>From 51def5ff599a25dea16bbb1a76ca061b5e41c5a6 Mon Sep 17 00:00:00 2001 From: Paul Eggert Date: Tue, 7 Jul 2020 14:41:56 -0700 Subject: [PATCH] factor: simplify by assuming libgmp Since coreutils now always has libgmp available, if only via compiling mini-gmp.c, there's not a big need for special-purpose code for factoring 128-bit integers. On platforms that care about performance libgmp will be linked in, and on other platforms mini-gmp.c is good enough. Although there is a performance advantage on the special-purpose code for factoring 128-bit integers, it's not big enough to justify the software engineering hassle of maintaining it. * cfg.mk (_ll): Remove. (exclude_file_name_regexp--sc_useless_cpp_parens) (exclude_file_name_regexp--sc_space_before_open_paren) (exclude_file_name_regexp--sc_preprocessor_indentation) (exclude_file_name_regexp--sc_ensure_comma_after_id_est) (exclude_file_name_regexp--sc_long_lines): Omit mention of $(_ll). * src/factor.c (USE_SQUFOF, STAT_SQUFOF, USE_LONGLONG_H) (W_TYPE_SIZE, UWtype, UHWtype, UQItype, SItype, USItype, DItype) (UDItype, UQItype, SItype, USItype, DItype, UDItype) (LONGLONG_STANDALONE, ASSERT, __GMP_DECLSPEC, __clz_tab) (__GMP_GNUC_PREREQ, HAVE_HOST_CPU_FAMILY_powerpc) (factor_clz_tab, __ll_B, __ll_lowpart, __ll_highpart) (MAX_NFACTS, struct factors, umul_ppmm, udiv_qrnnd, add_ssaaaa) (rsh2, lsh2, ge2, gt2, sub_ddmmss, count_leading_zeros) (count_trailing_zeros, submod, addmod, addmod2, submod2) (HIGHBIT_TO_MASK, mod2, gcd_odd, gcd2_odd) (factor_insert_multiplicity, factor_insert, factor_insert_large) (primes_diff8, struct primes_dtab, primes_dtab) (factor_insert_refind, factor_using_division, DIVBLOCK) (binvert_table, binv, divexact_21, redcify, redcify2, mulredc) (mulredc2, powm, powm2, millerrabin, millerrabin2, prime_p) (prime2_p, factor_using_pollard_rho, factor_using_pollard_rho2) (isqrt, MAGIC64, MAGIC63, MAGIC65, MAGIC11, is_square, invtab) (div_smallq, QUEUE_SIZE, Q_FREQ_SIZE, q_freq, MIN) (factor_using_squfof, factor, strto2uintmax, struct lbuf_, lbuf) (lbuf_alloc, lbuf_flush, lbuf_putc, lbuf_putint) (print_uintmaxes, print_factors_single): Remove; no longer used. (print_factors, main): Simplify by assuming libgmp. * src/local.mk (noinst_HEADERS): Remove src/longlong.h. * src/longlong.h: Remove. --- cfg.mk | 13 +- src/factor.c | 2070 +------------------------------------------ src/local.mk | 1 - src/longlong.h | 2267 ------------------------------------------------ 4 files changed, 45 insertions(+), 4306 deletions(-) delete mode 100644 src/longlong.h diff --git a/cfg.mk b/cfg.mk index d352aac94..b89f2e9cd 100644 --- a/cfg.mk +++ b/cfg.mk @@ -855,19 +855,14 @@ exclude_file_name_regexp--sc_prohibit_fail_0 = \ exclude_file_name_regexp--sc_prohibit_test_minus_ao = *\.texi$$ exclude_file_name_regexp--sc_prohibit_atoi_atof = ^lib/euidaccess-stat\.c$$ -# longlong.h is maintained elsewhere. -_ll = ^src/longlong\.h$$ -exclude_file_name_regexp--sc_useless_cpp_parens = $(_ll) -exclude_file_name_regexp--sc_space_before_open_paren = $(_ll) - tbi_1 = ^tests/pr/|(\.mk|^man/help2man)$$ tbi_2 = ^scripts/git-hooks/(pre-commit|pre-applypatch|applypatch-msg)$$ -tbi_3 = (GNU)?[Mm]akefile(\.am)?$$|$(_ll) +tbi_3 = (GNU)?[Mm]akefile(\.am)?$$ exclude_file_name_regexp--sc_prohibit_tab_based_indentation = \ $(tbi_1)|$(tbi_2)|$(tbi_3) exclude_file_name_regexp--sc_preprocessor_indentation = \ - ^(gl/lib/rand-isaac\.[ch]|gl/tests/test-rand-isaac\.c)$$|$(_ll) + ^(gl/lib/rand-isaac\.[ch]|gl/tests/test-rand-isaac\.c)$$ exclude_file_name_regexp--sc_prohibit_stat_st_blocks = \ ^(src/system\.h|tests/du/2g\.sh)$$ @@ -889,8 +884,8 @@ exclude_file_name_regexp--sc_prohibit-gl-attributes = ^src/libstdbuf\.c$$ exclude_file_name_regexp--sc_prohibit_uppercase_id_est = \.diff$$ exclude_file_name_regexp--sc_ensure_dblspace_after_dot_before_id_est = \.diff$$ -exclude_file_name_regexp--sc_ensure_comma_after_id_est = \.diff|$(_ll)$$ -exclude_file_name_regexp--sc_long_lines = \.diff$$|$(_ll) +exclude_file_name_regexp--sc_ensure_comma_after_id_est = \.diff$$ +exclude_file_name_regexp--sc_long_lines = \.diff$$ # Augment AM_CFLAGS to include our per-directory options: AM_CFLAGS += $($(@D)_CFLAGS) diff --git a/src/factor.c b/src/factor.c index c1c35a562..2b4b419d7 100644 --- a/src/factor.c +++ b/src/factor.c @@ -15,36 +15,20 @@ along with this program. If not, see . */ /* Originally written by Paul Rubin . - Adapted for GNU, fixed to factor UINT_MAX by Jim Meyering. + Adapted for GNU by Jim Meyering. Arbitrary-precision code adapted by James Youngman from Torbjörn Granlund's factorize.c, from GNU MP version 4.2.2. In 2012, the core was rewritten by Torbjörn Granlund and Niels Möller. Contains code from GNU MP. */ -/* Efficiently factor numbers that fit in one or two words (word = uintmax_t), - or, with GMP, numbers of any size. - - Code organisation: - - There are several variants of many functions, for handling one word, two - words, and GMP's mpz_t type. If the one-word variant is called foo, the - two-word variant will be foo2, and the one for mpz_t will be mp_foo. In - some cases, the plain function variants will handle both one-word and - two-word numbers, evidenced by function arguments. - - The factoring code for two words will fall into the code for one word when - progress allows that. +/* Efficiently factor numbers of any size. Algorithm: - (1) Perform trial division using a small primes table, but without hardware - division since the primes table store inverses modulo the word base. - (The GMP variant of this code doesn't make use of the precomputed - inverses, but instead relies on GMP for fast divisibility testing.) + (1) Perform trial division. (2) Check the nature of any non-factored part using Miller-Rabin for detecting composites, and Lucas for detecting primes. - (3) Factor any remaining composite part using the Pollard-Brent rho - algorithm or if USE_SQUFOF is defined to 1, try that first. + (3) Factor any remaining composite part using Pollard-Brent rho. Status of found factors are checked again using Miller-Rabin and Lucas. We prefer using Hensel norm in the divisions, not the more familiar @@ -59,23 +43,9 @@ elsewhere. A problem is to locate the inverses not from an index, but from a prime. We might instead compute the inverse on-the-fly. - * Tune trial division table size (not forgetting that this is a standalone - program where the table will be read from disk for each invocation). - * Implement less naive powm, using k-ary exponentiation for k = 3 or perhaps k = 4. - * Try to speed trial division code for single uintmax_t numbers, i.e., the - code using DIVBLOCK. It currently runs at 2 cycles per prime (Intel SBR, - IBR), 3 cycles per prime (AMD Stars) and 5 cycles per prime (AMD BD) when - using gcc 4.6 and 4.7. Some software pipelining should help; 1, 2, and 4 - respectively cycles ought to be possible. - - * The redcify function could be vastly improved by using (plain Euclidian) - pre-inversion (such as GMP's invert_limb) and udiv_qrnnd_preinv (from - GMP's gmp-impl.h). The redcify2 function could be vastly improved using - similar methoods. These functions currently dominate run time when using - the -w option. */ /* Whether to recursively factor to prove primality, @@ -84,17 +54,6 @@ # define PROVE_PRIMALITY 1 #endif -/* Faster for certain ranges but less general. */ -#ifndef USE_SQUFOF -# define USE_SQUFOF 0 -#endif - -/* Output SQUFOF statistics. */ -#ifndef STAT_SQUFOF -# define STAT_SQUFOF 0 -#endif - - #include #include #include @@ -120,99 +79,6 @@ /* Token delimiters when reading from a file. */ #define DELIM "\n\t " -#ifndef USE_LONGLONG_H -/* With the way we use longlong.h, it's only safe to use - when UWtype = UHWtype, as there were various cases - (as can be seen in the history for longlong.h) where - for example, _LP64 was required to enable W_TYPE_SIZE==64 code, - to avoid compile time or run time issues. */ -# if LONG_MAX == INTMAX_MAX -# define USE_LONGLONG_H 1 -# endif -#endif - -#if USE_LONGLONG_H - -/* Make definitions for longlong.h to make it do what it can do for us */ - -/* bitcount for uintmax_t */ -# if UINTMAX_MAX == UINT32_MAX -# define W_TYPE_SIZE 32 -# elif UINTMAX_MAX == UINT64_MAX -# define W_TYPE_SIZE 64 -# elif UINTMAX_MAX == UINT128_MAX -# define W_TYPE_SIZE 128 -# endif - -# define UWtype uintmax_t -# define UHWtype unsigned long int -# undef UDWtype -# if HAVE_ATTRIBUTE_MODE -typedef unsigned int UQItype __attribute__ ((mode (QI))); -typedef int SItype __attribute__ ((mode (SI))); -typedef unsigned int USItype __attribute__ ((mode (SI))); -typedef int DItype __attribute__ ((mode (DI))); -typedef unsigned int UDItype __attribute__ ((mode (DI))); -# else -typedef unsigned char UQItype; -typedef long SItype; -typedef unsigned long int USItype; -# if HAVE_LONG_LONG_INT -typedef long long int DItype; -typedef unsigned long long int UDItype; -# else /* Assume `long' gives us a wide enough type. Needed for hppa2.0w. */ -typedef long int DItype; -typedef unsigned long int UDItype; -# endif -# endif -# define LONGLONG_STANDALONE /* Don't require GMP's longlong.h mdep files */ -# define ASSERT(x) /* FIXME make longlong.h really standalone */ -# define __GMP_DECLSPEC /* FIXME make longlong.h really standalone */ -# define __clz_tab factor_clz_tab /* Rename to avoid glibc collision */ -# ifndef __GMP_GNUC_PREREQ -# define __GMP_GNUC_PREREQ(a,b) 1 -# endif - -/* These stub macros are only used in longlong.h in certain system compiler - combinations, so ensure usage to avoid -Wunused-macros warnings. */ -# if __GMP_GNUC_PREREQ (1,1) && defined __clz_tab -ASSERT (1) -__GMP_DECLSPEC -# endif - -# if _ARCH_PPC -# define HAVE_HOST_CPU_FAMILY_powerpc 1 -# endif -# include "longlong.h" -# ifdef COUNT_LEADING_ZEROS_NEED_CLZ_TAB -const unsigned char factor_clz_tab[129] = -{ - 1,2,3,3,4,4,4,4,5,5,5,5,5,5,5,5,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6, - 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, - 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, - 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, - 9 -}; -# endif - -#else /* not USE_LONGLONG_H */ - -# define W_TYPE_SIZE (8 * sizeof (uintmax_t)) -# define __ll_B ((uintmax_t) 1 << (W_TYPE_SIZE / 2)) -# define __ll_lowpart(t) ((uintmax_t) (t) & (__ll_B - 1)) -# define __ll_highpart(t) ((uintmax_t) (t) >> (W_TYPE_SIZE / 2)) - -#endif - -#if !defined __clz_tab && !defined UHWtype -/* Without this seemingly useless conditional, gcc -Wunused-macros - warns that each of the two tested macros is unused on Fedora 18. - FIXME: this is just an ugly band-aid. Fix it properly. */ -#endif - -/* 2*3*5*7*11...*101 is 128 bits, and has 26 prime factors */ -#define MAX_NFACTS 26 - enum { DEV_DEBUG_OPTION = CHAR_MAX + 1 @@ -226,14 +92,6 @@ static struct option const long_options[] = {NULL, 0, NULL, 0} }; -struct factors -{ - uintmax_t plarge[2]; /* Can have a single large factor */ - uintmax_t p[MAX_NFACTS]; - unsigned char e[MAX_NFACTS]; - unsigned char nfactors; -}; - struct mp_factors { mpz_t *p; @@ -241,321 +99,6 @@ struct mp_factors unsigned long int nfactors; }; -static void factor (uintmax_t, uintmax_t, struct factors *); - -#ifndef umul_ppmm -# define umul_ppmm(w1, w0, u, v) \ - do { \ - uintmax_t __x0, __x1, __x2, __x3; \ - unsigned long int __ul, __vl, __uh, __vh; \ - uintmax_t __u = (u), __v = (v); \ - \ - __ul = __ll_lowpart (__u); \ - __uh = __ll_highpart (__u); \ - __vl = __ll_lowpart (__v); \ - __vh = __ll_highpart (__v); \ - \ - __x0 = (uintmax_t) __ul * __vl; \ - __x1 = (uintmax_t) __ul * __vh; \ - __x2 = (uintmax_t) __uh * __vl; \ - __x3 = (uintmax_t) __uh * __vh; \ - \ - __x1 += __ll_highpart (__x0);/* this can't give carry */ \ - __x1 += __x2; /* but this indeed can */ \ - if (__x1 < __x2) /* did we get it? */ \ - __x3 += __ll_B; /* yes, add it in the proper pos. */ \ - \ - (w1) = __x3 + __ll_highpart (__x1); \ - (w0) = (__x1 << W_TYPE_SIZE / 2) + __ll_lowpart (__x0); \ - } while (0) -#endif - -#if !defined udiv_qrnnd || defined UDIV_NEEDS_NORMALIZATION -/* Define our own, not needing normalization. This function is - currently not performance critical, so keep it simple. Similar to - the mod macro below. */ -# undef udiv_qrnnd -# define udiv_qrnnd(q, r, n1, n0, d) \ - do { \ - uintmax_t __d1, __d0, __q, __r1, __r0; \ - \ - assert ((n1) < (d)); \ - __d1 = (d); __d0 = 0; \ - __r1 = (n1); __r0 = (n0); \ - __q = 0; \ - for (unsigned int __i = W_TYPE_SIZE; __i > 0; __i--) \ - { \ - rsh2 (__d1, __d0, __d1, __d0, 1); \ - __q <<= 1; \ - if (ge2 (__r1, __r0, __d1, __d0)) \ - { \ - __q++; \ - sub_ddmmss (__r1, __r0, __r1, __r0, __d1, __d0); \ - } \ - } \ - (r) = __r0; \ - (q) = __q; \ - } while (0) -#endif - -#if !defined add_ssaaaa -# define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - do { \ - uintmax_t _add_x; \ - _add_x = (al) + (bl); \ - (sh) = (ah) + (bh) + (_add_x < (al)); \ - (sl) = _add_x; \ - } while (0) -#endif - -#define rsh2(rh, rl, ah, al, cnt) \ - do { \ - (rl) = ((ah) << (W_TYPE_SIZE - (cnt))) | ((al) >> (cnt)); \ - (rh) = (ah) >> (cnt); \ - } while (0) - -#define lsh2(rh, rl, ah, al, cnt) \ - do { \ - (rh) = ((ah) << cnt) | ((al) >> (W_TYPE_SIZE - (cnt))); \ - (rl) = (al) << (cnt); \ - } while (0) - -#define ge2(ah, al, bh, bl) \ - ((ah) > (bh) || ((ah) == (bh) && (al) >= (bl))) - -#define gt2(ah, al, bh, bl) \ - ((ah) > (bh) || ((ah) == (bh) && (al) > (bl))) - -#ifndef sub_ddmmss -# define sub_ddmmss(rh, rl, ah, al, bh, bl) \ - do { \ - uintmax_t _cy; \ - _cy = (al) < (bl); \ - (rl) = (al) - (bl); \ - (rh) = (ah) - (bh) - _cy; \ - } while (0) -#endif - -#ifndef count_leading_zeros -# define count_leading_zeros(count, x) do { \ - uintmax_t __clz_x = (x); \ - unsigned int __clz_c; \ - for (__clz_c = 0; \ - (__clz_x & ((uintmax_t) 0xff << (W_TYPE_SIZE - 8))) == 0; \ - __clz_c += 8) \ - __clz_x <<= 8; \ - for (; (intmax_t)__clz_x >= 0; __clz_c++) \ - __clz_x <<= 1; \ - (count) = __clz_c; \ - } while (0) -#endif - -#ifndef count_trailing_zeros -# define count_trailing_zeros(count, x) do { \ - uintmax_t __ctz_x = (x); \ - unsigned int __ctz_c = 0; \ - while ((__ctz_x & 1) == 0) \ - { \ - __ctz_x >>= 1; \ - __ctz_c++; \ - } \ - (count) = __ctz_c; \ - } while (0) -#endif - -/* Requires that a < n and b <= n */ -#define submod(r,a,b,n) \ - do { \ - uintmax_t _t = - (uintmax_t) (a < b); \ - (r) = ((n) & _t) + (a) - (b); \ - } while (0) - -#define addmod(r,a,b,n) \ - submod ((r), (a), ((n) - (b)), (n)) - -/* Modular two-word addition and subtraction. For performance reasons, the - most significant bit of n1 must be clear. The destination variables must be - distinct from the mod operand. */ -#define addmod2(r1, r0, a1, a0, b1, b0, n1, n0) \ - do { \ - add_ssaaaa ((r1), (r0), (a1), (a0), (b1), (b0)); \ - if (ge2 ((r1), (r0), (n1), (n0))) \ - sub_ddmmss ((r1), (r0), (r1), (r0), (n1), (n0)); \ - } while (0) -#define submod2(r1, r0, a1, a0, b1, b0, n1, n0) \ - do { \ - sub_ddmmss ((r1), (r0), (a1), (a0), (b1), (b0)); \ - if ((intmax_t) (r1) < 0) \ - add_ssaaaa ((r1), (r0), (r1), (r0), (n1), (n0)); \ - } while (0) - -#define HIGHBIT_TO_MASK(x) \ - (((intmax_t)-1 >> 1) < 0 \ - ? (uintmax_t)((intmax_t)(x) >> (W_TYPE_SIZE - 1)) \ - : ((x) & ((uintmax_t) 1 << (W_TYPE_SIZE - 