/* Definitions of numeric types for XEmacs using the MP library. Copyright (C) 2004 Jerry James. This file is part of XEmacs. XEmacs 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. XEmacs 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 XEmacs. If not, see . */ /* Synched up with: Not in FSF. */ /* This library defines the following types: bignum = MINT The MP library does not include support for ratios or bigfloats. */ #ifndef INCLUDED_number_mp_h_ #define INCLUDED_number_mp_h_ /* BSD MP libraries without MP_PREFIX define a function named pow in mp.h that has a different prototype from the one in math.h. We don't use that function anyway, so we do this for safety purposes. However, this means that number-mp.h must always be included before math.h. */ #define pow mp_pow BEGIN_C_DECLS #include END_C_DECLS #undef pow #ifdef MP_PREFIX #define MP_GCD mp_gcd #define MP_ITOM mp_itom #define MP_XTOM mp_xtom #define MP_MADD mp_madd #define MP_MCMP mp_mcmp #define MP_MDIV mp_mdiv #define MP_MFREE mp_mfree #define MP_MSUB mp_msub #define MP_MULT mp_mult #define MP_SDIV mp_sdiv #ifdef HAVE_MP_MOVE #define MP_MOVE(x,y) mp_move (x, y) #else #define MP_MOVE(x,y) mp_madd (x, bignum_zero, y) #endif #else #define MP_GCD gcd #define MP_ITOM itom #define MP_XTOM xtom #define MP_MADD madd #define MP_MCMP mcmp #define MP_MDIV mdiv #define MP_MFREE mfree #define MP_MSUB msub #define MP_MULT mult #define MP_SDIV sdiv #ifdef HAVE_MP_MOVE #define MP_MOVE(x,y) move (x, y) #else #define MP_MOVE(x,y) madd (x, bignum_zero, y) #endif #endif typedef MINT *bignum; extern void init_number_mp(void); /********************************* Bignums **********************************/ #define HAVE_BIGNUM 1 extern MINT *bignum_zero, *intern_bignum; extern MINT *bignum_min_int, *bignum_max_int, *bignum_max_uint; extern MINT *bignum_min_long, *bignum_max_long, *bignum_max_ulong; extern MINT *bignum_min_llong, *bignum_max_llong, *bignum_max_ullong; extern short div_rem; /***** Bignum: basic functions *****/ #define bignum_init(b) (b = MP_ITOM (0)) #define bignum_fini(b) MP_MFREE (b) #define bignum_hashcode(b) bignum_to_uint (b) #define bignum_sign(b) MP_MCMP (b, bignum_zero) #define bignum_evenp(b) (MP_SDIV (b, 2, intern_bignum, &div_rem), \ div_rem == 0) #define bignum_oddp(b) (MP_SDIV (b, 2, intern_bignum, &div_rem), \ div_rem != 0) /***** Bignum: size *****/ #define bignum_fits_int_p(b) (MP_MCMP (b, bignum_min_int) >= 0 && \ MP_MCMP (b, bignum_max_int) <= 0) #define bignum_fits_uint_p(b) (MP_MCMP (b, bignum_zero) >= 0 && \ MP_MCMP (b, bignum_max_uint) <= 0) #define bignum_fits_long_p(b) (MP_MCMP (b, bignum_min_long) >= 0 && \ MP_MCMP (b, bignum_max_long) <= 0) #define bignum_fits_ulong_p(b) (MP_MCMP (b, bignum_zero) >= 0 && \ MP_MCMP (b, bignum_max_ulong) <= 0) #define bignum_fits_llong_p(b) (MP_MCMP (b, bignum_min_llong) >= 0 && \ MP_MCMP (b, bignum_max_llong) <= 0) #define bignum_fits_ullong_p(b) (MP_MCMP (b, bignum_zero) >= 0 && \ MP_MCMP (b, bignum_max_ullong) <= 0) /***** Bignum: conversions *****/ extern char *bignum_to_string(bignum, int); extern int bignum_to_int(bignum); extern unsigned int bignum_to_uint(bignum); extern long bignum_to_long(bignum); extern unsigned long bignum_to_ulong(bignum); extern long long bignum_to_llong(bignum); extern unsigned long long bignum_to_ullong(bignum); extern double bignum_to_double(bignum); /***** Bignum: converting assignments *****/ #define bignum_set(b1, b2) MP_MOVE (b2, b1) extern int bignum_set_string(bignum, const char *, int); extern void bignum_set_long(bignum, long); extern void bignum_set_ulong(bignum, unsigned long); extern void bignum_set_llong(bignum, long long); extern void bignum_set_ullong(bignum, unsigned long long); extern void bignum_set_double(bignum, double); /***** Bignum: comparisons *****/ #define bignum_cmp(b1,b2) MP_MCMP (b1, b2) #define bignum_lt(b1,b2) (MP_MCMP (b1, b2) < 0) #define bignum_le(b1,b2) (MP_MCMP (b1, b2) <= 0) #define bignum_eql(b1,b2) (MP_MCMP (b1, b2) == 0) #define bignum_ge(b1,b2) (MP_MCMP (b1, b2) >= 0) #define bignum_gt(b1,b2) (MP_MCMP (b1, b2) > 0) /***** Bignum: arithmetic *****/ #define bignum_neg(b,b2) MP_MSUB (bignum_zero, b2, b) #define bignum_abs(b,b2) (MP_MCMP (b2, bignum_zero) < 0 \ ? MP_MSUB (bignum_zero, b2, b) \ : MP_MADD (bignum_zero, b2, b)) #define bignum_add(b,b1,b2) MP_MADD (b1, b2, b) #define bignum_sub(b,b1,b2) MP_MSUB (b1, b2, b) #define bignum_mul(b,b1,b2) MP_MULT (b1, b2, b) extern int bignum_divisible_p(bignum, bignum); #define bignum_div(b,b1,b2) MP_MDIV (b1, b2, b, intern_bignum) extern void bignum_ceil(bignum, bignum, bignum); extern void bignum_floor(bignum, bignum, bignum); #define bignum_mod(b,b1,b2) MP_MDIV (b1, b2, intern_bignum, b) extern void bignum_pow(bignum, bignum, unsigned long); #define bignum_gcd(res,b1,b2) MP_GCD (b1, b2, res) extern void bignum_lcm(bignum, bignum, bignum); /***** Bignum: bit manipulations *****/ extern void bignum_and(bignum, bignum, bignum); extern void bignum_ior(bignum, bignum, bignum); extern void bignum_xor(bignum, bignum, bignum); extern void bignum_not(bignum, bignum); extern void bignum_setbit(bignum, unsigned long); extern void bignum_clrbit(bignum, unsigned long); extern int bignum_testbit(bignum, unsigned long); extern void bignum_lshift(bignum, bignum, unsigned long); extern void bignum_rshift(bignum, bignum, unsigned long); /***** Bignum: random numbers *****/ #define bignum_random_seed(s) extern void bignum_random(bignum, bignum); #endif /* INCLUDED_number_mp_h_ */