpz_limbs_finish (result, negative ? -n : n);
}
188
void
189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
mpz_set_uintmax_slow (mpz_t result, uintmax_t v)
{
  int maxlimbs = (UINTMAX_WIDTH + GMP_NUMB_BITS - 1) / GMP_NUMB_BITS;
  mp_limb_t *limb = mpz_limbs_write (result, maxlimbs);
  int n = 0;

  do
    {
      limb[n++] = v;
      v = GMP_NUMB_BITS < INTMAX_WIDTH ? v >> GMP_NUMB_BITS : 0;
    }
  while (v != 0);

  mpz_limbs_finish (result, n);
}

205 206 207 208
/* If Z fits into *PI, store its value there and return true.
   Return false otherwise.  */
bool
mpz_to_intmax (mpz_t const z, intmax_t *pi)
209
{
210 211
  ptrdiff_t bits = mpz_sizeinbase (z, 2);
  bool negative = mpz_sgn (z) < 0;
212 213 214 215 216 217 218 219

  if (bits < INTMAX_WIDTH)
    {
      intmax_t v = 0;
      int i = 0, shift = 0;

      do
	{
220
	  intmax_t limb = mpz_getlimbn (z, i++);
221 222 223 224 225
	  v += limb << shift;
	  shift += GMP_NUMB_BITS;
	}
      while (shift < bits);

226 227 228 229 230 231 232 233
      *pi = negative ? -v : v;
      return true;
    }
  if (bits == INTMAX_WIDTH && INTMAX_MIN < -INTMAX_MAX && negative
      && mpz_scan1 (z, 0) == INTMAX_WIDTH - 1)
    {
      *pi = INTMAX_MIN;
      return true;
234
    }
235
  return false;
236
}
237 238
bool
mpz_to_uintmax (mpz_t const z, uintmax_t *pi)
239
{
240 241 242 243 244 245
  if (mpz_sgn (z) < 0)
    return false;
  ptrdiff_t bits = mpz_sizeinbase (z, 2);
  if (UINTMAX_WIDTH < bits)
    return false;

246
  uintmax_t v = 0;
247 248 249
  int i = 0, shift = 0;

  do
250
    {
251 252 253
      uintmax_t limb = mpz_getlimbn (z, i++);
      v += limb << shift;
      shift += GMP_NUMB_BITS;
254
    }
255 256 257 258 259 260 261 262 263 264 265 266
  while (shift < bits);

  *pi = v;
  return true;
}

/* Return the value of the bignum X if it fits, 0 otherwise.
   A bignum cannot be zero, so 0 indicates failure reliably.  */
intmax_t
bignum_to_intmax (Lisp_Object x)
{
  intmax_t i;
267
  return mpz_to_intmax (*xbignum_val (x), &i) ? i : 0;
268 269 270 271 272
}
uintmax_t
bignum_to_uintmax (Lisp_Object x)
{
  uintmax_t i;
273
  return mpz_to_uintmax (*xbignum_val (x), &i) ? i : 0;
Paul Eggert's avatar
Paul Eggert committed
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

/* Multiply and exponentiate mpz_t values without aborting due to size
   limits.  */

/* GMP tests for this value and aborts (!) if it is exceeded.
   This is as of GMP 6.1.2 (2016); perhaps future versions will differ.  */
enum { GMP_NLIMBS_MAX = min (INT_MAX, ULONG_MAX / GMP_NUMB_BITS) };

/* An upper bound on limb counts, needed to prevent libgmp and/or
   Emacs from aborting or otherwise misbehaving.  This bound applies
   to estimates of mpz_t sizes before the mpz_t objects are created,
   as opposed to integer-width which operates on mpz_t values after
   creation and before conversion to Lisp bignums.  */
enum
  {
   NLIMBS_LIMIT = min (min (/* libgmp needs to store limb counts.  */
			    GMP_NLIMBS_MAX,

			    /* Size calculations need to work.  */
			    min (PTRDIFF_MAX, SIZE_MAX) / sizeof (mp_limb_t)),

		       /* Emacs puts bit counts into fixnums.  */
		       MOST_POSITIVE_FIXNUM / GMP_NUMB_BITS)
  };

/* Like mpz_size, but tell the compiler the result is a nonnegative int.  */

static int
emacs_mpz_size (mpz_t const op)
{
  mp_size_t size = mpz_size (op);
  eassume (0 <= size && size <= INT_MAX);
  return size;
}

/* Wrappers to work around GMP limitations.  As of GMP 6.1.2 (2016),
   the library code aborts when a number is too large.  These wrappers
   avoid the problem for functions that can return numbers much larger
   than their arguments.  For slowly-growing numbers, the integer
   width checks in bignum.c should suffice.  */

void
emacs_mpz_mul (mpz_t rop, mpz_t const op1, mpz_t const op2)
{
  if (NLIMBS_LIMIT - emacs_mpz_size (op1) < emacs_mpz_size (op2))
    overflow_error ();
  mpz_mul (rop, op1, op2);
}

void
emacs_mpz_mul_2exp (mpz_t rop, mpz_t const op1, EMACS_INT op2)
{
  /* Fudge factor derived from GMP 6.1.2, to avoid an abort in
     mpz_mul_2exp (look for the '+ 1' in its source code).  */
  enum { mul_2exp_extra_limbs = 1 };
  enum { lim = min (NLIMBS_LIMIT, GMP_NLIMBS_MAX - mul_2exp_extra_limbs) };