1)) \ - ? UINTMAX_MAX : (uintmax_t) 0)) - -/* Compute r = a mod d, where r = <*t1,retval>, a = , d = . - Requires that d1 != 0. */ -static uintmax_t -mod2 (uintmax_t *r1, uintmax_t a1, uintmax_t a0, uintmax_t d1, uintmax_t d0) -{ - int cntd, cnta; - - assert (d1 != 0); - - if (a1 == 0) - { - *r1 = 0; - return a0; - } - - count_leading_zeros (cntd, d1); - count_leading_zeros (cnta, a1); - int cnt = cntd - cnta; - lsh2 (d1, d0, d1, d0, cnt); - for (int i = 0; i < cnt; i++) - { - if (ge2 (a1, a0, d1, d0)) - sub_ddmmss (a1, a0, a1, a0, d1, d0); - rsh2 (d1, d0, d1, d0, 1); - } - - *r1 = a1; - return a0; -} - -static uintmax_t _GL_ATTRIBUTE_CONST -gcd_odd (uintmax_t a, uintmax_t b) -{ - if ( (b & 1) == 0) - { - uintmax_t t = b; - b = a; - a = t; - } - if (a == 0) - return b; - - /* Take out least significant one bit, to make room for sign */ - b >>= 1; - - for (;;) - { - uintmax_t t; - uintmax_t bgta; - - while ((a & 1) == 0) - a >>= 1; - a >>= 1; - - t = a - b; - if (t == 0) - return (a << 1) + 1; - - bgta = HIGHBIT_TO_MASK (t); - - /* b <-- min (a, b) */ - b += (bgta & t); - - /* a <-- |a - b| */ - a = (t ^ bgta) - bgta; - } -} - -static uintmax_t -gcd2_odd (uintmax_t *r1, uintmax_t a1, uintmax_t a0, uintmax_t b1, uintmax_t b0) -{ - assert (b0 & 1); - - if ( (a0 | a1) == 0) - { - *r1 = b1; - return b0; - } - - while ((a0 & 1) == 0) - rsh2 (a1, a0, a1, a0, 1); - - for (;;) - { - if ((b1 | a1) == 0) - { - *r1 = 0; - return gcd_odd (b0, a0); - } - - if (gt2 (a1, a0, b1, b0)) - { - sub_ddmmss (a1, a0, a1, a0, b1, b0); - do - rsh2 (a1, a0, a1, a0, 1); - while ((a0 & 1) == 0); - } - else if (gt2 (b1, b0, a1, a0)) - { - sub_ddmmss (b1, b0, b1, b0, a1, a0); - do - rsh2 (b1, b0, b1, b0, 1); - while ((b0 & 1) == 0); - } - else - break; - } - - *r1 = a1; - return a0; -} - -static void -factor_insert_multiplicity (struct factors *factors, - uintmax_t prime, unsigned int m) -{ - unsigned int nfactors = factors->nfactors; - uintmax_t *p = factors->p; - unsigned char *e = factors->e; - - /* Locate position for insert new or increment e. */ - int i; - for (i = nfactors - 1; i >= 0; i--) - { - if (p[i] <= prime) - break; - } - - if (i < 0 || p[i] != prime) - { - for (int j = nfactors - 1; j > i; j--) - { - p[j + 1] = p[j]; - e[j + 1] = e[j]; - } - p[i + 1] = prime; - e[i + 1] = m; - factors->nfactors = nfactors + 1; - } - else - { - e[i] += m; - } -} - -#define factor_insert(f, p) factor_insert_multiplicity (f, p, 1) - -static void -factor_insert_large (struct factors *factors, - uintmax_t p1, uintmax_t p0) -{ - if (p1 > 0) - { - assert (factors->plarge[1] == 0); - factors->plarge[0] = p0; - factors->plarge[1] = p1; - } - else - factor_insert (factors, p0); -} - #ifndef mpz_inits # include @@ -664,31 +207,11 @@ static const unsigned char primes_diff[] = { #define PRIMES_PTAB_ENTRIES \ (sizeof (primes_diff) / sizeof (primes_diff[0]) - 8 + 1) -#define P(a,b,c,d) b, -static const unsigned char primes_diff8[] = { -#include "primes.h" -0,0,0,0,0,0,0 /* 7 sentinels for 8-way loop */ -}; -#undef P - -struct primes_dtab -{ - uintmax_t binv, lim; -}; - -#define P(a,b,c,d) {c,d}, -static const struct primes_dtab primes_dtab[] = { -#include "primes.h" -{1,0},{1,0},{1,0},{1,0},{1,0},{1,0},{1,0} /* 7 sentinels for 8-way loop */ -}; -#undef P - /* Verify that uintmax_t is not wider than the integers used to generate primes.h. */ verify (W <= WIDE_UINT_BITS); -/* debugging for developers. Enables devmsg(). - This flag is used only in the GMP code. */ +/* debugging for developers. Enables devmsg(). */ static bool dev_debug = false; /* Prove primality or run probabilistic tests. */ @@ -697,15 +220,6 @@ static bool flag_prove_primality = PROVE_PRIMALITY; /* Number of Miller-Rabin tests to run when not proving primality. */ #define MR_REPS 25 -static void -factor_insert_refind (struct factors *factors, uintmax_t p, unsigned int i, - unsigned int off) -{ - for (unsigned int j = 0; j < off; j++) - p += primes_diff[i + j]; - factor_insert (factors, p); -} - /* Trial division with odd primes uses the following trick. Let p be an odd prime, and B = 2^{W_TYPE_SIZE}. For simplicity, @@ -739,87 +253,6 @@ factor_insert_refind (struct factors *factors, uintmax_t p, unsigned int i, order, and the non-multiples of p onto the range lim < q < B. */ -static uintmax_t -factor_using_division (uintmax_t *t1p, uintmax_t t1, uintmax_t t0, - struct factors *factors) -{ - if (t0 % 2 == 0) - { - unsigned int cnt; - - if (t0 == 0) - { - count_trailing_zeros (cnt, t1); - t0 = t1 >> cnt; - t1 = 0; - cnt += W_TYPE_SIZE; - } - else - { - count_trailing_zeros (cnt, t0); - rsh2 (t1, t0, t1, t0, cnt); - } - - factor_insert_multiplicity (factors, 2, cnt); - } - - uintmax_t p = 3; - unsigned int i; - for (i = 0; t1 > 0 && i < PRIMES_PTAB_ENTRIES; i++) - { - for (;;) - { - uintmax_t q1, q0, hi, lo _GL_UNUSED; - - q0 = t0 * primes_dtab[i].binv; - umul_ppmm (hi, lo, q0, p); - if (hi > t1) - break; - hi = t1 - hi; - q1 = hi * primes_dtab[i].binv; - if (LIKELY (q1 > primes_dtab[i].lim)) - break; - t1 = q1; t0 = q0; - factor_insert (factors, p); - } - p += primes_diff[i + 1]; - } - if (t1p) - *t1p = t1; - -#define DIVBLOCK(I) \ - do { \ - for (;;) \ - { \ - q = t0 * pd[I].binv; \ - if (LIKELY (q > pd[I].lim)) \ - break; \ - t0 = q; \ - factor_insert_refind (factors, p, i + 1, I); \ - } \ - } while (0) - - for (; i < PRIMES_PTAB_ENTRIES; i += 8) - { - uintmax_t q; - const struct primes_dtab *pd = &primes_dtab[i]; - DIVBLOCK (0); - DIVBLOCK (1); - DIVBLOCK (2); - DIVBLOCK (3); - DIVBLOCK (4); - DIVBLOCK (5); - DIVBLOCK (6); - DIVBLOCK (7); - - p += primes_diff8[i]; - if (p * p > t0) - break; - } - - return t0; -} - static void mp_factor_using_division (mpz_t t, struct mp_factors *factors) { @@ -857,303 +290,6 @@ mp_factor_using_division (mpz_t t, struct mp_factors *factors) mpz_clear (q); } -/* Entry i contains (2i+1)^(-1) mod 2^8. */ -static const unsigned char binvert_table[128] = -{ - 0x01, 0xAB, 0xCD, 0xB7, 0x39, 0xA3, 0xC5, 0xEF, - 0xF1, 0x1B, 0x3D, 0xA7, 0x29, 0x13, 0x35, 0xDF, - 0xE1, 0x8B, 0xAD, 0x97, 0x19, 0x83, 0xA5, 0xCF, - 0xD1, 0xFB, 0x1D, 0x87, 0x09, 0xF3, 0x15, 0xBF, - 0xC1, 0x6B, 0x8D, 0x77, 0xF9, 0x63, 0x85, 0xAF, - 0xB1, 0xDB, 0xFD, 0x67, 0xE9, 0xD3, 0xF5, 0x9F, - 0xA1, 0x4B, 0x6D, 0x57, 0xD9, 0x43, 0x65, 0x8F, - 0x91, 0xBB, 0xDD, 0x47, 0xC9, 0xB3, 0xD5, 0x7F, - 0x81, 0x2B, 0x4D, 0x37, 0xB9, 0x23, 0x45, 0x6F, - 0x71, 0x9B, 0xBD, 0x27, 0xA9, 0x93, 0xB5, 0x5F, - 0x61, 0x0B, 0x2D, 0x17, 0x99, 0x03, 0x25, 0x4F, - 0x51, 0x7B, 0x9D, 0x07, 0x89, 0x73, 0x95, 0x3F, - 0x41, 0xEB, 0x0D, 0xF7, 0x79, 0xE3, 0x05, 0x2F, - 0x31, 0x5B, 0x7D, 0xE7, 0x69, 0x53, 0x75, 0x1F, - 0x21, 0xCB, 0xED, 0xD7, 0x59, 0xC3, 0xE5, 0x0F, - 0x11, 0x3B, 0x5D, 0xC7, 0x49, 0x33, 0x55, 0xFF -}; - -/* Compute n^(-1) mod B, using a Newton iteration. */ -#define binv(inv,n) \ - do { \ - uintmax_t __n = (n); \ - uintmax_t __inv; \ - \ - __inv = binvert_table[(__n / 2) & 0x7F]; /* 8 */ \ - if (W_TYPE_SIZE > 8) __inv = 2 * __inv - __inv * __inv * __n; \ - if (W_TYPE_SIZE > 16) __inv = 2 * __inv - __inv * __inv * __n; \ - if (W_TYPE_SIZE > 32) __inv = 2 * __inv - __inv * __inv * __n; \ - \ - if (W_TYPE_SIZE > 64) \ - { \ - int __invbits = 64; \ - do { \ - __inv = 2 * __inv - __inv * __inv * __n; \ - __invbits *= 2; \ - } while (__invbits < W_TYPE_SIZE); \ - } \ - \ - (inv) = __inv; \ - } while (0) - -/* q = u / d, assuming d|u. */ -#define divexact_21(q1, q0, u1, u0, d) \ - do { \ - uintmax_t _di, _q0; \ - binv (_di, (d)); \ - _q0 = (u0) * _di; \ - if ((u1) >= (d)) \ - { \ - uintmax_t _p1, _p0 _GL_UNUSED; \ - umul_ppmm (_p1, _p0, _q0, d); \ - (q1) = ((u1) - _p1) * _di; \ - (q0) = _q0; \ - } \ - else \ - { \ - (q0) = _q0; \ - (q1) = 0; \ - } \ - } while (0) - -/* x B (mod n). */ -#define redcify(r_prim, r, n) \ - do { \ - uintmax_t _redcify_q _GL_UNUSED; \ - udiv_qrnnd (_redcify_q, r_prim, r, 0, n); \ - } while (0) - -/* x B^2 (mod n). Requires x > 0, n1 < B/2 */ -#define redcify2(r1, r0, x, n1, n0) \ - do { \ - uintmax_t _r1, _r0, _i; \ - if ((x) < (n1)) \ - { \ - _r1 = (x); _r0 = 0; \ - _i = W_TYPE_SIZE; \ - } \ - else \ - { \ - _r1 = 0; _r0 = (x); \ - _i = 2*W_TYPE_SIZE; \ - } \ - while (_i-- > 0) \ - { \ - lsh2 (_r1, _r0, _r1, _r0, 1); \ - if (ge2 (_r1, _r0, (n1), (n0))) \ - sub_ddmmss (_r1, _r0, _r1, _r0, (n1), (n0)); \ - } \ - (r1) = _r1; \ - (r0) = _r0; \ - } while (0) - -/* Modular two-word multiplication, r = a * b mod m, with mi = m^(-1) mod B. - Both a and b must be in redc form, the result will be in redc form too. */ -static inline uintmax_t -mulredc (uintmax_t a, uintmax_t b, uintmax_t m, uintmax_t mi) -{ - uintmax_t rh, rl, q, th, tl _GL_UNUSED, xh; - - umul_ppmm (rh, rl, a, b); - q = rl * mi; - umul_ppmm (th, tl, q, m); - xh = rh - th; - if (rh < th) - xh += m; - - return xh; -} - -/* Modular two-word multiplication, r = a * b mod m, with mi = m^(-1) mod B. - Both a and b must be in redc form, the result will be in redc form too. - For performance reasons, the most significant bit of m must be clear. */ -static uintmax_t -mulredc2 (uintmax_t *r1p, - uintmax_t a1, uintmax_t a0, uintmax_t b1, uintmax_t b0, - uintmax_t m1, uintmax_t m0, uintmax_t mi) -{ - uintmax_t r1, r0, q, p1, p0 _GL_UNUSED, t1, t0, s1, s0; - mi = -mi; - assert ( (a1 >> (W_TYPE_SIZE - 1)) == 0); - assert ( (b1 >> (W_TYPE_SIZE - 1)) == 0); - assert ( (m1 >> (W_TYPE_SIZE - 1)) == 0); - - /* First compute a0 * B^{-1} - +-----+ - |a0 b0| - +--+--+--+ - |a0 b1| - +--+--+--+ - |q0 m0| - +--+--+--+ - |q0 m1| - -+--+--+--+ - |r1|r0| 0| - +--+--+--+ - */ - umul_ppmm (t1, t0, a0, b0); - umul_ppmm (r1, r0, a0, b1); - q = mi * t0; - umul_ppmm (p1, p0, q, m0); - umul_ppmm (s1, s0, q, m1); - r0 += (t0 != 0); /* Carry */ - add_ssaaaa (r1, r0, r1, r0, 0, p1); - add_ssaaaa (r1, r0, r1, r0, 0, t1); - add_ssaaaa (r1, r0, r1, r0, s1, s0); - - /* Next, (a1 * + B^{-1} - +-----+ - |a1 b0| - +--+--+ - |r1|r0| - +--+--+--+ - |a1 b1| - +--+--+--+ - |q1 m0| - +--+--+--+ - |q1 m1| - -+--+--+--+ - |r1|r0| 0| - +--+--+--+ - */ - umul_ppmm (t1, t0, a1, b0); - umul_ppmm (s1, s0, a1, b1); - add_ssaaaa (t1, t0, t1, t0, 0, r0); - q = mi * t0; - add_ssaaaa (r1, r0, s1, s0, 0, r1); - umul_ppmm (p1, p0, q, m0); - umul_ppmm (s1, s0, q, m1); - r0 += (t0 != 0); /* Carry */ - add_ssaaaa (r1, r0, r1, r0, 0, p1); - add_ssaaaa (r1, r0, r1, r0, 0, t1); - add_ssaaaa (r1, r0, r1, r0, s1, s0); - - if (ge2 (r1, r0, m1, m0)) - sub_ddmmss (r1, r0, r1, r0, m1, m0); - - *r1p = r1; - return r0; -} - -static uintmax_t _GL_ATTRIBUTE_CONST -powm (uintmax_t b, uintmax_t e, uintmax_t n, uintmax_t ni, uintmax_t one) -{ - uintmax_t y = one; - - if (e & 1) - y = b; - - while (e != 0) - { - b = mulredc (b, b, n, ni); - e >>= 1; - - if (e & 1) - y = mulredc (y, b, n, ni); - } - - return y; -} - -static uintmax_t -powm2 (uintmax_t *r1m, - const uintmax_t *bp, const uintmax_t *ep, const uintmax_t *np, - uintmax_t ni, const uintmax_t *one) -{ - uintmax_t r1, r0, b1, b0, n1, n0; - unsigned int i; - uintmax_t e; - - b0 = bp[0]; - b1 = bp[1]; - n0 = np[0]; - n1 = np[1]; - - r0 = one[0]; - r1 = one[1]; - - for (e = ep[0], i = W_TYPE_SIZE; i > 0; i--, e >>= 1) - { - if (e & 1) - { - r0 = mulredc2 (r1m, r1, r0, b1, b0, n1, n0, ni); - r1 = *r1m; - } - b0 = mulredc2 (r1m, b1, b0, b1, b0, n1, n0, ni); - b1 = *r1m; - } - for (e = ep[1]; e > 0; e >>= 1) - { - if (e & 1) - { - r0 = mulredc2 (r1m, r1, r0, b1, b0, n1, n0, ni); - r1 = *r1m; - } - b0 = mulredc2 (r1m, b1, b0, b1, b0, n1, n0, ni); - b1 = *r1m; - } - *r1m = r1; - return r0; -} - -static bool _GL_ATTRIBUTE_CONST -millerrabin (uintmax_t n, uintmax_t ni, uintmax_t b, uintmax_t q, - unsigned int k, uintmax_t one) -{ - uintmax_t y = powm (b, q, n, ni, one); - - uintmax_t nm1 = n - one; /* -1, but in redc representation. */ - - if (y == one || y == nm1) - return true; - - for (unsigned int i = 1; i < k; i++) - { - y = mulredc (y, y, n, ni); - - if (y == nm1) - return true; - if (y == one) - return false; - } - return false; -} - -static bool -millerrabin2 (const uintmax_t *np, uintmax_t ni, const uintmax_t *bp, - const uintmax_t *qp, unsigned int k, const uintmax_t *one) -{ - uintmax_t y1, y0, nm1_1, nm1_0, r1m; - - y0 = powm2 (&r1m, bp, qp, np, ni, one); - y1 = r1m; - - if (y0 == one[0] && y1 == one[1]) - return true; - - sub_ddmmss (nm1_1, nm1_0, np[1], np[0], one[1], one[0]); - - if (y0 == nm1_0 && y1 == nm1_1) - return true; - - for (unsigned int i = 1; i < k; i++) - { - y0 = mulredc2 (&r1m, y1, y0, y1, y0, np[1], np[0], ni); - y1 = r1m; - - if (y0 == nm1_0 && y1 == nm1_1) - return true; - if (y0 == one[0] && y1 == one[1]) - return false; - } - return false; -} - static bool mp_millerrabin (mpz_srcptr n, mpz_srcptr nm1, mpz_ptr x, mpz_ptr y, mpz_srcptr q, unsigned long int k) @@ -1174,41 +310,43 @@ mp_millerrabin (mpz_srcptr n, mpz_srcptr nm1, mpz_ptr x, mpz_ptr y, return false; } -/* Lucas' prime test. The number of iterations vary greatly, up to a few dozen - have been observed. The average seem to be about 2. */ static bool -prime_p (uintmax_t n) +mp_prime_p (mpz_t n) { - int k; bool is_prime; - uintmax_t a_prim, one, ni; - struct factors factors; + mpz_t q, a, nm1, tmp; + struct mp_factors factors; - if (n <= 1) + if (mpz_cmp_ui (n, 1) <= 0) return false; /* We have already casted out small primes. */ - if (n < (uintmax_t) FIRST_OMITTED_PRIME * FIRST_OMITTED_PRIME) + if (mpz_cmp_ui (n, (long) FIRST_OMITTED_PRIME * FIRST_OMITTED_PRIME) < 0) return true; + mpz_inits (q, a, nm1, tmp, NULL); + /* Precomputation for Miller-Rabin. */ - uintmax_t q = n - 1; - for (k = 0; (q & 1) == 0; k++) - q >>= 1; + mpz_sub_ui (nm1, n, 1); - uintmax_t a = 2; - binv (ni, n); /* ni <- 1/n mod B */ - redcify (one, 1, n); - addmod (a_prim, one, one, n); /* i.e., redcify a = 2 */ + /* Find q and k, where q is odd and n = 1 + 2**k * q. */ + unsigned long int k = mpz_scan1 (nm1, 0); + mpz_tdiv_q_2exp (q, nm1, k); + + mpz_set_ui (a, 2); /* Perform a Miller-Rabin test, finds most composites quickly. */ - if (!millerrabin (n, ni, a_prim, q, k, one)) - return false; + if (!mp_millerrabin (n, nm1, a, tmp, q, k)) + { + is_prime = false; + goto ret2; + } if (flag_prove_primality) { /* Factor n-1 for Lucas. */ - factor (0, n - 1, &factors); + mpz_set (tmp, nm1); + mp_factor (tmp, &factors); } /* Loop until Lucas proves our number prime, or Miller-Rabin proves our @@ -1218,10 +356,11 @@ prime_p (uintmax_t n) if (flag_prove_primality) { is_prime = true; - for (unsigned int i = 0; i < factors.nfactors && is_prime; i++) + for (unsigned long int i = 0; i < factors.nfactors && is_prime; i++) { - is_prime - = powm (a_prim, (n - 1) / factors.p[i], n, ni, one) != one; + mpz_divexact (tmp, nm1, factors.p[i]); + mpz_powm (tmp, a, tmp, n); + is_prime = mpz_cmp_ui (tmp, 1) != 0; } } else @@ -1231,202 +370,15 @@ prime_p (uintmax_t n) } if (is_prime) - return true; + goto ret1; - a += primes_diff[r]; /* Establish new base. */ - - /* The following is equivalent to redcify (a_prim, a, n). It runs faster - on most processors, since it avoids udiv_qrnnd. If we go down the - udiv_qrnnd_preinv path, this code should be replaced. */ - { - uintmax_t s1, s0; - umul_ppmm (s1, s0, one, a); - if (LIKELY (s1 == 0)) - a_prim = s0 % n; - else - { - uintmax_t dummy _GL_UNUSED; - udiv_qrnnd (dummy, a_prim, s1, s0, n); - } - } - - if (!millerrabin (n, ni, a_prim, q, k, one)) - return false; - } - - error (0, 0, _("Lucas prime test failure. This should not happen")); - abort (); -} - -static bool -prime2_p (uintmax_t n1, uintmax_t n0) -{ - uintmax_t q[2], nm1[2]; - uintmax_t a_prim[2]; - uintmax_t one[2]; - uintmax_t na[2]; - uintmax_t ni; - unsigned int k; - struct factors factors; - - if (n1 == 0) - return prime_p (n0); - - nm1[1] = n1 - (n0 == 0); - nm1[0] = n0 - 1; - if (nm1[0] == 0) - { - count_trailing_zeros (k, nm1[1]); - - q[0] = nm1[1] >> k; - q[1] = 0; - k += W_TYPE_SIZE; - } - else - { - count_trailing_zeros (k, nm1[0]); - rsh2 (q[1], q[0], nm1[1], nm1[0], k); - } - - uintmax_t a = 2; - binv (ni, n0); - redcify2 (one[1], one[0], 1, n1, n0); - addmod2 (a_prim[1], a_prim[0], one[1], one[0], one[1], one[0], n1, n0); - - /* FIXME: Use scalars or pointers in arguments? Some consistency needed. */ - na[0] = n0; - na[1] = n1; - - if (!millerrabin2 (na, ni, a_prim, q, k, one)) - return false; - - if (flag_prove_primality) - { - /* Factor n-1 for Lucas. */ - factor (nm1[1], nm1[0], &factors); - } - - /* Loop until Lucas proves our number prime, or Miller-Rabin proves our - number composite. */ - for (unsigned int r = 0; r < PRIMES_PTAB_ENTRIES; r++) - { - bool is_prime; - uintmax_t e[2], y[2]; - - if (flag_prove_primality) - { - is_prime = true; - if (factors.plarge[1]) - { - uintmax_t pi; - binv (pi, factors.plarge[0]); - e[0] = pi * nm1[0]; - e[1] = 0; - y[0] = powm2 (&y[1], a_prim, e, na, ni, one); - is_prime = (y[0] != one[0] || y[1] != one[1]); - } - for (unsigned int i = 0; i < factors.nfactors && is_prime; i++) - { - /* FIXME: We always have the factor 2. Do we really need to - handle it here? We have done the same powering as part - of millerrabin. */ - if (factors.p[i] == 2) - rsh2 (e[1], e[0], nm1[1], nm1[0], 1); - else - divexact_21 (e[1], e[0], nm1[1], nm1[0], factors.p[i]); - y[0] = powm2 (&y[1], a_prim, e, na, ni, one); - is_prime = (y[0] != one[0] || y[1] != one[1]); - } - } - else - { - /* After enough Miller-Rabin runs, be content. */ - is_prime = (r == MR_REPS - 1); - } - - if (is_prime) - return true; - - a += primes_diff[r]; /* Establish new base. */ - redcify2 (a_prim[1], a_prim[0], a, n1, n0); - - if (!millerrabin2 (na, ni, a_prim, q, k, one)) - return false; - } - - error (0, 0, _("Lucas prime test failure. This should not happen")); - abort (); -} - -static bool -mp_prime_p (mpz_t n) -{ - bool is_prime; - mpz_t q, a, nm1, tmp; - struct mp_factors factors; - - if (mpz_cmp_ui (n, 1) <= 0) - return false; - - /* We have already casted out small primes. */ - if (mpz_cmp_ui (n, (long) FIRST_OMITTED_PRIME * FIRST_OMITTED_PRIME) < 0) - return true; - - mpz_inits (q, a, nm1, tmp, NULL); - - /* Precomputation for Miller-Rabin. */ - mpz_sub_ui (nm1, n, 1); - - /* Find q and k, where q is odd and n = 1 + 2**k * q. */ - unsigned long int k = mpz_scan1 (nm1, 0); - mpz_tdiv_q_2exp (q, nm1, k); - - mpz_set_ui (a, 2); - - /* Perform a Miller-Rabin test, finds most composites quickly. */ - if (!mp_millerrabin (n, nm1, a, tmp, q, k)) - { - is_prime = false; - goto ret2; - } - - if (flag_prove_primality) - { - /* Factor n-1 for Lucas. */ - mpz_set (tmp, nm1); - mp_factor (tmp, &factors); - } - - /* Loop until Lucas proves our number prime, or Miller-Rabin proves our - number composite. */ - for (unsigned int r = 0; r < PRIMES_PTAB_ENTRIES; r++) - { - if (flag_prove_primality) - { - is_prime = true; - for (unsigned long int i = 0; i < factors.nfactors && is_prime; i++) - { - mpz_divexact (tmp, nm1, factors.p[i]); - mpz_powm (tmp, a, tmp, n); - is_prime = mpz_cmp_ui (tmp, 1) != 0; - } - } - else - { - /* After enough Miller-Rabin runs, be content. */ - is_prime = (r == MR_REPS - 1); - } - - if (is_prime) - goto ret1; - - mpz_add_ui (a, a, primes_diff[r]); /* Establish new base. */ - - if (!mp_millerrabin (n, nm1, a, tmp, q, k)) - { - is_prime = false; - goto ret1; - } + mpz_add_ui (a, a, primes_diff[r]); /* Establish new base. */ + + if (!mp_millerrabin (n, nm1, a, tmp, q, k)) + { + is_prime = false; + goto ret1; + } } error (0, 0, _("Lucas prime test failure. This should not happen")); @@ -1441,213 +393,6 @@ mp_prime_p (mpz_t n) return is_prime; } -static void -factor_using_pollard_rho (uintmax_t n, unsigned long int a, - struct factors *factors) -{ - uintmax_t x, z, y, P, t, ni, g; - - unsigned long int k = 1; - unsigned long int l = 1; - - redcify (P, 1, n); - addmod (x, P, P, n); /* i.e., redcify(2) */ - y = z = x; - - while (n != 1) - { - assert (a < n); - - binv (ni, n); /* FIXME: when could we use old 'ni' value? */ - - for (;;) - { - do - { - x = mulredc (x, x, n, ni); - addmod (x, x, a, n); - - submod (t, z, x, n); - P = mulredc (P, t, n, ni); - - if (k % 32 == 1) - { - if (gcd_odd (P, n) != 1) - goto factor_found; - y = x; - } - } - while (--k != 0); - - z = x; - k = l; - l = 2 * l; - for (unsigned long int i = 0; i < k; i++) - { - x = mulredc (x, x, n, ni); - addmod (x, x, a, n); - } - y = x; - } - - factor_found: - do - { - y = mulredc (y, y, n, ni); - addmod (y, y, a, n); - - submod (t, z, y, n); - g = gcd_odd (t, n); - } - while (g == 1); - - if (n == g) - { - /* Found n itself as factor. Restart with different params. */ - factor_using_pollard_rho (n, a + 1, factors); - return; - } - - n = n / g; - - if (!prime_p (g)) - factor_using_pollard_rho (g, a + 1, factors); - else - factor_insert (factors, g); - - if (prime_p (n)) - { - factor_insert (factors, n); - break; - } - - x = x % n; - z = z % n; - y = y % n; - } -} - -static void -factor_using_pollard_rho2 (uintmax_t n1, uintmax_t n0, unsigned long int a, - struct factors *factors) -{ - uintmax_t x1, x0, z1, z0, y1, y0, P1, P0, t1, t0, ni, g1, g0, r1m; - - unsigned long int k = 1; - unsigned long int l = 1; - - redcify2 (P1, P0, 1, n1, n0); - addmod2 (x1, x0, P1, P0, P1, P0, n1, n0); /* i.e., redcify(2) */ - y1 = z1 = x1; - y0 = z0 = x0; - - while (n1 != 0 || n0 != 1) - { - binv (ni, n0); - - for (;;) - { - do - { - x0 = mulredc2 (&r1m, x1, x0, x1, x0, n1, n0, ni); - x1 = r1m; - addmod2 (x1, x0, x1, x0, 0, (uintmax_t) a, n1, n0); - - submod2 (t1, t0, z1, z0, x1, x0, n1, n0); - P0 = mulredc2 (&r1m, P1, P0, t1, t0, n1, n0, ni); - P1 = r1m; - - if (k % 32 == 1) - { - g0 = gcd2_odd (&g1, P1, P0, n1, n0); - if (g1 != 0 || g0 != 1) - goto factor_found; - y1 = x1; y0 = x0; - } - } - while (--k != 0); - - z1 = x1; z0 = x0; - k = l; - l = 2 * l; - for (unsigned long int i = 0; i < k; i++) - { - x0 = mulredc2 (&r1m, x1, x0, x1, x0, n1, n0, ni); - x1 = r1m; - addmod2 (x1, x0, x1, x0, 0, (uintmax_t) a, n1, n0); - } - y1 = x1; y0 = x0; - } - - factor_found: - do - { - y0 = mulredc2 (&r1m, y1, y0, y1, y0, n1, n0, ni); - y1 = r1m; - addmod2 (y1, y0, y1, y0, 0, (uintmax_t) a, n1, n0); - - submod2 (t1, t0, z1, z0, y1, y0, n1, n0); - g0 = gcd2_odd (&g1, t1, t0, n1, n0); - } - while (g1 == 0 && g0 == 1); - - if (g1 == 0) - { - /* The found factor is one word, and > 1. */ - divexact_21 (n1, n0, n1, n0, g0); /* n = n / g */ - - if (!prime_p (g0)) - factor_using_pollard_rho (g0, a + 1, factors); - else - factor_insert (factors, g0); - } - else - { - /* The found factor is two words. This is highly unlikely, thus hard - to trigger. Please be careful before you change this code! */ - uintmax_t ginv; - - if (n1 == g1 && n0 == g0) - { - /* Found n itself as factor. Restart with different params. */ - factor_using_pollard_rho2 (n1, n0, a + 1, factors); - return; - } - - binv (ginv, g0); /* Compute n = n / g. Since the result will */ - n0 = ginv * n0; /* fit one word, we can compute the quotient */ - n1 = 0; /* modulo B, ignoring the high divisor word. */ - - if (!prime2_p (g1, g0)) - factor_using_pollard_rho2 (g1, g0, a + 1, factors); - else - factor_insert_large (factors, g1, g0); - } - - if (n1 == 0) - { - if (prime_p (n0)) - { - factor_insert (factors, n0); - break; - } - - factor_using_pollard_rho (n0, a, factors); - return; - } - - if (prime2_p (n1, n0)) - { - factor_insert_large (factors, n1, n0); - break; - } - - x0 = mod2 (&x1, x1, x0, n1, n0); - z0 = mod2 (&z1, z1, z0, n1, n0); - y0 = mod2 (&y1, y1, y0, n1, n0); - } -} - static void mp_factor_using_pollard_rho (mpz_t n, unsigned long int a, struct mp_factors *factors) @@ -1740,494 +485,6 @@ mp_factor_using_pollard_rho (mpz_t n, unsigned long int a, mpz_clears (P, t2, t, z, x, y, NULL); } -#if USE_SQUFOF -/* FIXME: Maybe better to use an iteration converging to 1/sqrt(n)? If - algorithm is replaced, consider also returning the remainder. */ -static uintmax_t _GL_ATTRIBUTE_CONST -isqrt (uintmax_t n) -{ - uintmax_t x; - unsigned c; - if (n == 0) - return 0; - - count_leading_zeros (c, n); - - /* Make x > sqrt(n). This will be invariant through the loop. */ - x = (uintmax_t) 1 << ((W_TYPE_SIZE + 1 - c) / 2); - - for (;;) - { - uintmax_t y = (x + n/x) / 2; - if (y >= x) - return x; - - x = y; - } -} - -static uintmax_t _GL_ATTRIBUTE_CONST -isqrt2 (uintmax_t nh, uintmax_t nl) -{ - unsigned int shift; - uintmax_t x; - - /* Ensures the remainder fits in an uintmax_t. */ - assert (nh < ((uintmax_t) 1 << (W_TYPE_SIZE - 2))); - - if (nh == 0) - return isqrt (nl); - - count_leading_zeros (shift, nh); - shift &= ~1; - - /* Make x > sqrt(n) */ - x = isqrt ( (nh << shift) + (nl >> (W_TYPE_SIZE - shift))) + 1; - x <<= (W_TYPE_SIZE - shift) / 2; - - /* Do we need more than one iteration? */ - for (;;) - { - uintmax_t r _GL_UNUSED; - uintmax_t q, y; - udiv_qrnnd (q, r, nh, nl, x); - y = (x + q) / 2; - - if (y >= x) - { - uintmax_t hi, lo; - umul_ppmm (hi, lo, x + 1, x + 1); - assert (gt2 (hi, lo, nh, nl)); - - umul_ppmm (hi, lo, x, x); - assert (ge2 (nh, nl, hi, lo)); - sub_ddmmss (hi, lo, nh, nl, hi, lo); - assert (hi == 0); - - return x; - } - - x = y; - } -} - -/* MAGIC[N] has a bit i set iff i is a quadratic residue mod N. */ -# define MAGIC64 0x0202021202030213ULL -# define MAGIC63 0x0402483012450293ULL -# define MAGIC65 0x218a019866014613ULL -# define MAGIC11 0x23b - -/* Return the square root if the input is a square, otherwise 0. */ -static uintmax_t _GL_ATTRIBUTE_CONST -is_square (uintmax_t x) -{ - /* Uses the tests suggested by Cohen. Excludes 99% of the non-squares before - computing the square root. */ - if (((MAGIC64 >> (x & 63)) & 1) - && ((MAGIC63 >> (x % 63)) & 1) - /* Both 0 and 64 are squares mod (65) */ - && ((MAGIC65 >> ((x % 65) & 63)) & 1) - && ((MAGIC11 >> (x % 11) & 1))) - { - uintmax_t r = isqrt (x); - if (r*r == x) - return r; - } - return 0; -} - -/* invtab[i] = floor(0x10000 / (0x100 + i) */ -static const unsigned short invtab[0x81] = - { - 0x200, - 0x1fc, 0x1f8, 0x1f4, 0x1f0, 0x1ec, 0x1e9, 0x1e5, 0x1e1, - 0x1de, 0x1da, 0x1d7, 0x1d4, 0x1d0, 0x1cd, 0x1ca, 0x1c7, - 0x1c3, 0x1c0, 0x1bd, 0x1ba, 0x1b7, 0x1b4, 0x1b2, 0x1af, - 0x1ac, 0x1a9, 0x1a6, 0x1a4, 0x1a1, 0x19e, 0x19c, 0x199, - 0x197, 0x194, 0x192, 0x18f, 0x18d, 0x18a, 0x188, 0x186, - 0x183, 0x181, 0x17f, 0x17d, 0x17a, 0x178, 0x176, 0x174, - 0x172, 0x170, 0x16e, 0x16c, 0x16a, 0x168, 0x166, 0x164, - 0x162, 0x160, 0x15e, 0x15c, 0x15a, 0x158, 0x157, 0x155, - 0x153, 0x151, 0x150, 0x14e, 0x14c, 0x14a, 0x149, 0x147, - 0x146, 0x144, 0x142, 0x141, 0x13f, 0x13e, 0x13c, 0x13b, - 0x139, 0x138, 0x136, 0x135, 0x133, 0x132, 0x130, 0x12f, - 0x12e, 0x12c, 0x12b, 0x129, 0x128, 0x127, 0x125, 0x124, - 0x123, 0x121, 0x120, 0x11f, 0x11e, 0x11c, 0x11b, 0x11a, - 0x119, 0x118, 0x116, 0x115, 0x114, 0x113, 0x112, 0x111, - 0x10f, 0x10e, 0x10d, 0x10c, 0x10b, 0x10a, 0x109, 0x108, - 0x107, 0x106, 0x105, 0x104, 0x103, 0x102, 0x101, 0x100, - }; - -/* Compute q = [u/d], r = u mod d. Avoids slow hardware division for the case - that q < 0x40; here it instead uses a table of (Euclidian) inverses. */ -# define div_smallq(q, r, u, d) \ - do { \ - if ((u) / 0x40 < (d)) \ - { \ - int _cnt; \ - uintmax_t _dinv, _mask, _q, _r; \ - count_leading_zeros (_cnt, (d)); \ - _r = (u); \ - if (UNLIKELY (_cnt > (W_TYPE_SIZE - 8))) \ - { \ - _dinv = invtab[((d) << (_cnt + 8 - W_TYPE_SIZE)) - 0x80]; \ - _q = _dinv * _r >> (8 + W_TYPE_SIZE - _cnt); \ - } \ - else \ - { \ - _dinv = invtab[((d) >> (W_TYPE_SIZE - 8 - _cnt)) - 0x7f]; \ - _q = _dinv * (_r >> (W_TYPE_SIZE - 3 - _cnt)) >> 11; \ - } \ - _r -= _q*(d); \ - \ - _mask = -(uintmax_t) (_r >= (d)); \ - (r) = _r - (_mask & (d)); \ - (q) = _q - _mask; \ - assert ( (q) * (d) + (r) == u); \ - } \ - else \ - { \ - uintmax_t _q = (u) / (d); \ - (r) = (u) - _q * (d); \ - (q) = _q; \ - } \ - } while (0) - -/* Notes: Example N = 22117019. After first phase we find Q1 = 6314, Q - = 3025, P = 1737, representing F_{18} = (-6314, 2* 1737, 3025), - with 3025 = 55^2. - - Constructing the square root, we get Q1 = 55, Q = 8653, P = 4652, - representing G_0 = (-55, 2*4652, 8653). - - In the notation of the paper: - - S_{-1} = 55, S_0 = 8653, R_0 = 4652 - - Put - - t_0 = floor([q_0 + R_0] / S0) = 1 - R_1 = t_0 * S_0 - R_0 = 4001 - S_1 = S_{-1} +t_0 (R_0 - R_1) = 706 -*/ - -/* Multipliers, in order of efficiency: - 0.7268 3*5*7*11 = 1155 = 3 (mod 4) - 0.7317 3*5*7 = 105 = 1 - 0.7820 3*5*11 = 165 = 1 - 0.7872 3*5 = 15 = 3 - 0.8101 3*7*11 = 231 = 3 - 0.8155 3*7 = 21 = 1 - 0.8284 5*7*11 = 385 = 1 - 0.8339 5*7 = 35 = 3 - 0.8716 3*11 = 33 = 1 - 0.8774 3 = 3 = 3 - 0.8913 5*11 = 55 = 3 - 0.8972 5 = 5 = 1 - 0.9233 7*11 = 77 = 1 - 0.9295 7 = 7 = 3 - 0.9934 11 = 11 = 3 -*/ -# define QUEUE_SIZE 50 -#endif - -#if STAT_SQUFOF -# define Q_FREQ_SIZE 50 -/* Element 0 keeps the total */ -static unsigned int q_freq[Q_FREQ_SIZE + 1]; -# define MIN(a,b) ((a) < (b) ? (a) : (b)) -#endif - -#if USE_SQUFOF -/* Return true on success. Expected to fail only for numbers - >= 2^{2*W_TYPE_SIZE - 2}, or close to that limit. */ -static bool -factor_using_squfof (uintmax_t n1, uintmax_t n0, struct factors *factors) -{ - /* Uses algorithm and notation from - - SQUARE FORM FACTORIZATION - JASON E. GOWER AND SAMUEL S. WAGSTAFF, JR. - - https://homes.cerias.purdue.edu/~ssw/squfof.pdf - */ - - static const unsigned int multipliers_1[] = - { /* = 1 (mod 4) */ - 105, 165, 21, 385, 33, 5, 77, 1, 0 - }; - static const unsigned int multipliers_3[] = - { /* = 3 (mod 4) */ - 1155, 15, 231, 35, 3, 55, 7, 11, 0 - }; - - const unsigned int *m; - - struct { uintmax_t Q; uintmax_t P; } queue[QUEUE_SIZE]; - - if (n1 >= ((uintmax_t) 1 << (W_TYPE_SIZE - 2))) - return false; - - uintmax_t sqrt_n = isqrt2 (n1, n0); - - if (n0 == sqrt_n * sqrt_n) - { - uintmax_t p1, p0; - - umul_ppmm (p1, p0, sqrt_n, sqrt_n); - assert (p0 == n0); - - if (n1 == p1) - { - if (prime_p (sqrt_n)) - factor_insert_multiplicity (factors, sqrt_n, 2); - else - { - struct factors f; - - f.nfactors = 0; - if (!factor_using_squfof (0, sqrt_n, &f)) - { - /* Try pollard rho instead */ - factor_using_pollard_rho (sqrt_n, 1, &f); - } - /* Duplicate the new factors */ - for (unsigned int i = 0; i < f.nfactors; i++) - factor_insert_multiplicity (factors, f.p[i], 2*f.e[i]); - } - return true; - } - } - - /* Select multipliers so we always get n * mu = 3 (mod 4) */ - for (m = (n0 % 4 == 1) ? multipliers_3 : multipliers_1; - *m; m++) - { - uintmax_t S, Dh, Dl, Q1, Q, P, L, L1, B; - unsigned int i; - unsigned int mu = *m; - unsigned int qpos = 0; - - assert (mu * n0 % 4 == 3); - - /* In the notation of the paper, with mu * n == 3 (mod 4), we - get \Delta = 4 mu * n, and the paper's \mu is 2 mu. As far as - I understand it, the necessary bound is 4 \mu^3 < n, or 32 - mu^3 < n. - - However, this seems insufficient: With n = 37243139 and mu = - 105, we get a trivial factor, from the square 38809 = 197^2, - without any corresponding Q earlier in the iteration. - - Requiring 64 mu^3 < n seems sufficient. */ - if (n1 == 0) - { - if ((uintmax_t) mu*mu*mu >= n0 / 64) - continue; - } - else - { - if (n1 > ((uintmax_t) 1 << (W_TYPE_SIZE - 2)) / mu) - continue; - } - umul_ppmm (Dh, Dl, n0, mu); - Dh += n1 * mu; - - assert (Dl % 4 != 1); - assert (Dh < (uintmax_t) 1 << (W_TYPE_SIZE - 2)); - - S = isqrt2 (Dh, Dl); - - Q1 = 1; - P = S; - - /* Square root remainder fits in one word, so ignore high part. */ - Q = Dl - P*P; - /* FIXME: When can this differ from floor(sqrt(2 sqrt(D)))? */ - L = isqrt (2*S); - B = 2*L; - L1 = mu * 2 * L; - - /* The form is (+/- Q1, 2P, -/+ Q), of discriminant 4 (P^2 + Q Q1) = - 4 D. */ - - for (i = 0; i <= B; i++) - { - uintmax_t q, P1, t, rem; - - div_smallq (q, rem, S+P, Q); - P1 = S - rem; /* P1 = q*Q - P */ - - IF_LINT (assert (q > 0 && Q > 0)); - -# if STAT_SQUFOF - q_freq[0]++; - q_freq[MIN (q, Q_FREQ_SIZE)]++; -# endif - - if (Q <= L1) - { - uintmax_t g = Q; - - if ( (Q & 1) == 0) - g /= 2; - - g /= gcd_odd (g, mu); - - if (g <= L) - { - if (qpos >= QUEUE_SIZE) - die (EXIT_FAILURE, 0, _("squfof queue overflow")); - queue[qpos].Q = g; - queue[qpos].P = P % g; - qpos++; - } - } - - /* I think the difference can be either sign, but mod - 2^W_TYPE_SIZE arithmetic should be fine. */ - t = Q1 + q * (P - P1); - Q1 = Q; - Q = t; - P = P1; - - if ( (i & 1) == 0) - { - uintmax_t r = is_square (Q); - if (r) - { - for (unsigned int j = 0; j < qpos; j++) - { - if (queue[j].Q == r) - { - if (r == 1) - /* Traversed entire cycle. */ - goto next_multiplier; - - /* Need the absolute value for divisibility test. */ - if (P >= queue[j].P) - t = P - queue[j].P; - else - t = queue[j].P - P; - if (t % r == 0) - { - /* Delete entries up to and including entry - j, which matched. */ - memmove (queue, queue + j + 1, - (qpos - j - 1) * sizeof (queue[0])); - qpos -= (j + 1); - } - goto next_i; - } - } - - /* We have found a square form, which should give a - factor. */ - Q1 = r; - assert (S >= P); /* What signs are possible? */ - P += r * ((S - P) / r); - - /* Note: Paper says (N - P*P) / Q1, that seems incorrect - for the case D = 2N. */ - /* Compute Q = (D - P*P) / Q1, but we need double - precision. */ - uintmax_t hi, lo; - umul_ppmm (hi, lo, P, P); - sub_ddmmss (hi, lo, Dh, Dl, hi, lo); - udiv_qrnnd (Q, rem, hi, lo, Q1); - assert (rem == 0); - - for (;;) - { - /* Note: There appears to by a typo in the paper, - Step 4a in the algorithm description says q <-- - floor([S+P]/\hat Q), but looking at the equations - in Sec. 3.1, it should be q <-- floor([S+P] / Q). - (In this code, \hat Q is Q1). */ - div_smallq (q, rem, S+P, Q); - P1 = S - rem; /* P1 = q*Q - P */ - -# if STAT_SQUFOF - q_freq[0]++; - q_freq[MIN (q, Q_FREQ_SIZE)]++; -# endif - if (P == P1) - break; - t = Q1 + q * (P - P1); - Q1 = Q; - Q = t; - P = P1; - } - - if ( (Q & 1) == 0) - Q /= 2; - Q /= gcd_odd (Q, mu); - - assert (Q > 1 && (n1 || Q < n0)); - - if (prime_p (Q)) - factor_insert (factors, Q); - else if (!factor_using_squfof (0, Q, factors)) - factor_using_pollard_rho (Q, 2, factors); - - divexact_21 (n1, n0, n1, n0, Q); - - if (prime2_p (n1, n0)) - factor_insert_large (factors, n1, n0); - else - { - if (!factor_using_squfof (n1, n0, factors)) - { - if (n1 == 0) - factor_using_pollard_rho (n0, 1, factors); - else - factor_using_pollard_rho2 (n1, n0, 1, factors); - } - } - - return true; - } - } - next_i:; - } - next_multiplier:; - } - return false; -} -#endif - -/* Compute the prime factors of the 128-bit number (T1,T0), and put the - results in FACTORS. */ -static void -factor (uintmax_t t1, uintmax_t t0, struct factors *factors) -{ - factors->nfactors = 0; - factors->plarge[1] = 0; - - if (t1 == 0 && t0 < 2) - return; - - t0 = factor_using_division (&t1, t1, t0, factors); - - if (t1 == 0 && t0 < 2) - return; - - if (prime2_p (t1, t0)) - factor_insert_large (factors, t1, t0); - else - { -#if USE_SQUFOF - if (factor_using_squfof (t1, t0, factors)) - return; -#endif - - if (t1 == 0) - factor_using_pollard_rho (t0, 1, factors); - else - factor_using_pollard_rho2 (t1, t0, 1, factors); - } -} - /* Use Pollard-rho to compute the prime factors of arbitrary-precision T, and put the results in FACTORS. */ static void @@ -2250,206 +507,6 @@ mp_factor (mpz_t t, struct mp_factors *factors) } } -static strtol_error -strto2uintmax (uintmax_t *hip, uintmax_t *lop, const char *s) -{ - unsigned int lo_carry; - uintmax_t hi = 0, lo = 0; - - strtol_error err = LONGINT_INVALID; - - /* Initial scan for invalid digits. */ - const char *p = s; - for (;;) - { - unsigned int c = *p++; - if (c == 0) - break; - - if (UNLIKELY (!ISDIGIT (c))) - { - err = LONGINT_INVALID; - break; - } - - err = LONGINT_OK; /* we've seen at least one valid digit */ - } - - while (err == LONGINT_OK) - { - unsigned int c = *s++; - if (c == 0) - break; - - c -= '0'; - - if (UNLIKELY (hi > ~(uintmax_t)0 / 10)) - { - err = LONGINT_OVERFLOW; - break; - } - hi = 10 * hi; - - lo_carry = (lo >> (W_TYPE_SIZE - 3)) + (lo >> (W_TYPE_SIZE - 1)); - lo_carry += 10 * lo < 2 * lo; - - lo = 10 * lo; - lo += c; - - lo_carry += lo < c; - hi += lo_carry; - if (UNLIKELY (hi < lo_carry)) - { - err = LONGINT_OVERFLOW; - break; - } - } - - *hip = hi; - *lop = lo; - - return err; -} - -/* Structure and routines for buffering and outputting full lines, - to support parallel operation efficiently. */ -static struct lbuf_ -{ - char *buf; - char *end; -} lbuf; - -/* 512 is chosen to give good performance, - and also is the max guaranteed size that - consumers can read atomically through pipes. - Also it's big enough to cater for max line length - even with 128 bit uintmax_t. */ -#define FACTOR_PIPE_BUF 512 - -static void -lbuf_alloc (void) -{ - if (lbuf.buf) - return; - - /* Double to ensure enough space for - previous numbers + next number. */ - lbuf.buf = xmalloc (FACTOR_PIPE_BUF * 2); - lbuf.end = lbuf.buf; -} - -/* Write complete LBUF to standard output. */ -static void -lbuf_flush (void) -{ - size_t size = lbuf.end - lbuf.buf; - if (full_write (STDOUT_FILENO, lbuf.buf, size) != size) - die (EXIT_FAILURE, errno, "%s", _("write error")); - lbuf.end = lbuf.buf; -} - -/* Add a character C to LBUF and if it's a newline - and enough bytes are already buffered, - then write atomically to standard output. */ -static void -lbuf_putc (char c) -{ - *lbuf.end++ = c; - - if (c == '\n') - { - size_t buffered = lbuf.end - lbuf.buf; - - /* Provide immediate output for interactive use. */ - static int line_buffered = -1; - if (line_buffered == -1) - line_buffered = isatty (STDIN_FILENO) || isatty (STDOUT_FILENO); - if (line_buffered) - lbuf_flush (); - else if (buffered >= FACTOR_PIPE_BUF) - { - /* Write output in <= PIPE_BUF chunks - so consumers can read atomically. */ - char const *tend = lbuf.end; - - /* Since a umaxint_t's factors must fit in 512 - we're guaranteed to find a newline here. */ - char *tlend = lbuf.buf + FACTOR_PIPE_BUF; - while (*--tlend != '\n'); - tlend++; - - lbuf.end = tlend; - lbuf_flush (); - - /* Buffer the remainder. */ - memcpy (lbuf.buf, tlend, tend - tlend); - lbuf.end = lbuf.buf + (tend - tlend); - } - } -} - -/* Buffer an int to the internal LBUF. */ -static void -lbuf_putint (uintmax_t i, size_t min_width) -{ - char buf[INT_BUFSIZE_BOUND (uintmax_t)]; - char const *umaxstr = umaxtostr (i, buf); - size_t width = sizeof (buf) - (umaxstr - buf) - 1; - size_t z = width; - - for (; z < min_width; z++) - *lbuf.end++ = '0'; - - memcpy (lbuf.end, umaxstr, width); - lbuf.end += width; -} - -static void -print_uintmaxes (uintmax_t t1, uintmax_t t0) -{ - uintmax_t q, r; - - if (t1 == 0) - lbuf_putint (t0, 0); - else - { - /* Use very plain code here since it seems hard to write fast code - without assuming a specific word size. */ - q = t1 / 1000000000; - r = t1 % 1000000000; - udiv_qrnnd (t0, r, r, t0, 1000000000); - print_uintmaxes (q, t0); - lbuf_putint (r, 9); - } -} - -/* Single-precision factoring */ -static void -print_factors_single (uintmax_t t1, uintmax_t t0) -{ - struct factors factors; - - print_uintmaxes (t1, t0); - lbuf_putc (':'); - - factor (t1, t0, &factors); - - for (unsigned int j = 0; j < factors.nfactors; j++) - for (unsigned int k = 0; k < factors.e[j]; k++) - { - lbuf_putc (' '); - print_uintmaxes (0, factors.p[j]); - } - - if (factors.plarge[1]) - { - lbuf_putc (' '); - print_uintmaxes (factors.plarge[1], factors.plarge[0]); - } - - lbuf_putc ('\n'); -} - /* Emit the factors of the indicated number. If we have the option of using either algorithm, we select on the basis of the length of the number. For longer numbers, we prefer the MP algorithm even if the native algorithm @@ -2464,40 +521,16 @@ print_factors (const char *input) str++; str += *str == '+'; - uintmax_t t1, t0; - - /* Try converting the number to one or two words. If it fails, use GMP or - print an error message. The 2nd condition checks that the most - significant bit of the two-word number is clear, in a typesize neutral - way. */ - strtol_error err = strto2uintmax (&t1, &t0, str); + devmsg ("[using arbitrary-precision arithmetic] "); + mpz_t t; + struct mp_factors factors; - switch (err) + if (*str == '-' || mpz_init_set_str (t, str, 10) != 0) { - case LONGINT_OK: - if (((t1 << 1) >> 1) == t1) - { - devmsg ("[using single-precision arithmetic] "); - print_factors_single (t1, t0); - return true; - } - break; - - case LONGINT_OVERFLOW: - /* Try GMP. */ - break; - - default: error (0, 0, _("%s is not a valid positive integer"), quote (input)); return false; } - devmsg ("[using arbitrary-precision arithmetic] "); - mpz_t t; - struct mp_factors factors; - - mpz_init_set_str (t, str, 10); - gmp_printf ("%Zd:", t); mp_factor (t, &factors); @@ -2566,9 +599,7 @@ main (int argc, char **argv) bindtextdomain (PACKAGE, LOCALEDIR); textdomain (PACKAGE); - lbuf_alloc (); atexit (close_stdout); - atexit (lbuf_flush); int c; while ((c = getopt_long (argc, argv, "", long_options, NULL)) != -1) @@ -2588,10 +619,6 @@ main (int argc, char **argv) } } -#if STAT_SQUFOF - memset (q_freq, 0, sizeof (q_freq)); -#endif - bool ok; if (argc <= optind) ok = do_stdin (); @@ -2603,20 +630,5 @@ main (int argc, char **argv) ok = false; } -#if STAT_SQUFOF - if (q_freq[0] > 0) - { - double acc_f; - printf ("q freq. cum. freq.(total: %d)\n", q_freq[0]); - for (unsigned int i = 1, acc_f = 0.0; i <= Q_FREQ_SIZE; i++) - { - double f = (double) q_freq[i] / q_freq[0]; - acc_f += f; - printf ("%s%d %.2f%% %.2f%%\n", i == Q_FREQ_SIZE ? ">=" : "", i, - 100.0 * f, 100.0 * acc_f); - } - } -#endif - return ok ? EXIT_SUCCESS : EXIT_FAILURE; } diff --git a/src/local.mk b/src/local.mk index 72db9c704..b4cd8e5b3 100644 --- a/src/local.mk +++ b/src/local.mk @@ -52,7 +52,6 @@ noinst_HEADERS = \ src/fs-is-local.h \ src/group-list.h \ src/ioblksize.h \ - src/longlong.h \ src/ls.h \ src/operand2sig.h \ src/prog-fprintf.h \ diff --git a/src/longlong.h b/src/longlong.h deleted file mode 100644 index e57ba7821..000000000 --- a/src/longlong.h +++ /dev/null @@ -1,2267 +0,0 @@ -/* longlong.h -- definitions for mixed size 32/64 bit arithmetic. - -Copyright 1991-2020 Free Software Foundation, Inc. - -This file 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 3 of the License, or (at your option) any -later version. - -This file 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 this file. If not, see https://www.gnu.org/licenses/. */ - -/* You have to define the following before including this file: - - UWtype -- An unsigned type, default type for operations (typically a "word") - UHWtype -- An unsigned type, at least half the size of UWtype - UDWtype -- An unsigned type, at least twice as large a UWtype - W_TYPE_SIZE -- size in bits of UWtype - - SItype, USItype -- Signed and unsigned 32 bit types - DItype, UDItype -- Signed and unsigned 64 bit types - - On a 32 bit machine UWtype should typically be USItype; - on a 64 bit machine, UWtype should typically be UDItype. - - Optionally, define: - - LONGLONG_STANDALONE -- Avoid code that needs machine-dependent support files - NO_ASM -- Disable inline asm - - - CAUTION! Using this version of longlong.h outside of GMP is not safe. You - need to include gmp.h and gmp-impl.h, or certain things might not work as - expected. -*/ - -#define __BITS4 (W_TYPE_SIZE / 4) -#define __ll_B ((UWtype) 1 << (W_TYPE_SIZE / 2)) -#define __ll_lowpart(t) ((UWtype) (t) & (__ll_B - 1)) -#define __ll_highpart(t) ((UWtype) (t) >> (W_TYPE_SIZE / 2)) - -/* This is used to make sure no undesirable sharing between different libraries - that use this file takes place. */ -#ifndef __MPN -#define __MPN(x) __##x -#endif - -/* Define auxiliary asm macros. - - 1) umul_ppmm(high_prod, low_prod, multiplier, multiplicand) multiplies two - UWtype integers MULTIPLIER and MULTIPLICAND, and generates a two UWtype - word product in HIGH_PROD and LOW_PROD. - - 2) __umulsidi3(a,b) multiplies two UWtype integers A and B, and returns a - UDWtype product. This is just a variant of umul_ppmm. - - 3) udiv_qrnnd(quotient, remainder, high_numerator, low_numerator, - denominator) divides a UDWtype, composed by the UWtype integers - HIGH_NUMERATOR and LOW_NUMERATOR, by DENOMINATOR and places the quotient - in QUOTIENT and the remainder in REMAINDER. HIGH_NUMERATOR must be less - than DENOMINATOR for correct operation. If, in addition, the most - significant bit of DENOMINATOR must be 1, then the pre-processor symbol - UDIV_NEEDS_NORMALIZATION is defined to 1. - - 4) sdiv_qrnnd(quotient, remainder, high_numerator, low_numerator, - denominator). Like udiv_qrnnd but the numbers are signed. The quotient - is rounded towards 0. - - 5) count_leading_zeros(count, x) counts the number of zero-bits from the - msb to the first non-zero bit in the UWtype X. This is the number of - steps X needs to be shifted left to set the msb. Undefined for X == 0, - unless the symbol COUNT_LEADING_ZEROS_0 is defined to some value. - - 6) count_trailing_zeros(count, x) like count_leading_zeros, but counts - from the least significant end. - - 7) add_ssaaaa(high_sum, low_sum, high_addend_1, low_addend_1, - high_addend_2, low_addend_2) adds two UWtype integers, composed by - HIGH_ADDEND_1 and LOW_ADDEND_1, and HIGH_ADDEND_2 and LOW_ADDEND_2 - respectively. The result is placed in HIGH_SUM and LOW_SUM. Overflow - (i.e. carry out) is not stored anywhere, and is lost. - - 8) sub_ddmmss(high_difference, low_difference, high_minuend, low_minuend, - high_subtrahend, low_subtrahend) subtracts two two-word UWtype integers, - composed by HIGH_MINUEND_1 and LOW_MINUEND_1, and HIGH_SUBTRAHEND_2 and - LOW_SUBTRAHEND_2 respectively. The result is placed in HIGH_DIFFERENCE - and LOW_DIFFERENCE. Overflow (i.e. carry out) is not stored anywhere, - and is lost. - - If any of these macros are left undefined for a particular CPU, - C macros are used. - - - Notes: - - For add_ssaaaa the two high and two low addends can both commute, but - unfortunately gcc only supports one "%" commutative in each asm block. - This has always been so but is only documented in recent versions - (eg. pre-release 3.3). Having two or more "%"s can cause an internal - compiler error in certain rare circumstances. - - Apparently it was only the last "%" that was ever actually respected, so - the code has been updated to leave just that. Clearly there's a free - choice whether high or low should get it, if there's a reason to favour - one over the other. Also obviously when the constraints on the two - operands are identical there's no benefit to the reloader in any "%" at - all. - - */ - -/* The CPUs come in alphabetical order below. - - Please add support for more CPUs here, or improve the current support - for the CPUs below! */ - - -/* count_leading_zeros_gcc_clz is count_leading_zeros implemented with gcc - 3.4 __builtin_clzl or __builtin_clzll, according to our limb size. - Similarly count_trailing_zeros_gcc_ctz using __builtin_ctzl or - __builtin_ctzll. - - These builtins are only used when we check what code comes out, on some - chips they're merely libgcc calls, where we will instead want an inline - in that case (either asm or generic C). - - These builtins are better than an asm block of the same insn, since an - asm block doesn't give gcc any information about scheduling or resource - usage. We keep an asm block for use on prior versions of gcc though. - - For reference, __builtin_ffs existed in gcc prior to __builtin_clz, but - it's not used (for count_leading_zeros) because it generally gives extra - code to ensure the result is 0 when the input is 0, which we don't need - or want. */ - -#ifdef _LONG_LONG_LIMB -#define count_leading_zeros_gcc_clz(count,x) \ - do { \ - ASSERT ((x) != 0); \ - (count) = __builtin_clzll (x); \ - } while (0) -#else -#define count_leading_zeros_gcc_clz(count,x) \ - do { \ - ASSERT ((x) != 0); \ - (count) = __builtin_clzl (x); \ - } while (0) -#endif - -#ifdef _LONG_LONG_LIMB -#define count_trailing_zeros_gcc_ctz(count,x) \ - do { \ - ASSERT ((x) != 0); \ - (count) = __builtin_ctzll (x); \ - } while (0) -#else -#define count_trailing_zeros_gcc_ctz(count,x) \ - do { \ - ASSERT ((x) != 0); \ - (count) = __builtin_ctzl (x); \ - } while (0) -#endif - - -/* FIXME: The macros using external routines like __MPN(count_leading_zeros) - don't need to be under !NO_ASM */ -#if ! defined (NO_ASM) - -#if defined (__alpha) && W_TYPE_SIZE == 64 -/* Most alpha-based machines, except Cray systems. */ -#if defined (__GNUC__) -#if __GMP_GNUC_PREREQ (3,3) -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UDItype __m0 = (m0), __m1 = (m1); \ - (ph) = __builtin_alpha_umulh (__m0, __m1); \ - (pl) = __m0 * __m1; \ - } while (0) -#else -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UDItype __m0 = (m0), __m1 = (m1); \ - __asm__ ("umulh %r1,%2,%0" \ - : "=r" (ph) \ - : "%rJ" (__m0), "rI" (__m1)); \ - (pl) = __m0 * __m1; \ - } while (0) -#endif -#else /* ! __GNUC__ */ -#include -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UDItype __m0 = (m0), __m1 = (m1); \ - (ph) = __UMULH (__m0, __m1); \ - (pl) = __m0 * __m1; \ - } while (0) -#endif -#ifndef LONGLONG_STANDALONE -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { UWtype __di; \ - __di = __MPN(invert_limb) (d); \ - udiv_qrnnd_preinv (q, r, n1, n0, d, __di); \ - } while (0) -#define UDIV_PREINV_ALWAYS 1 -#define UDIV_NEEDS_NORMALIZATION 1 -#endif /* LONGLONG_STANDALONE */ - -/* clz_tab is required in all configurations, since mpn/alpha/cntlz.asm - always goes into libgmp.so, even when not actually used. */ -#define COUNT_LEADING_ZEROS_NEED_CLZ_TAB - -#if defined (__GNUC__) && HAVE_HOST_CPU_alpha_CIX -#define count_leading_zeros(COUNT,X) \ - __asm__("ctlz %1,%0" : "=r"(COUNT) : "r"(X)) -#define count_trailing_zeros(COUNT,X) \ - __asm__("cttz %1,%0" : "=r"(COUNT) : "r"(X)) -#endif /* clz/ctz using cix */ - -#if ! defined (count_leading_zeros) \ - && defined (__GNUC__) && ! defined (LONGLONG_STANDALONE) -/* ALPHA_CMPBGE_0 gives "cmpbge $31,src,dst", ie. test src bytes == 0. - "$31" is written explicitly in the asm, since an "r" constraint won't - select reg 31. There seems no need to worry about "r31" syntax for cray, - since gcc itself (pre-release 3.4) emits just $31 in various places. */ -#define ALPHA_CMPBGE_0(dst, src) \ - do { asm ("cmpbge $31, %1, %0" : "=r" (dst) : "r" (src)); } while (0) -/* Zero bytes are turned into bits with cmpbge, a __clz_tab lookup counts - them, locating the highest non-zero byte. A second __clz_tab lookup - counts the leading zero bits in that byte, giving the result. */ -#define count_leading_zeros(count, x) \ - do { \ - UWtype __clz__b, __clz__c, __clz__x = (x); \ - ALPHA_CMPBGE_0 (__clz__b, __clz__x); /* zero bytes */ \ - __clz__b = __clz_tab [(__clz__b >> 1) ^ 0x7F]; /* 8 to 1 byte */ \ - __clz__b = __clz__b * 8 - 7; /* 57 to 1 shift */ \ - __clz__x >>= __clz__b; \ - __clz__c = __clz_tab [__clz__x]; /* 8 to 1 bit */ \ - __clz__b = 65 - __clz__b; \ - (count) = __clz__b - __clz__c; \ - } while (0) -#define COUNT_LEADING_ZEROS_NEED_CLZ_TAB -#endif /* clz using cmpbge */ - -#if ! defined (count_leading_zeros) && ! defined (LONGLONG_STANDALONE) -#if HAVE_ATTRIBUTE_CONST -long __MPN(count_leading_zeros) (UDItype) __attribute__ ((const)); -#else -long __MPN(count_leading_zeros) (UDItype); -#endif -#define count_leading_zeros(count, x) \ - ((count) = __MPN(count_leading_zeros) (x)) -#endif /* clz using mpn */ -#endif /* __alpha */ - -#if defined (__AVR) && W_TYPE_SIZE == 8 -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - unsigned short __p = (unsigned short) (m0) * (m1); \ - (ph) = __p >> 8; \ - (pl) = __p; \ - } while (0) -#endif /* AVR */ - -#if defined (_CRAY) && W_TYPE_SIZE == 64 -#include -#define UDIV_PREINV_ALWAYS 1 -#define UDIV_NEEDS_NORMALIZATION 1 -long __MPN(count_leading_zeros) (UDItype); -#define count_leading_zeros(count, x) \ - ((count) = _leadz ((UWtype) (x))) -#if defined (_CRAYIEEE) /* I.e., Cray T90/ieee, T3D, and T3E */ -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UDItype __m0 = (m0), __m1 = (m1); \ - (ph) = _int_mult_upper (__m0, __m1); \ - (pl) = __m0 * __m1; \ - } while (0) -#ifndef LONGLONG_STANDALONE -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { UWtype __di; \ - __di = __MPN(invert_limb) (d); \ - udiv_qrnnd_preinv (q, r, n1, n0, d, __di); \ - } while (0) -#endif /* LONGLONG_STANDALONE */ -#endif /* _CRAYIEEE */ -#endif /* _CRAY */ - -#if defined (__ia64) && W_TYPE_SIZE == 64 -/* This form encourages gcc (pre-release 3.4 at least) to emit predicated - "sub r=r,r" and "sub r=r,r,1", giving a 2 cycle latency. The generic - code using "al>= _c; \ - if (_x >= 1 << 4) \ - _x >>= 4, _c += 4; \ - if (_x >= 1 << 2) \ - _x >>= 2, _c += 2; \ - _c += _x >> 1; \ - (count) = W_TYPE_SIZE - 1 - _c; \ - } while (0) -/* similar to what gcc does for __builtin_ffs, but 0 based rather than 1 - based, and we don't need a special case for x==0 here */ -#define count_trailing_zeros(count, x) \ - do { \ - UWtype __ctz_x = (x); \ - __asm__ ("popcnt %0 = %1" \ - : "=r" (count) \ - : "r" ((__ctz_x-1) & ~__ctz_x)); \ - } while (0) -#endif -#if defined (__INTEL_COMPILER) -#include -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UWtype __m0 = (m0), __m1 = (m1); \ - ph = _m64_xmahu (__m0, __m1, 0); \ - pl = __m0 * __m1; \ - } while (0) -#endif -#ifndef LONGLONG_STANDALONE -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { UWtype __di; \ - __di = __MPN(invert_limb) (d); \ - udiv_qrnnd_preinv (q, r, n1, n0, d, __di); \ - } while (0) -#define UDIV_PREINV_ALWAYS 1 -#define UDIV_NEEDS_NORMALIZATION 1 -#endif -#endif - - -#if defined (__GNUC__) - -/* We sometimes need to clobber "cc" with gcc2, but that would not be - understood by gcc1. Use cpp to avoid major code duplication. */ -#if __GNUC__ < 2 -#define __CLOBBER_CC -#define __AND_CLOBBER_CC -#else /* __GNUC__ >= 2 */ -#define __CLOBBER_CC : "cc" -#define __AND_CLOBBER_CC , "cc" -#endif /* __GNUC__ < 2 */ - -#if (defined (__a29k__) || defined (_AM29K)) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("add %1,%4,%5\n\taddc %0,%2,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "%r" (al), "rI" (bl)) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("sub %1,%4,%5\n\tsubc %0,%2,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "r" (al), "rI" (bl)) -#define umul_ppmm(xh, xl, m0, m1) \ - do { \ - USItype __m0 = (m0), __m1 = (m1); \ - __asm__ ("multiplu %0,%1,%2" \ - : "=r" (xl) \ - : "r" (__m0), "r" (__m1)); \ - __asm__ ("multmu %0,%1,%2" \ - : "=r" (xh) \ - : "r" (__m0), "r" (__m1)); \ - } while (0) -#define udiv_qrnnd(q, r, n1, n0, d) \ - __asm__ ("dividu %0,%3,%4" \ - : "=r" (q), "=q" (r) \ - : "1" (n1), "r" (n0), "r" (d)) -#define count_leading_zeros(count, x) \ - __asm__ ("clz %0,%1" \ - : "=r" (count) \ - : "r" (x)) -#define COUNT_LEADING_ZEROS_0 32 -#endif /* __a29k__ */ - -#if defined (__arc__) -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("add.f\t%1, %4, %5\n\tadc\t%0, %2, %3" \ - : "=r" (sh), \ - "=&r" (sl) \ - : "r" ((USItype) (ah)), \ - "rICal" ((USItype) (bh)), \ - "%r" ((USItype) (al)), \ - "rICal" ((USItype) (bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("sub.f\t%1, %4, %5\n\tsbc\t%0, %2, %3" \ - : "=r" (sh), \ - "=&r" (sl) \ - : "r" ((USItype) (ah)), \ - "rICal" ((USItype) (bh)), \ - "r" ((USItype) (al)), \ - "rICal" ((USItype) (bl))) -#endif - -#if defined (__arm__) && (defined (__thumb2__) || !defined (__thumb__)) \ - && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (bl) && -(USItype)(bl) < 0x100) \ - __asm__ ("subs\t%1, %4, %5\n\tadc\t%0, %2, %3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), \ - "%r" (al), "rI" (-(USItype)(bl)) __CLOBBER_CC); \ - else \ - __asm__ ("adds\t%1, %4, %5\n\tadc\t%0, %2, %3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "%r" (al), "rI" (bl) __CLOBBER_CC); \ - } while (0) -/* FIXME: Extend the immediate range for the low word by using both ADDS and - SUBS, since they set carry in the same way. Note: We need separate - definitions for thumb and non-thumb due to the absence of RSC on thumb. */ -#if defined (__thumb__) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (ah) && __builtin_constant_p (bh) \ - && (ah) == (bh)) \ - __asm__ ("subs\t%1, %2, %3\n\tsbc\t%0, %0, %0" \ - : "=r" (sh), "=r" (sl) \ - : "r" (al), "rI" (bl) __CLOBBER_CC); \ - else if (__builtin_constant_p (al)) \ - __asm__ ("rsbs\t%1, %5, %4\n\tsbc\t%0, %2, %3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "rI" (al), "r" (bl) __CLOBBER_CC); \ - else if (__builtin_constant_p (bl)) \ - __asm__ ("subs\t%1, %4, %5\n\tsbc\t%0, %2, %3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "r" (al), "rI" (bl) __CLOBBER_CC); \ - else \ - __asm__ ("subs\t%1, %4, %5\n\tsbc\t%0, %2, %3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "r" (al), "rI" (bl) __CLOBBER_CC); \ - } while (0) -#else -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (ah) && __builtin_constant_p (bh) \ - && (ah) == (bh)) \ - __asm__ ("subs\t%1, %2, %3\n\tsbc\t%0, %0, %0" \ - : "=r" (sh), "=r" (sl) \ - : "r" (al), "rI" (bl) __CLOBBER_CC); \ - else if (__builtin_constant_p (al)) \ - { \ - if (__builtin_constant_p (ah)) \ - __asm__ ("rsbs\t%1, %5, %4\n\trsc\t%0, %3, %2" \ - : "=r" (sh), "=&r" (sl) \ - : "rI" (ah), "r" (bh), "rI" (al), "r" (bl) __CLOBBER_CC); \ - else \ - __asm__ ("rsbs\t%1, %5, %4\n\tsbc\t%0, %2, %3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "rI" (al), "r" (bl) __CLOBBER_CC); \ - } \ - else if (__builtin_constant_p (ah)) \ - { \ - if (__builtin_constant_p (bl)) \ - __asm__ ("subs\t%1, %4, %5\n\trsc\t%0, %3, %2" \ - : "=r" (sh), "=&r" (sl) \ - : "rI" (ah), "r" (bh), "r" (al), "rI" (bl) __CLOBBER_CC); \ - else \ - __asm__ ("rsbs\t%1, %5, %4\n\trsc\t%0, %3, %2" \ - : "=r" (sh), "=&r" (sl) \ - : "rI" (ah), "r" (bh), "rI" (al), "r" (bl) __CLOBBER_CC); \ - } \ - else if (__builtin_constant_p (bl)) \ - __asm__ ("subs\t%1, %4, %5\n\tsbc\t%0, %2, %3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "r" (al), "rI" (bl) __CLOBBER_CC); \ - else /* only bh might be a constant */ \ - __asm__ ("subs\t%1, %4, %5\n\tsbc\t%0, %2, %3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "rI" (bh), "r" (al), "rI" (bl) __CLOBBER_CC); \ - } while (0) -#endif -#if defined (__ARM_ARCH_2__) || defined (__ARM_ARCH_2A__) \ - || defined (__ARM_ARCH_3__) -#define umul_ppmm(xh, xl, a, b) \ - do { \ - register USItype __t0, __t1, __t2; \ - __asm__ ("%@ Inlined umul_ppmm\n" \ - " mov %2, %5, lsr #16\n" \ - " mov %0, %6, lsr #16\n" \ - " bic %3, %5, %2, lsl #16\n" \ - " bic %4, %6, %0, lsl #16\n" \ - " mul %1, %3, %4\n" \ - " mul %4, %2, %4\n" \ - " mul %3, %0, %3\n" \ - " mul %0, %2, %0\n" \ - " adds %3, %4, %3\n" \ - " addcs %0, %0, #65536\n" \ - " adds %1, %1, %3, lsl #16\n" \ - " adc %0, %0, %3, lsr #16" \ - : "=&r" ((USItype) (xh)), "=r" ((USItype) (xl)), \ - "=&r" (__t0), "=&r" (__t1), "=r" (__t2) \ - : "r" ((USItype) (a)), "r" ((USItype) (b)) __CLOBBER_CC); \ - } while (0) -#ifndef LONGLONG_STANDALONE -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { UWtype __r; \ - (q) = __MPN(udiv_qrnnd) (&__r, (n1), (n0), (d)); \ - (r) = __r; \ - } while (0) -extern UWtype __MPN(udiv_qrnnd) (UWtype *, UWtype, UWtype, UWtype); -#endif /* LONGLONG_STANDALONE */ -#else /* ARMv4 or newer */ -#define umul_ppmm(xh, xl, a, b) \ - __asm__ ("umull %0,%1,%2,%3" : "=&r" (xl), "=&r" (xh) : "r" (a), "r" (b)) -#define smul_ppmm(xh, xl, a, b) \ - __asm__ ("smull %0,%1,%2,%3" : "=&r" (xl), "=&r" (xh) : "r" (a), "r" (b)) -#ifndef LONGLONG_STANDALONE -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { UWtype __di; \ - __di = __MPN(invert_limb) (d); \ - udiv_qrnnd_preinv (q, r, n1, n0, d, __di); \ - } while (0) -#define UDIV_PREINV_ALWAYS 1 -#define UDIV_NEEDS_NORMALIZATION 1 -#endif /* LONGLONG_STANDALONE */ -#endif /* defined(__ARM_ARCH_2__) ... */ -#define count_leading_zeros(count, x) count_leading_zeros_gcc_clz(count, x) -#define count_trailing_zeros(count, x) count_trailing_zeros_gcc_ctz(count, x) -#endif /* __arm__ */ - -#if defined (__aarch64__) && W_TYPE_SIZE == 64 -/* FIXME: Extend the immediate range for the low word by using both - ADDS and SUBS, since they set carry in the same way. */ -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (bl) && -(UDItype)(bl) < 0x1000) \ - __asm__ ("subs\t%1, %x4, %5\n\tadc\t%0, %x2, %x3" \ - : "=r" (sh), "=&r" (sl) \ - : "rZ" ((UDItype)(ah)), "rZ" ((UDItype)(bh)), \ - "%r" ((UDItype)(al)), "rI" (-(UDItype)(bl)) __CLOBBER_CC);\ - else \ - __asm__ ("adds\t%1, %x4, %5\n\tadc\t%0, %x2, %x3" \ - : "=r" (sh), "=&r" (sl) \ - : "rZ" ((UDItype)(ah)), "rZ" ((UDItype)(bh)), \ - "%r" ((UDItype)(al)), "rI" ((UDItype)(bl)) __CLOBBER_CC);\ - } while (0) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (bl) && -(UDItype)(bl) < 0x1000) \ - __asm__ ("adds\t%1, %x4, %5\n\tsbc\t%0, %x2, %x3" \ - : "=r,r" (sh), "=&r,&r" (sl) \ - : "rZ,rZ" ((UDItype)(ah)), "rZ,rZ" ((UDItype)(bh)), \ - "r,Z" ((UDItype)(al)), "rI,r" (-(UDItype)(bl)) __CLOBBER_CC);\ - else \ - __asm__ ("subs\t%1, %x4, %5\n\tsbc\t%0, %x2, %x3" \ - : "=r,r" (sh), "=&r,&r" (sl) \ - : "rZ,rZ" ((UDItype)(ah)), "rZ,rZ" ((UDItype)(bh)), \ - "r,Z" ((UDItype)(al)), "rI,r" ((UDItype)(bl)) __CLOBBER_CC);\ - } while(0); -#if __GMP_GNUC_PREREQ (4,9) -#define umul_ppmm(w1, w0, u, v) \ - do { \ - typedef unsigned int __ll_UTItype __attribute__((mode(TI))); \ - __ll_UTItype __ll = (__ll_UTItype)(u) * (v); \ - w1 = __ll >> 64; \ - w0 = __ll; \ - } while (0) -#endif -#if !defined (umul_ppmm) -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UDItype __m0 = (m0), __m1 = (m1); \ - __asm__ ("umulh\t%0, %1, %2" : "=r" (ph) : "r" (__m0), "r" (__m1)); \ - (pl) = __m0 * __m1; \ - } while (0) -#endif -#define count_leading_zeros(count, x) count_leading_zeros_gcc_clz(count, x) -#define count_trailing_zeros(count, x) count_trailing_zeros_gcc_ctz(count, x) -#endif /* __aarch64__ */ - -#if defined (__clipper__) && W_TYPE_SIZE == 32 -#define umul_ppmm(w1, w0, u, v) \ - ({union {UDItype __ll; \ - struct {USItype __l, __h;} __i; \ - } __x; \ - __asm__ ("mulwux %2,%0" \ - : "=r" (__x.