  EMACS_INT op2limbs = op2 / GMP_NUMB_BITS;
  if (lim - emacs_mpz_size (op1) < op2limbs)
    overflow_error ();
  mpz_mul_2exp (rop, op1, op2);
}

void
emacs_mpz_pow_ui (mpz_t rop, mpz_t const base, unsigned long exp)
{
  /* This fudge factor is derived from GMP 6.1.2, to avoid an abort in
     mpz_n_pow_ui (look for the '5' in its source code).  */
  enum { pow_ui_extra_limbs = 5 };
  enum { lim = min (NLIMBS_LIMIT, GMP_NLIMBS_MAX - pow_ui_extra_limbs) };

  int nbase = emacs_mpz_size (base), n;
  if (INT_MULTIPLY_WRAPV (nbase, exp, &n) || lim < n)
    overflow_error ();
  mpz_pow_ui (rop, base, exp);
}


354
/* Yield an upper bound on the buffer size needed to contain a C
355
   string representing the NUM in base BASE.  This includes any
356
   preceding '-' and the terminating NUL.  */
357 358 359 360 361
static ptrdiff_t
mpz_bufsize (mpz_t const num, int base)
{
  return mpz_sizeinbase (num, base) + 2;
}
362 363 364
ptrdiff_t
bignum_bufsize (Lisp_Object num, int base)
{
365
  return mpz_bufsize (*xbignum_val (num), base);
366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388
}

/* Convert NUM to a nearest double, as opposed to mpz_get_d which
   truncates toward zero.  */
double
mpz_get_d_rounded (mpz_t const num)
{
  ptrdiff_t size = mpz_bufsize (num, 10);

  /* Use mpz_get_d as a shortcut for a bignum so small that rounding
     errors cannot occur, which is possible if EMACS_INT (not counting
     sign) has fewer bits than a double significand.  */
  if (! ((FLT_RADIX == 2 && DBL_MANT_DIG <= FIXNUM_BITS - 1)
	 || (FLT_RADIX == 16 && DBL_MANT_DIG * 4 <= FIXNUM_BITS - 1))
      && size <= DBL_DIG + 2)
    return mpz_get_d (num);

  USE_SAFE_ALLOCA;
  char *buf = SAFE_ALLOCA (size);
  mpz_get_str (buf, 10, num);
  double result = strtod (buf, NULL);
  SAFE_FREE ();
  return result;
389 390 391 392 393 394 395 396 397 398
}

/* Store into BUF (of size SIZE) the value of NUM as a base-BASE string.
   If BASE is negative, use upper-case digits in base -BASE.
   Return the string's length.
   SIZE must equal bignum_bufsize (NUM, abs (BASE)).  */
ptrdiff_t
bignum_to_c_string (char *buf, ptrdiff_t size, Lisp_Object num, int base)
{
  eassert (bignum_bufsize (num, abs (base)) == size);
399
  mpz_get_str (buf, base, *xbignum_val (num));
400 401 402 403 404 405
  ptrdiff_t n = size - 2;
  return !buf[n - 1] ? n - 1 : n + !!buf[n];
}

/* Convert NUM to a base-BASE Lisp string.
   If BASE is negative, use upper-case digits in base -BASE.  */
Paul Eggert's avatar
Paul Eggert committed
406 407 408 409

Lisp_Object
bignum_to_string (Lisp_Object num, int base)
{
410
  ptrdiff_t size = bignum_bufsize (num, abs (base));
Paul Eggert's avatar
Paul Eggert committed
411
  USE_SAFE_ALLOCA;
412 413 414
  char *str = SAFE_ALLOCA (size);
  ptrdiff_t len = bignum_to_c_string (str, size, num, base);
  Lisp_Object result = make_unibyte_string (str, len);
Paul Eggert's avatar
Paul Eggert committed
415 416 417 418 419 420
  SAFE_FREE ();
  return result;
}

/* Create a bignum by scanning NUM, with digits in BASE.
   NUM must consist of an optional '-', a nonempty sequence
421
   of base-BASE digits, and a terminating NUL byte, and
Paul Eggert's avatar
Paul Eggert committed
422 423 424 425 426
   the represented number must not be in fixnum range.  */

Lisp_Object
make_bignum_str (char const *num, int base)
{
427 428
  struct Lisp_Bignum *b = ALLOCATE_PLAIN_PSEUDOVECTOR (struct Lisp_Bignum,
						       PVEC_BIGNUM);
Paul Eggert's avatar
Paul Eggert committed
429 430 431 432 433
  mpz_init (b->value);
  int check = mpz_set_str (b->value, num, base);
  eassert (check == 0);
  return make_lisp_ptr (b, Lisp_Vectorlike);
}
Paul Eggert's avatar
Paul Eggert committed
434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467

/* Check that X is a Lisp integer in the range LO..HI.
   Return X's value as an intmax_t.  */

intmax_t
check_integer_range (Lisp_Object x, intmax_t lo, intmax_t hi)
{
  CHECK_INTEGER (x);
  intmax_t i;
  if (! (integer_to_intmax (x, &i) && lo <= i && i <= hi))
    args_out_of_range_3 (x, make_int (lo), make_int (hi));
  return i;
}

/* Check that X is a Lisp integer in the range 0..HI.
   Return X's value as an uintmax_t.  */

uintmax_t
check_uinteger_max (Lisp_Object x, uintmax_t hi)
{
  CHECK_INTEGER (x);
  uintmax_t i;
  if (! (integer_to_uintmax (x, &i) && i <= hi))
    args_out_of_range_3 (x, make_fixnum (0), make_uint (hi));
  return i;
}

/* Check that X is a Lisp integer no greater than INT_MAX,
   and return its value or zero, whichever is greater.  */

int
check_int_nonnegative (Lisp_Object x)
{
  CHECK_INTEGER (x);
468
  return NILP (Fnatnump (x)) ? 0 : check_integer_range (x, 0, INT_MAX);
Paul Eggert's avatar
Paul Eggert committed
469
}