__ll) \ - : "%0" ((USItype)(u)), "r" ((USItype)(v))); \ - (w1) = __x.__i.__h; (w0) = __x.__i.__l;}) -#define smul_ppmm(w1, w0, u, v) \ - ({union {DItype __ll; \ - struct {SItype __l, __h;} __i; \ - } __x; \ - __asm__ ("mulwx %2,%0" \ - : "=r" (__x.__ll) \ - : "%0" ((SItype)(u)), "r" ((SItype)(v))); \ - (w1) = __x.__i.__h; (w0) = __x.__i.__l;}) -#define __umulsidi3(u, v) \ - ({UDItype __w; \ - __asm__ ("mulwux %2,%0" \ - : "=r" (__w) : "%0" ((USItype)(u)), "r" ((USItype)(v))); \ - __w; }) -#endif /* __clipper__ */ - -/* Fujitsu vector computers. */ -#if defined (__uxp__) && W_TYPE_SIZE == 32 -#define umul_ppmm(ph, pl, u, v) \ - do { \ - union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __asm__ ("mult.lu %1,%2,%0" : "=r" (__x.__ll) : "%r" (u), "rK" (v));\ - (ph) = __x.__i.__h; \ - (pl) = __x.__i.__l; \ - } while (0) -#define smul_ppmm(ph, pl, u, v) \ - do { \ - union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __asm__ ("mult.l %1,%2,%0" : "=r" (__x.__ll) : "%r" (u), "rK" (v)); \ - (ph) = __x.__i.__h; \ - (pl) = __x.__i.__l; \ - } while (0) -#endif - -#if defined (__gmicro__) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("add.w %5,%1\n\taddx %3,%0" \ - : "=g" (sh), "=&g" (sl) \ - : "0" ((USItype)(ah)), "g" ((USItype)(bh)), \ - "%1" ((USItype)(al)), "g" ((USItype)(bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("sub.w %5,%1\n\tsubx %3,%0" \ - : "=g" (sh), "=&g" (sl) \ - : "0" ((USItype)(ah)), "g" ((USItype)(bh)), \ - "1" ((USItype)(al)), "g" ((USItype)(bl))) -#define umul_ppmm(ph, pl, m0, m1) \ - __asm__ ("mulx %3,%0,%1" \ - : "=g" (ph), "=r" (pl) \ - : "%0" ((USItype)(m0)), "g" ((USItype)(m1))) -#define udiv_qrnnd(q, r, nh, nl, d) \ - __asm__ ("divx %4,%0,%1" \ - : "=g" (q), "=r" (r) \ - : "1" ((USItype)(nh)), "0" ((USItype)(nl)), "g" ((USItype)(d))) -#define count_leading_zeros(count, x) \ - __asm__ ("bsch/1 %1,%0" \ - : "=g" (count) : "g" ((USItype)(x)), "0" ((USItype)0)) -#endif - -#if defined (__hppa) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("add%I5 %5,%r4,%1\n\taddc %r2,%r3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "rM" (ah), "rM" (bh), "%rM" (al), "rI" (bl)) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("sub%I4 %4,%r5,%1\n\tsubb %r2,%r3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "rM" (ah), "rM" (bh), "rI" (al), "rM" (bl)) -#if defined (_PA_RISC1_1) -#define umul_ppmm(wh, wl, u, v) \ - do { \ - union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __asm__ ("xmpyu %1,%2,%0" : "=*f" (__x.__ll) : "*f" (u), "*f" (v)); \ - (wh) = __x.__i.__h; \ - (wl) = __x.__i.__l; \ - } while (0) -#endif -#define count_leading_zeros(count, x) \ - do { \ - USItype __tmp; \ - __asm__ ( \ - "ldi 1,%0\n" \ -" extru,= %1,15,16,%%r0 ; Bits 31..16 zero?\n" \ -" extru,tr %1,15,16,%1 ; No. Shift down, skip add.\n" \ -" ldo 16(%0),%0 ; Yes. Perform add.\n" \ -" extru,= %1,23,8,%%r0 ; Bits 15..8 zero?\n" \ -" extru,tr %1,23,8,%1 ; No. Shift down, skip add.\n" \ -" ldo 8(%0),%0 ; Yes. Perform add.\n" \ -" extru,= %1,27,4,%%r0 ; Bits 7..4 zero?\n" \ -" extru,tr %1,27,4,%1 ; No. Shift down, skip add.\n" \ -" ldo 4(%0),%0 ; Yes. Perform add.\n" \ -" extru,= %1,29,2,%%r0 ; Bits 3..2 zero?\n" \ -" extru,tr %1,29,2,%1 ; No. Shift down, skip add.\n" \ -" ldo 2(%0),%0 ; Yes. Perform add.\n" \ -" extru %1,30,1,%1 ; Extract bit 1.\n" \ -" sub %0,%1,%0 ; Subtract it.\n" \ - : "=r" (count), "=r" (__tmp) : "1" (x)); \ - } while (0) -#endif /* hppa */ - -/* These macros are for ABI=2.0w. In ABI=2.0n they can't be used, since GCC - (3.2) puts longlong into two adjacent 32-bit registers. Presumably this - is just a case of no direct support for 2.0n but treating it like 1.0. */ -#if defined (__hppa) && W_TYPE_SIZE == 64 && ! defined (_LONG_LONG_LIMB) -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("add%I5 %5,%r4,%1\n\tadd,dc %r2,%r3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "rM" (ah), "rM" (bh), "%rM" (al), "rI" (bl)) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("sub%I4 %4,%r5,%1\n\tsub,db %r2,%r3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "rM" (ah), "rM" (bh), "rI" (al), "rM" (bl)) -#endif /* hppa */ - -#if (defined (__i370__) || defined (__s390__) || defined (__mvs__)) && W_TYPE_SIZE == 32 -#if defined (__zarch__) || defined (HAVE_HOST_CPU_s390_zarch) -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - do { \ -/* if (__builtin_constant_p (bl)) \ - __asm__ ("alfi\t%1,%o5\n\talcr\t%0,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" (ah), "r" (bh), "%1" (al), "n" (bl) __CLOBBER_CC);\ - else \ -*/ __asm__ ("alr\t%1,%5\n\talcr\t%0,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" (ah), "r" (bh), "%1" (al), "r" (bl)__CLOBBER_CC); \ - } while (0) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - do { \ -/* if (__builtin_constant_p (bl)) \ - __asm__ ("slfi\t%1,%o5\n\tslbr\t%0,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" (ah), "r" (bh), "1" (al), "n" (bl) __CLOBBER_CC); \ - else \ -*/ __asm__ ("slr\t%1,%5\n\tslbr\t%0,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" (ah), "r" (bh), "1" (al), "r" (bl) __CLOBBER_CC); \ - } while (0) -#if __GMP_GNUC_PREREQ (4,5) -#define umul_ppmm(xh, xl, m0, m1) \ - do { \ - union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __x.__ll = (UDItype) (m0) * (UDItype) (m1); \ - (xh) = __x.__i.__h; (xl) = __x.__i.__l; \ - } while (0) -#else -#if 0 -/* FIXME: this fails if gcc knows about the 64-bit registers. Use only - with a new enough processor pretending we have 32-bit registers. */ -#define umul_ppmm(xh, xl, m0, m1) \ - do { \ - union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __asm__ ("mlr\t%0,%2" \ - : "=r" (__x.__ll) \ - : "%0" (m0), "r" (m1)); \ - (xh) = __x.__i.__h; (xl) = __x.__i.__l; \ - } while (0) -#else -#define umul_ppmm(xh, xl, m0, m1) \ - do { \ - /* When we have 64-bit regs and gcc is aware of that, we cannot simply use - DImode for the product, since that would be allocated to a single 64-bit - register, whereas mlr uses the low 32-bits of an even-odd register pair. - */ \ - register USItype __r0 __asm__ ("0"); \ - register USItype __r1 __asm__ ("1") = (m0); \ - __asm__ ("mlr\t%0,%3" \ - : "=r" (__r0), "=r" (__r1) \ - : "r" (__r1), "r" (m1)); \ - (xh) = __r0; (xl) = __r1; \ - } while (0) -#endif /* if 0 */ -#endif -#if 0 -/* FIXME: this fails if gcc knows about the 64-bit registers. Use only - with a new enough processor pretending we have 32-bit registers. */ -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { \ - union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __x.__i.__h = n1; __x.__i.__l = n0; \ - __asm__ ("dlr\t%0,%2" \ - : "=r" (__x.__ll) \ - : "0" (__x.__ll), "r" (d)); \ - (q) = __x.__i.__l; (r) = __x.__i.__h; \ - } while (0) -#else -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { \ - register USItype __r0 __asm__ ("0") = (n1); \ - register USItype __r1 __asm__ ("1") = (n0); \ - __asm__ ("dlr\t%0,%4" \ - : "=r" (__r0), "=r" (__r1) \ - : "r" (__r0), "r" (__r1), "r" (d)); \ - (q) = __r1; (r) = __r0; \ - } while (0) -#endif /* if 0 */ -#else /* if __zarch__ */ -/* FIXME: this fails if gcc knows about the 64-bit registers. */ -#define smul_ppmm(xh, xl, m0, m1) \ - do { \ - union {DItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __asm__ ("mr\t%0,%2" \ - : "=r" (__x.__ll) \ - : "%0" (m0), "r" (m1)); \ - (xh) = __x.__i.__h; (xl) = __x.__i.__l; \ - } while (0) -/* FIXME: this fails if gcc knows about the 64-bit registers. */ -#define sdiv_qrnnd(q, r, n1, n0, d) \ - do { \ - union {DItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __x.__i.__h = n1; __x.__i.__l = n0; \ - __asm__ ("dr\t%0,%2" \ - : "=r" (__x.__ll) \ - : "0" (__x.__ll), "r" (d)); \ - (q) = __x.__i.__l; (r) = __x.__i.__h; \ - } while (0) -#endif /* if __zarch__ */ -#endif - -#if defined (__s390x__) && W_TYPE_SIZE == 64 -/* We need to cast operands with register constraints, otherwise their types - will be assumed to be SImode by gcc. For these machines, such operations - will insert a value into the low 32 bits, and leave the high 32 bits with - garbage. */ -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - do { \ - __asm__ ("algr\t%1,%5\n\talcgr\t%0,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((UDItype)(ah)), "r" ((UDItype)(bh)), \ - "%1" ((UDItype)(al)), "r" ((UDItype)(bl)) __CLOBBER_CC); \ - } while (0) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - do { \ - __asm__ ("slgr\t%1,%5\n\tslbgr\t%0,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((UDItype)(ah)), "r" ((UDItype)(bh)), \ - "1" ((UDItype)(al)), "r" ((UDItype)(bl)) __CLOBBER_CC); \ - } while (0) -#define umul_ppmm(xh, xl, m0, m1) \ - do { \ - union {unsigned int __attribute__ ((mode(TI))) __ll; \ - struct {UDItype __h, __l;} __i; \ - } __x; \ - __asm__ ("mlgr\t%0,%2" \ - : "=r" (__x.__ll) \ - : "%0" ((UDItype)(m0)), "r" ((UDItype)(m1))); \ - (xh) = __x.__i.__h; (xl) = __x.__i.__l; \ - } while (0) -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { \ - union {unsigned int __attribute__ ((mode(TI))) __ll; \ - struct {UDItype __h, __l;} __i; \ - } __x; \ - __x.__i.__h = n1; __x.__i.__l = n0; \ - __asm__ ("dlgr\t%0,%2" \ - : "=r" (__x.__ll) \ - : "0" (__x.__ll), "r" ((UDItype)(d))); \ - (q) = __x.__i.__l; (r) = __x.__i.__h; \ - } while (0) -#if 0 /* FIXME: Enable for z10 (?) */ -#define count_leading_zeros(cnt, x) \ - do { \ - union {unsigned int __attribute__ ((mode(TI))) __ll; \ - struct {UDItype __h, __l;} __i; \ - } __clr_cnt; \ - __asm__ ("flogr\t%0,%1" \ - : "=r" (__clr_cnt.__ll) \ - : "r" (x) __CLOBBER_CC); \ - (cnt) = __clr_cnt.__i.__h; \ - } while (0) -#endif -#endif - -/* On x86 and x86_64, every asm implicitly clobbers "flags" and "fpsr", - so we don't need __CLOBBER_CC. */ -#if (defined (__i386__) || defined (__i486__)) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("addl %5,%k1\n\tadcl %3,%k0" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((USItype)(ah)), "g" ((USItype)(bh)), \ - "%1" ((USItype)(al)), "g" ((USItype)(bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("subl %5,%k1\n\tsbbl %3,%k0" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((USItype)(ah)), "g" ((USItype)(bh)), \ - "1" ((USItype)(al)), "g" ((USItype)(bl))) -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("mull %3" \ - : "=a" (w0), "=d" (w1) \ - : "%0" ((USItype)(u)), "rm" ((USItype)(v))) -#define udiv_qrnnd(q, r, n1, n0, dx) /* d renamed to dx avoiding "=d" */\ - __asm__ ("divl %4" /* stringification in K&R C */ \ - : "=a" (q), "=d" (r) \ - : "0" ((USItype)(n0)), "1" ((USItype)(n1)), "rm" ((USItype)(dx))) - -#if HAVE_HOST_CPU_i586 || HAVE_HOST_CPU_pentium || HAVE_HOST_CPU_pentiummmx -/* Pentium bsrl takes between 10 and 72 cycles depending where the most - significant 1 bit is, hence the use of the following alternatives. bsfl - is slow too, between 18 and 42 depending where the least significant 1 - bit is, so let the generic count_trailing_zeros below make use of the - count_leading_zeros here too. */ - -#if HAVE_HOST_CPU_pentiummmx && ! defined (LONGLONG_STANDALONE) -/* The following should be a fixed 14 or 15 cycles, but possibly plus an L1 - cache miss reading from __clz_tab. For P55 it's favoured over the float - below so as to avoid mixing MMX and x87, since the penalty for switching - between the two is about 100 cycles. - - The asm block sets __shift to -3 if the high 24 bits are clear, -2 for - 16, -1 for 8, or 0 otherwise. This could be written equivalently as - follows, but as of gcc 2.95.2 it results in conditional jumps. - - __shift = -(__n < 0x1000000); - __shift -= (__n < 0x10000); - __shift -= (__n < 0x100); - - The middle two sbbl and cmpl's pair, and with luck something gcc - generates might pair with the first cmpl and the last sbbl. The "32+1" - constant could be folded into __clz_tab[], but it doesn't seem worth - making a different table just for that. */ - -#define count_leading_zeros(c,n) \ - do { \ - USItype __n = (n); \ - USItype __shift; \ - __asm__ ("cmpl $0x1000000, %1\n" \ - "sbbl %0, %0\n" \ - "cmpl $0x10000, %1\n" \ - "sbbl $0, %0\n" \ - "cmpl $0x100, %1\n" \ - "sbbl $0, %0\n" \ - : "=&r" (__shift) : "r" (__n)); \ - __shift = __shift*8 + 24 + 1; \ - (c) = 32 + 1 - __shift - __clz_tab[__n >> __shift]; \ - } while (0) -#define COUNT_LEADING_ZEROS_NEED_CLZ_TAB -#define COUNT_LEADING_ZEROS_0 31 /* n==0 indistinguishable from n==1 */ - -#else /* ! pentiummmx || LONGLONG_STANDALONE */ -/* The following should be a fixed 14 cycles or so. Some scheduling - opportunities should be available between the float load/store too. This - sort of code is used in gcc 3 for __builtin_ffs (with "n&-n") and is - apparently suggested by the Intel optimizing manual (don't know exactly - where). gcc 2.95 or up will be best for this, so the "double" is - correctly aligned on the stack. */ -#define count_leading_zeros(c,n) \ - do { \ - union { \ - double d; \ - unsigned a[2]; \ - } __u; \ - __u.d = (UWtype) (n); \ - (c) = 0x3FF + 31 - (__u.a[1] >> 20); \ - } while (0) -#define COUNT_LEADING_ZEROS_0 (0x3FF + 31) -#endif /* pentiummx */ - -#else /* ! pentium */ - -#if __GMP_GNUC_PREREQ (3,4) /* using bsrl */ -#define count_leading_zeros(count,x) count_leading_zeros_gcc_clz(count,x) -#endif /* gcc clz */ - -/* On P6, gcc prior to 3.0 generates a partial register stall for - __cbtmp^31, due to using "xorb $31" instead of "xorl $31", the former - being 1 code byte smaller. "31-__cbtmp" is a workaround, probably at the - cost of one extra instruction. Do this for "i386" too, since that means - generic x86. */ -#if ! defined (count_leading_zeros) && __GNUC__ < 3 \ - && (HAVE_HOST_CPU_i386 \ - || HAVE_HOST_CPU_i686 \ - || HAVE_HOST_CPU_pentiumpro \ - || HAVE_HOST_CPU_pentium2 \ - || HAVE_HOST_CPU_pentium3) -#define count_leading_zeros(count, x) \ - do { \ - USItype __cbtmp; \ - ASSERT ((x) != 0); \ - __asm__ ("bsrl %1,%0" : "=r" (__cbtmp) : "rm" ((USItype)(x))); \ - (count) = 31 - __cbtmp; \ - } while (0) -#endif /* gcc<3 asm bsrl */ - -#ifndef count_leading_zeros -#define count_leading_zeros(count, x) \ - do { \ - USItype __cbtmp; \ - ASSERT ((x) != 0); \ - __asm__ ("bsrl %1,%0" : "=r" (__cbtmp) : "rm" ((USItype)(x))); \ - (count) = __cbtmp ^ 31; \ - } while (0) -#endif /* asm bsrl */ - -#if __GMP_GNUC_PREREQ (3,4) /* using bsfl */ -#define count_trailing_zeros(count,x) count_trailing_zeros_gcc_ctz(count,x) -#endif /* gcc ctz */ - -#ifndef count_trailing_zeros -#define count_trailing_zeros(count, x) \ - do { \ - ASSERT ((x) != 0); \ - __asm__ ("bsfl %1,%k0" : "=r" (count) : "rm" ((USItype)(x))); \ - } while (0) -#endif /* asm bsfl */ - -#endif /* ! pentium */ - -#endif /* 80x86 */ - -#if defined (__amd64__) && W_TYPE_SIZE == 64 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("addq %5,%q1\n\tadcq %3,%q0" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((UDItype)(ah)), "rme" ((UDItype)(bh)), \ - "%1" ((UDItype)(al)), "rme" ((UDItype)(bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("subq %5,%q1\n\tsbbq %3,%q0" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((UDItype)(ah)), "rme" ((UDItype)(bh)), \ - "1" ((UDItype)(al)), "rme" ((UDItype)(bl))) -#if X86_ASM_MULX \ - && (HAVE_HOST_CPU_haswell || HAVE_HOST_CPU_broadwell \ - || HAVE_HOST_CPU_skylake || HAVE_HOST_CPU_bd4 || HAVE_HOST_CPU_zen) -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("mulx\t%3, %0, %1" \ - : "=r" (w0), "=r" (w1) \ - : "%d" ((UDItype)(u)), "rm" ((UDItype)(v))) -#else -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("mulq\t%3" \ - : "=a" (w0), "=d" (w1) \ - : "%0" ((UDItype)(u)), "rm" ((UDItype)(v))) -#endif -#define udiv_qrnnd(q, r, n1, n0, dx) /* d renamed to dx avoiding "=d" */\ - __asm__ ("divq %4" /* stringification in K&R C */ \ - : "=a" (q), "=d" (r) \ - : "0" ((UDItype)(n0)), "1" ((UDItype)(n1)), "rm" ((UDItype)(dx))) - -#if HAVE_HOST_CPU_haswell || HAVE_HOST_CPU_broadwell || HAVE_HOST_CPU_skylake \ - || HAVE_HOST_CPU_k10 || HAVE_HOST_CPU_bd1 || HAVE_HOST_CPU_bd2 \ - || HAVE_HOST_CPU_bd3 || HAVE_HOST_CPU_bd4 || HAVE_HOST_CPU_zen \ - || HAVE_HOST_CPU_bobcat || HAVE_HOST_CPU_jaguar -#define count_leading_zeros(count, x) \ - do { \ - /* This is lzcnt, spelled for older assemblers. Destination and */ \ - /* source must be a 64-bit registers, hence cast and %q. */ \ - __asm__ ("rep;bsr\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \ - } while (0) -#define COUNT_LEADING_ZEROS_0 64 -#else -#define count_leading_zeros(count, x) \ - do { \ - UDItype __cbtmp; \ - ASSERT ((x) != 0); \ - __asm__ ("bsr\t%1,%0" : "=r" (__cbtmp) : "rm" ((UDItype)(x))); \ - (count) = __cbtmp ^ 63; \ - } while (0) -#endif - -#if HAVE_HOST_CPU_bd2 || HAVE_HOST_CPU_bd3 || HAVE_HOST_CPU_bd4 \ - || HAVE_HOST_CPU_zen || HAVE_HOST_CPU_jaguar -#define count_trailing_zeros(count, x) \ - do { \ - /* This is tzcnt, spelled for older assemblers. Destination and */ \ - /* source must be a 64-bit registers, hence cast and %q. */ \ - __asm__ ("rep;bsf\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \ - } while (0) -#define COUNT_TRAILING_ZEROS_0 64 -#else -#define count_trailing_zeros(count, x) \ - do { \ - ASSERT ((x) != 0); \ - __asm__ ("bsf\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \ - } while (0) -#endif -#endif /* __amd64__ */ - -#if defined (__i860__) && W_TYPE_SIZE == 32 -#define rshift_rhlc(r,h,l,c) \ - __asm__ ("shr %3,r0,r0\;shrd %1,%2,%0" \ - "=r" (r) : "r" (h), "r" (l), "rn" (c)) -#endif /* i860 */ - -#if defined (__i960__) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("cmpo 1,0\;addc %5,%4,%1\;addc %3,%2,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "dI" (ah), "dI" (bh), "%dI" (al), "dI" (bl)) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("cmpo 0,0\;subc %5,%4,%1\;subc %3,%2,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "dI" (ah), "dI" (bh), "dI" (al), "dI" (bl)) -#define umul_ppmm(w1, w0, u, v) \ - ({union {UDItype __ll; \ - struct {USItype __l, __h;} __i; \ - } __x; \ - __asm__ ("emul %2,%1,%0" \ - : "=d" (__x.__ll) : "%dI" (u), "dI" (v)); \ - (w1) = __x.__i.__h; (w0) = __x.__i.__l;}) -#define __umulsidi3(u, v) \ - ({UDItype __w; \ - __asm__ ("emul %2,%1,%0" : "=d" (__w) : "%dI" (u), "dI" (v)); \ - __w; }) -#define udiv_qrnnd(q, r, nh, nl, d) \ - do { \ - union {UDItype __ll; \ - struct {USItype __l, __h;} __i; \ - } __nn; \ - __nn.__i.__h = (nh); __nn.__i.__l = (nl); \ - __asm__ ("ediv %d,%n,%0" \ - : "=d" (__rq.__ll) : "dI" (__nn.__ll), "dI" (d)); \ - (r) = __rq.__i.__l; (q) = __rq.__i.__h; \ - } while (0) -#define count_leading_zeros(count, x) \ - do { \ - USItype __cbtmp; \ - __asm__ ("scanbit %1,%0" : "=r" (__cbtmp) : "r" (x)); \ - (count) = __cbtmp ^ 31; \ - } while (0) -#define COUNT_LEADING_ZEROS_0 (-32) /* sic */ -#if defined (__i960mx) /* what is the proper symbol to test??? */ -#define rshift_rhlc(r,h,l,c) \ - do { \ - union {UDItype __ll; \ - struct {USItype __l, __h;} __i; \ - } __nn; \ - __nn.__i.__h = (h); __nn.__i.__l = (l); \ - __asm__ ("shre %2,%1,%0" : "=d" (r) : "dI" (__nn.__ll), "dI" (c)); \ - } -#endif /* i960mx */ -#endif /* i960 */ - -#if (defined (__mc68000__) || defined (__mc68020__) || defined(mc68020) \ - || defined (__m68k__) || defined (__mc5200__) || defined (__mc5206e__) \ - || defined (__mc5307__)) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("add%.l %5,%1\n\taddx%.l %3,%0" \ - : "=d" (sh), "=&d" (sl) \ - : "0" ((USItype)(ah)), "d" ((USItype)(bh)), \ - "%1" ((USItype)(al)), "g" ((USItype)(bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("sub%.l %5,%1\n\tsubx%.l %3,%0" \ - : "=d" (sh), "=&d" (sl) \ - : "0" ((USItype)(ah)), "d" ((USItype)(bh)), \ - "1" ((USItype)(al)), "g" ((USItype)(bl))) -/* The '020, '030, '040 and CPU32 have 32x32->64 and 64/32->32q-32r. */ -#if defined (__mc68020__) || defined(mc68020) \ - || defined (__mc68030__) || defined (mc68030) \ - || defined (__mc68040__) || defined (mc68040) \ - || defined (__mcpu32__) || defined (mcpu32) \ - || defined (__NeXT__) -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("mulu%.l %3,%1:%0" \ - : "=d" (w0), "=d" (w1) \ - : "%0" ((USItype)(u)), "dmi" ((USItype)(v))) -#define udiv_qrnnd(q, r, n1, n0, d) \ - __asm__ ("divu%.l %4,%1:%0" \ - : "=d" (q), "=d" (r) \ - : "0" ((USItype)(n0)), "1" ((USItype)(n1)), "dmi" ((USItype)(d))) -#define sdiv_qrnnd(q, r, n1, n0, d) \ - __asm__ ("divs%.l %4,%1:%0" \ - : "=d" (q), "=d" (r) \ - : "0" ((USItype)(n0)), "1" ((USItype)(n1)), "dmi" ((USItype)(d))) -#else /* for other 68k family members use 16x16->32 multiplication */ -#define umul_ppmm(xh, xl, a, b) \ - do { USItype __umul_tmp1, __umul_tmp2; \ - __asm__ ("| Inlined umul_ppmm\n" \ -" move%.l %5,%3\n" \ -" move%.l %2,%0\n" \ -" move%.w %3,%1\n" \ -" swap %3\n" \ -" swap %0\n" \ -" mulu%.w %2,%1\n" \ -" mulu%.w %3,%0\n" \ -" mulu%.w %2,%3\n" \ -" swap %2\n" \ -" mulu%.w %5,%2\n" \ -" add%.l %3,%2\n" \ -" jcc 1f\n" \ -" add%.l %#0x10000,%0\n" \ -"1: move%.l %2,%3\n" \ -" clr%.w %2\n" \ -" swap %2\n" \ -" swap %3\n" \ -" clr%.w %3\n" \ -" add%.l %3,%1\n" \ -" addx%.l %2,%0\n" \ -" | End inlined umul_ppmm" \ - : "=&d" (xh), "=&d" (xl), \ - "=d" (__umul_tmp1), "=&d" (__umul_tmp2) \ - : "%2" ((USItype)(a)), "d" ((USItype)(b))); \ - } while (0) -#endif /* not mc68020 */ -/* The '020, '030, '040 and '060 have bitfield insns. - GCC 3.4 defines __mc68020__ when in CPU32 mode, check for __mcpu32__ to - exclude bfffo on that chip (bitfield insns not available). */ -#if (defined (__mc68020__) || defined (mc68020) \ - || defined (__mc68030__) || defined (mc68030) \ - || defined (__mc68040__) || defined (mc68040) \ - || defined (__mc68060__) || defined (mc68060) \ - || defined (__NeXT__)) \ - && ! defined (__mcpu32__) -#define count_leading_zeros(count, x) \ - __asm__ ("bfffo %1{%b2:%b2},%0" \ - : "=d" (count) \ - : "od" ((USItype) (x)), "n" (0)) -#define COUNT_LEADING_ZEROS_0 32 -#endif -#endif /* mc68000 */ - -#if defined (__m88000__) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("addu.co %1,%r4,%r5\n\taddu.ci %0,%r2,%r3" \ - : "=r" (sh), "=&r" (sl) \ - : "rJ" (ah), "rJ" (bh), "%rJ" (al), "rJ" (bl)) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("subu.co %1,%r4,%r5\n\tsubu.ci %0,%r2,%r3" \ - : "=r" (sh), "=&r" (sl) \ - : "rJ" (ah), "rJ" (bh), "rJ" (al), "rJ" (bl)) -#define count_leading_zeros(count, x) \ - do { \ - USItype __cbtmp; \ - __asm__ ("ff1 %0,%1" : "=r" (__cbtmp) : "r" (x)); \ - (count) = __cbtmp ^ 31; \ - } while (0) -#define COUNT_LEADING_ZEROS_0 63 /* sic */ -#if defined (__m88110__) -#define umul_ppmm(wh, wl, u, v) \ - do { \ - union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __asm__ ("mulu.d %0,%1,%2" : "=r" (__x.__ll) : "r" (u), "r" (v)); \ - (wh) = __x.__i.__h; \ - (wl) = __x.__i.__l; \ - } while (0) -#define udiv_qrnnd(q, r, n1, n0, d) \ - ({union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x, __q; \ - __x.__i.__h = (n1); __x.__i.__l = (n0); \ - __asm__ ("divu.d %0,%1,%2" \ - : "=r" (__q.__ll) : "r" (__x.__ll), "r" (d)); \ - (r) = (n0) - __q.__l * (d); (q) = __q.__l; }) -#endif /* __m88110__ */ -#endif /* __m88000__ */ - -#if defined (__mips) && W_TYPE_SIZE == 32 -#if __GMP_GNUC_PREREQ (4,4) -#define umul_ppmm(w1, w0, u, v) \ - do { \ - UDItype __ll = (UDItype)(u) * (v); \ - w1 = __ll >> 32; \ - w0 = __ll; \ - } while (0) -#endif -#if !defined (umul_ppmm) && __GMP_GNUC_PREREQ (2,7) && !defined (__clang__) -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("multu %2,%3" : "=l" (w0), "=h" (w1) : "d" (u), "d" (v)) -#endif -#if !defined (umul_ppmm) -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("multu %2,%3\n\tmflo %0\n\tmfhi %1" \ - : "=d" (w0), "=d" (w1) : "d" (u), "d" (v)) -#endif -#endif /* __mips */ - -#if (defined (__mips) && __mips >= 3) && W_TYPE_SIZE == 64 -#if defined (_MIPS_ARCH_MIPS64R6) -#define umul_ppmm(w1, w0, u, v) \ - do { \ - UDItype __m0 = (u), __m1 = (v); \ - (w0) = __m0 * __m1; \ - __asm__ ("dmuhu\t%0, %1, %2" : "=d" (w1) : "d" (__m0), "d" (__m1)); \ - } while (0) -#endif -#if !defined (umul_ppmm) && __GMP_GNUC_PREREQ (4,4) -#define umul_ppmm(w1, w0, u, v) \ - do { \ - typedef unsigned int __ll_UTItype __attribute__((mode(TI))); \ - __ll_UTItype __ll = (__ll_UTItype)(u) * (v); \ - w1 = __ll >> 64; \ - w0 = __ll; \ - } while (0) -#endif -#if !defined (umul_ppmm) && __GMP_GNUC_PREREQ (2,7) && !defined (__clang__) -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("dmultu %2,%3" \ - : "=l" (w0), "=h" (w1) \ - : "d" ((UDItype)(u)), "d" ((UDItype)(v))) -#endif -#if !defined (umul_ppmm) -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("dmultu %2,%3\n\tmflo %0\n\tmfhi %1" \ - : "=d" (w0), "=d" (w1) \ - : "d" ((UDItype)(u)), "d" ((UDItype)(v))) -#endif -#endif /* __mips */ - -#if defined (__mmix__) && W_TYPE_SIZE == 64 -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("MULU %0,%2,%3" : "=r" (w0), "=z" (w1) : "r" (u), "r" (v)) -#endif - -#if defined (__ns32000__) && W_TYPE_SIZE == 32 -#define umul_ppmm(w1, w0, u, v) \ - ({union {UDItype __ll; \ - struct {USItype __l, __h;} __i; \ - } __x; \ - __asm__ ("meid %2,%0" \ - : "=g" (__x.__ll) \ - : "%0" ((USItype)(u)), "g" ((USItype)(v))); \ - (w1) = __x.__i.__h; (w0) = __x.__i.__l;}) -#define __umulsidi3(u, v) \ - ({UDItype __w; \ - __asm__ ("meid %2,%0" \ - : "=g" (__w) \ - : "%0" ((USItype)(u)), "g" ((USItype)(v))); \ - __w; }) -#define udiv_qrnnd(q, r, n1, n0, d) \ - ({union {UDItype __ll; \ - struct {USItype __l, __h;} __i; \ - } __x; \ - __x.__i.__h = (n1); __x.__i.__l = (n0); \ - __asm__ ("deid %2,%0" \ - : "=g" (__x.__ll) \ - : "0" (__x.__ll), "g" ((USItype)(d))); \ - (r) = __x.__i.__l; (q) = __x.__i.__h; }) -#define count_trailing_zeros(count,x) \ - do { \ - __asm__ ("ffsd %2,%0" \ - : "=r" (count) \ - : "0" ((USItype) 0), "r" ((USItype) (x))); \ - } while (0) -#endif /* __ns32000__ */ - -/* In the past we had a block of various #defines tested - _ARCH_PPC - AIX - _ARCH_PWR - AIX - __powerpc__ - gcc - __POWERPC__ - BEOS - __ppc__ - Darwin - PPC - old gcc, GNU/Linux, SysV - The plain PPC test was not good for vxWorks, since PPC is defined on all - CPUs there (eg. m68k too), as a constant one is expected to compare - CPU_FAMILY against. - - At any rate, this was pretty unattractive and a bit fragile. The use of - HAVE_HOST_CPU_FAMILY is designed to cut through it all and be sure of - getting the desired effect. - - ENHANCE-ME: We should test _IBMR2 here when we add assembly support for - the system vendor compilers. (Is that vendor compilers with inline asm, - or what?) */ - -#if (HAVE_HOST_CPU_FAMILY_power || HAVE_HOST_CPU_FAMILY_powerpc) \ - && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (bh) && (bh) == 0) \ - __asm__ ("add%I4c %1,%3,%4\n\taddze %0,%2" \ - : "=r" (sh), "=&r" (sl) : "r" (ah), "%r" (al), "rI" (bl) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (bh) && (bh) == ~(USItype) 0) \ - __asm__ ("add%I4c %1,%3,%4\n\taddme %0,%2" \ - : "=r" (sh), "=&r" (sl) : "r" (ah), "%r" (al), "rI" (bl) \ - __CLOBBER_CC); \ - else \ - __asm__ ("add%I5c %1,%4,%5\n\tadde %0,%2,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "r" (bh), "%r" (al), "rI" (bl) \ - __CLOBBER_CC); \ - } while (0) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (ah) && (ah) == 0) \ - __asm__ ("subf%I3c %1,%4,%3\n\tsubfze %0,%2" \ - : "=r" (sh), "=&r" (sl) : "r" (bh), "rI" (al), "r" (bl) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (ah) && (ah) == ~(USItype) 0) \ - __asm__ ("subf%I3c %1,%4,%3\n\tsubfme %0,%2" \ - : "=r" (sh), "=&r" (sl) : "r" (bh), "rI" (al), "r" (bl) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (bh) && (bh) == 0) \ - __asm__ ("subf%I3c %1,%4,%3\n\taddme %0,%2" \ - : "=r" (sh), "=&r" (sl) : "r" (ah), "rI" (al), "r" (bl) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (bh) && (bh) == ~(USItype) 0) \ - __asm__ ("subf%I3c %1,%4,%3\n\taddze %0,%2" \ - : "=r" (sh), "=&r" (sl) : "r" (ah), "rI" (al), "r" (bl) \ - __CLOBBER_CC); \ - else \ - __asm__ ("subf%I4c %1,%5,%4\n\tsubfe %0,%3,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" (ah), "r" (bh), "rI" (al), "r" (bl) \ - __CLOBBER_CC); \ - } while (0) -#define count_leading_zeros(count, x) \ - __asm__ ("cntlzw %0,%1" : "=r" (count) : "r" (x)) -#define COUNT_LEADING_ZEROS_0 32 -#if HAVE_HOST_CPU_FAMILY_powerpc -#if __GMP_GNUC_PREREQ (4,4) -#define umul_ppmm(w1, w0, u, v) \ - do { \ - UDItype __ll = (UDItype)(u) * (v); \ - w1 = __ll >> 32; \ - w0 = __ll; \ - } while (0) -#endif -#if !defined (umul_ppmm) -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - USItype __m0 = (m0), __m1 = (m1); \ - __asm__ ("mulhwu %0,%1,%2" : "=r" (ph) : "%r" (m0), "r" (m1)); \ - (pl) = __m0 * __m1; \ - } while (0) -#endif -#define smul_ppmm(ph, pl, m0, m1) \ - do { \ - SItype __m0 = (m0), __m1 = (m1); \ - __asm__ ("mulhw %0,%1,%2" : "=r" (ph) : "%r" (m0), "r" (m1)); \ - (pl) = __m0 * __m1; \ - } while (0) -#else -#define smul_ppmm(xh, xl, m0, m1) \ - __asm__ ("mul %0,%2,%3" : "=r" (xh), "=q" (xl) : "r" (m0), "r" (m1)) -#define sdiv_qrnnd(q, r, nh, nl, d) \ - __asm__ ("div %0,%2,%4" : "=r" (q), "=q" (r) : "r" (nh), "1" (nl), "r" (d)) -#endif -#endif /* 32-bit POWER architecture variants. */ - -/* We should test _IBMR2 here when we add assembly support for the system - vendor compilers. */ -#if HAVE_HOST_CPU_FAMILY_powerpc && W_TYPE_SIZE == 64 -#if !defined (_LONG_LONG_LIMB) -/* _LONG_LONG_LIMB is ABI=mode32 where adde operates on 32-bit values. So - use adde etc only when not _LONG_LONG_LIMB. */ -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (bh) && (bh) == 0) \ - __asm__ ("add%I4c %1,%3,%4\n\taddze %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), \ - "%r" ((UDItype)(al)), "rI" ((UDItype)(bl)) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (bh) && (bh) == ~(UDItype) 0) \ - __asm__ ("add%I4c %1,%3,%4\n\taddme %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), \ - "%r" ((UDItype)(al)), "rI" ((UDItype)(bl)) \ - __CLOBBER_CC); \ - else \ - __asm__ ("add%I5c %1,%4,%5\n\tadde %0,%2,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), "r" ((UDItype)(bh)), \ - "%r" ((UDItype)(al)), "rI" ((UDItype)(bl)) \ - __CLOBBER_CC); \ - } while (0) -/* We use "*rI" for the constant operand here, since with just "I", gcc barfs. - This might seem strange, but gcc folds away the dead code late. */ -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - do { \ - if (__builtin_constant_p (bl) && bl > -0x8000 && bl <= 0x8000) { \ - if (__builtin_constant_p (ah) && (ah) == 0) \ - __asm__ ("addic %1,%3,%4\n\tsubfze %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(bh)), \ - "rI" ((UDItype)(al)), "*rI" (-((UDItype)(bl))) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (ah) && (ah) == ~(UDItype) 0) \ - __asm__ ("addic %1,%3,%4\n\tsubfme %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(bh)), \ - "rI" ((UDItype)(al)), "*rI" (-((UDItype)(bl))) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (bh) && (bh) == 0) \ - __asm__ ("addic %1,%3,%4\n\taddme %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), \ - "rI" ((UDItype)(al)), "*rI" (-((UDItype)(bl))) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (bh) && (bh) == ~(UDItype) 0) \ - __asm__ ("addic %1,%3,%4\n\taddze %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), \ - "rI" ((UDItype)(al)), "*rI" (-((UDItype)(bl))) \ - __CLOBBER_CC); \ - else \ - __asm__ ("addic %1,%4,%5\n\tsubfe %0,%3,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), "r" ((UDItype)(bh)), \ - "rI" ((UDItype)(al)), "*rI" (-((UDItype)(bl))) \ - __CLOBBER_CC); \ - } else { \ - if (__builtin_constant_p (ah) && (ah) == 0) \ - __asm__ ("subf%I3c %1,%4,%3\n\tsubfze %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(bh)), \ - "rI" ((UDItype)(al)), "r" ((UDItype)(bl)) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (ah) && (ah) == ~(UDItype) 0) \ - __asm__ ("subf%I3c %1,%4,%3\n\tsubfme %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(bh)), \ - "rI" ((UDItype)(al)), "r" ((UDItype)(bl)) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (bh) && (bh) == 0) \ - __asm__ ("subf%I3c %1,%4,%3\n\taddme %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), \ - "rI" ((UDItype)(al)), "r" ((UDItype)(bl)) \ - __CLOBBER_CC); \ - else if (__builtin_constant_p (bh) && (bh) == ~(UDItype) 0) \ - __asm__ ("subf%I3c %1,%4,%3\n\taddze %0,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), \ - "rI" ((UDItype)(al)), "r" ((UDItype)(bl)) \ - __CLOBBER_CC); \ - else \ - __asm__ ("subf%I4c %1,%5,%4\n\tsubfe %0,%3,%2" \ - : "=r" (sh), "=&r" (sl) \ - : "r" ((UDItype)(ah)), "r" ((UDItype)(bh)), \ - "rI" ((UDItype)(al)), "r" ((UDItype)(bl)) \ - __CLOBBER_CC); \ - } \ - } while (0) -#endif /* ! _LONG_LONG_LIMB */ -#define count_leading_zeros(count, x) \ - __asm__ ("cntlzd %0,%1" : "=r" (count) : "r" (x)) -#define COUNT_LEADING_ZEROS_0 64 -#if __GMP_GNUC_PREREQ (4,8) -#define umul_ppmm(w1, w0, u, v) \ - do { \ - typedef unsigned int __ll_UTItype __attribute__((mode(TI))); \ - __ll_UTItype __ll = (__ll_UTItype)(u) * (v); \ - w1 = __ll >> 64; \ - w0 = __ll; \ - } while (0) -#endif -#if !defined (umul_ppmm) -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UDItype __m0 = (m0), __m1 = (m1); \ - __asm__ ("mulhdu %0,%1,%2" : "=r" (ph) : "%r" (__m0), "r" (__m1)); \ - (pl) = __m0 * __m1; \ - } while (0) -#endif -#define smul_ppmm(ph, pl, m0, m1) \ - do { \ - DItype __m0 = (m0), __m1 = (m1); \ - __asm__ ("mulhd %0,%1,%2" : "=r" (ph) : "%r" (__m0), "r" (__m1)); \ - (pl) = __m0 * __m1; \ - } while (0) -#endif /* 64-bit PowerPC. */ - -#if defined (__pyr__) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("addw %5,%1\n\taddwc %3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((USItype)(ah)), "g" ((USItype)(bh)), \ - "%1" ((USItype)(al)), "g" ((USItype)(bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("subw %5,%1\n\tsubwb %3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((USItype)(ah)), "g" ((USItype)(bh)), \ - "1" ((USItype)(al)), "g" ((USItype)(bl))) -/* This insn works on Pyramids with AP, XP, or MI CPUs, but not with SP. */ -#define umul_ppmm(w1, w0, u, v) \ - ({union {UDItype __ll; \ - struct {USItype __h, __l;} __i; \ - } __x; \ - __asm__ ("movw %1,%R0\n\tuemul %2,%0" \ - : "=&r" (__x.__ll) \ - : "g" ((USItype) (u)), "g" ((USItype)(v))); \ - (w1) = __x.__i.__h; (w0) = __x.__i.__l;}) -#endif /* __pyr__ */ - -#if defined (__ibm032__) /* RT/ROMP */ && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("a %1,%5\n\tae %0,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((USItype)(ah)), "r" ((USItype)(bh)), \ - "%1" ((USItype)(al)), "r" ((USItype)(bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("s %1,%5\n\tse %0,%3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((USItype)(ah)), "r" ((USItype)(bh)), \ - "1" ((USItype)(al)), "r" ((USItype)(bl))) -#define smul_ppmm(ph, pl, m0, m1) \ - __asm__ ( \ - "s r2,r2\n" \ -" mts r10,%2\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" m r2,%3\n" \ -" cas %0,r2,r0\n" \ -" mfs r10,%1" \ - : "=r" (ph), "=r" (pl) \ - : "%r" ((USItype)(m0)), "r" ((USItype)(m1)) \ - : "r2") -#define count_leading_zeros(count, x) \ - do { \ - if ((x) >= 0x10000) \ - __asm__ ("clz %0,%1" \ - : "=r" (count) : "r" ((USItype)(x) >> 16)); \ - else \ - { \ - __asm__ ("clz %0,%1" \ - : "=r" (count) : "r" ((USItype)(x))); \ - (count) += 16; \ - } \ - } while (0) -#endif /* RT/ROMP */ - -#if defined (__riscv64) && W_TYPE_SIZE == 64 -#define umul_ppmm(ph, pl, u, v) \ - do { \ - UDItype __u = (u), __v = (v); \ - (pl) = __u * __v; \ - __asm__ ("mulhu\t%2, %1, %0" : "=r" (ph) : "%r" (__u), "r" (__v)); \ - } while (0) -#endif - -#if (defined (__SH2__) || defined (__SH3__) || defined (__SH4__)) && W_TYPE_SIZE == 32 -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("dmulu.l %2,%3\n\tsts macl,%1\n\tsts mach,%0" \ - : "=r" (w1), "=r" (w0) : "r" (u), "r" (v) : "macl", "mach") -#endif - -#if defined (__sparc__) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("addcc %r4,%5,%1\n\taddx %r2,%3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "rJ" (ah), "rI" (bh),"%rJ" (al), "rI" (bl) \ - __CLOBBER_CC) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("subcc %r4,%5,%1\n\tsubx %r2,%3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "rJ" (ah), "rI" (bh), "rJ" (al), "rI" (bl) \ - __CLOBBER_CC) -/* FIXME: When gcc -mcpu=v9 is used on solaris, gcc/config/sol2-sld-64.h - doesn't define anything to indicate that to us, it only sets __sparcv8. */ -#if defined (__sparc_v9__) || defined (__sparcv9) -/* Perhaps we should use floating-point operations here? */ -#if 0 -/* Triggers a bug making mpz/tests/t-gcd.c fail. - Perhaps we simply need explicitly zero-extend the inputs? */ -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("mulx %2,%3,%%g1; srl %%g1,0,%1; srlx %%g1,32,%0" : \ - "=r" (w1), "=r" (w0) : "r" (u), "r" (v) : "g1") -#else -/* Use v8 umul until above bug is fixed. */ -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("umul %2,%3,%1;rd %%y,%0" : "=r" (w1), "=r" (w0) : "r" (u), "r" (v)) -#endif -/* Use a plain v8 divide for v9. */ -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { \ - USItype __q; \ - __asm__ ("mov %1,%%y;nop;nop;nop;udiv %2,%3,%0" \ - : "=r" (__q) : "r" (n1), "r" (n0), "r" (d)); \ - (r) = (n0) - __q * (d); \ - (q) = __q; \ - } while (0) -#else -#if defined (__sparc_v8__) /* gcc normal */ \ - || defined (__sparcv8) /* gcc solaris */ \ - || HAVE_HOST_CPU_supersparc -/* Don't match immediate range because, 1) it is not often useful, - 2) the 'I' flag thinks of the range as a 13 bit signed interval, - while we want to match a 13 bit interval, sign extended to 32 bits, - but INTERPRETED AS UNSIGNED. */ -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("umul %2,%3,%1;rd %%y,%0" : "=r" (w1), "=r" (w0) : "r" (u), "r" (v)) - -#if HAVE_HOST_CPU_supersparc -#else -/* Don't use this on SuperSPARC because its udiv only handles 53 bit - dividends and will trap to the kernel for the rest. */ -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { \ - USItype __q; \ - __asm__ ("mov %1,%%y;nop;nop;nop;udiv %2,%3,%0" \ - : "=r" (__q) : "r" (n1), "r" (n0), "r" (d)); \ - (r) = (n0) - __q * (d); \ - (q) = __q; \ - } while (0) -#endif /* HAVE_HOST_CPU_supersparc */ - -#else /* ! __sparc_v8__ */ -#if defined (__sparclite__) -/* This has hardware multiply but not divide. It also has two additional - instructions scan (ffs from high bit) and divscc. */ -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("umul %2,%3,%1;rd %%y,%0" : "=r" (w1), "=r" (w0) : "r" (u), "r" (v)) -#define udiv_qrnnd(q, r, n1, n0, d) \ - __asm__ ("! Inlined udiv_qrnnd\n" \ -" wr %%g0,%2,%%y ! Not a delayed write for sparclite\n" \ -" tst %%g0\n" \ -" divscc %3,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%%g1\n" \ -" divscc %%g1,%4,%0\n" \ -" rd %%y,%1\n" \ -" bl,a 1f\n" \ -" add %1,%4,%1\n" \ -"1: ! End of inline udiv_qrnnd" \ - : "=r" (q), "=r" (r) : "r" (n1), "r" (n0), "rI" (d) \ - : "%g1" __AND_CLOBBER_CC) -#define count_leading_zeros(count, x) \ - __asm__ ("scan %1,1,%0" : "=r" (count) : "r" (x)) -/* Early sparclites return 63 for an argument of 0, but they warn that future - implementations might change this. Therefore, leave COUNT_LEADING_ZEROS_0 - undefined. */ -#endif /* __sparclite__ */ -#endif /* __sparc_v8__ */ -#endif /* __sparc_v9__ */ -/* Default to sparc v7 versions of umul_ppmm and udiv_qrnnd. */ -#ifndef umul_ppmm -#define umul_ppmm(w1, w0, u, v) \ - __asm__ ("! Inlined umul_ppmm\n" \ -" wr %%g0,%2,%%y ! SPARC has 0-3 delay insn after a wr\n" \ -" sra %3,31,%%g2 ! Don't move this insn\n" \ -" and %2,%%g2,%%g2 ! Don't move this insn\n" \ -" andcc %%g0,0,%%g1 ! Don't move this insn\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,%3,%%g1\n" \ -" mulscc %%g1,0,%%g1\n" \ -" add %%g1,%%g2,%0\n" \ -" rd %%y,%1" \ - : "=r" (w1), "=r" (w0) : "%rI" (u), "r" (v) \ - : "%g1", "%g2" __AND_CLOBBER_CC) -#endif -#ifndef udiv_qrnnd -#ifndef LONGLONG_STANDALONE -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { UWtype __r; \ - (q) = __MPN(udiv_qrnnd) (&__r, (n1), (n0), (d)); \ - (r) = __r; \ - } while (0) -extern UWtype __MPN(udiv_qrnnd) (UWtype *, UWtype, UWtype, UWtype); -#endif /* LONGLONG_STANDALONE */ -#endif /* udiv_qrnnd */ -#endif /* __sparc__ */ - -#if defined (__sparc__) && W_TYPE_SIZE == 64 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ( \ - "addcc %r4,%5,%1\n" \ - " addccc %r6,%7,%%g0\n" \ - " addc %r2,%3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "rJ" ((UDItype)(ah)), "rI" ((UDItype)(bh)), \ - "%rJ" ((UDItype)(al)), "rI" ((UDItype)(bl)), \ - "%rJ" ((UDItype)(al) >> 32), "rI" ((UDItype)(bl) >> 32) \ - __CLOBBER_CC) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ( \ - "subcc %r4,%5,%1\n" \ - " subccc %r6,%7,%%g0\n" \ - " subc %r2,%3,%0" \ - : "=r" (sh), "=&r" (sl) \ - : "rJ" ((UDItype)(ah)), "rI" ((UDItype)(bh)), \ - "rJ" ((UDItype)(al)), "rI" ((UDItype)(bl)), \ - "rJ" ((UDItype)(al) >> 32), "rI" ((UDItype)(bl) >> 32) \ - __CLOBBER_CC) -#if __VIS__ >= 0x300 -#undef add_ssaaaa -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ( \ - "addcc %r4, %5, %1\n" \ - " addxc %r2, %r3, %0" \ - : "=r" (sh), "=&r" (sl) \ - : "rJ" ((UDItype)(ah)), "rJ" ((UDItype)(bh)), \ - "%rJ" ((UDItype)(al)), "rI" ((UDItype)(bl)) __CLOBBER_CC) -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UDItype __m0 = (m0), __m1 = (m1); \ - (pl) = __m0 * __m1; \ - __asm__ ("umulxhi\t%2, %1, %0" \ - : "=r" (ph) \ - : "%r" (__m0), "r" (__m1)); \ - } while (0) -#define count_leading_zeros(count, x) \ - __asm__ ("lzd\t%1,%0" : "=r" (count) : "r" (x)) -/* Needed by count_leading_zeros_32 in sparc64.h. */ -#define COUNT_LEADING_ZEROS_NEED_CLZ_TAB -#endif -#endif - -#if (defined (__vax) || defined (__vax__)) && W_TYPE_SIZE == 32 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("addl2 %5,%1\n\tadwc %3,%0" \ - : "=g" (sh), "=&g" (sl) \ - : "0" ((USItype)(ah)), "g" ((USItype)(bh)), \ - "%1" ((USItype)(al)), "g" ((USItype)(bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("subl2 %5,%1\n\tsbwc %3,%0" \ - : "=g" (sh), "=&g" (sl) \ - : "0" ((USItype)(ah)), "g" ((USItype)(bh)), \ - "1" ((USItype)(al)), "g" ((USItype)(bl))) -#define smul_ppmm(xh, xl, m0, m1) \ - do { \ - union {UDItype __ll; \ - struct {USItype __l, __h;} __i; \ - } __x; \ - USItype __m0 = (m0), __m1 = (m1); \ - __asm__ ("emul %1,%2,$0,%0" \ - : "=g" (__x.__ll) : "g" (__m0), "g" (__m1)); \ - (xh) = __x.__i.__h; (xl) = __x.__i.__l; \ - } while (0) -#define sdiv_qrnnd(q, r, n1, n0, d) \ - do { \ - union {DItype __ll; \ - struct {SItype __l, __h;} __i; \ - } __x; \ - __x.__i.__h = n1; __x.__i.__l = n0; \ - __asm__ ("ediv %3,%2,%0,%1" \ - : "=g" (q), "=g" (r) : "g" (__x.__ll), "g" (d)); \ - } while (0) -#if 0 -/* FIXME: This instruction appears to be unimplemented on some systems (vax - 8800 maybe). */ -#define count_trailing_zeros(count,x) \ - do { \ - __asm__ ("ffs 0, 31, %1, %0" \ - : "=g" (count) \ - : "g" ((USItype) (x))); \ - } while (0) -#endif -#endif /* vax */ - -#if defined (__z8000__) && W_TYPE_SIZE == 16 -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - __asm__ ("add %H1,%H5\n\tadc %H0,%H3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((unsigned int)(ah)), "r" ((unsigned int)(bh)), \ - "%1" ((unsigned int)(al)), "rQR" ((unsigned int)(bl))) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - __asm__ ("sub %H1,%H5\n\tsbc %H0,%H3" \ - : "=r" (sh), "=&r" (sl) \ - : "0" ((unsigned int)(ah)), "r" ((unsigned int)(bh)), \ - "1" ((unsigned int)(al)), "rQR" ((unsigned int)(bl))) -#define umul_ppmm(xh, xl, m0, m1) \ - do { \ - union {long int __ll; \ - struct {unsigned int __h, __l;} __i; \ - } __x; \ - unsigned int __m0 = (m0), __m1 = (m1); \ - __asm__ ("mult %S0,%H3" \ - : "=r" (__x.__i.__h), "=r" (__x.__i.__l) \ - : "%1" (m0), "rQR" (m1)); \ - (xh) = __x.__i.__h; (xl) = __x.__i.__l; \ - (xh) += ((((signed int) __m0 >> 15) & __m1) \ - + (((signed int) __m1 >> 15) & __m0)); \ - } while (0) -#endif /* __z8000__ */ - -#endif /* __GNUC__ */ - -#endif /* NO_ASM */ - - -/* FIXME: "sidi" here is highly doubtful, should sometimes be "diti". */ -#if !defined (umul_ppmm) && defined (__umulsidi3) -#define umul_ppmm(ph, pl, m0, m1) \ - do { \ - UDWtype __ll = __umulsidi3 (m0, m1); \ - ph = (UWtype) (__ll >> W_TYPE_SIZE); \ - pl = (UWtype) __ll; \ - } while (0) -#endif - -#if !defined (__umulsidi3) -#define __umulsidi3(u, v) \ - ({UWtype __hi, __lo; \ - umul_ppmm (__hi, __lo, u, v); \ - ((UDWtype) __hi << W_TYPE_SIZE) | __lo; }) -#endif - - -#if defined (__cplusplus) -#define __longlong_h_C "C" -#else -#define __longlong_h_C -#endif - -/* Use mpn_umul_ppmm or mpn_udiv_qrnnd functions, if they exist. The "_r" - forms have "reversed" arguments, meaning the pointer is last, which - sometimes allows better parameter passing, in particular on 64-bit - hppa. */ - -#define mpn_umul_ppmm __MPN(umul_ppmm) -extern __longlong_h_C UWtype mpn_umul_ppmm (UWtype *, UWtype, UWtype); - -#if ! defined (umul_ppmm) && HAVE_NATIVE_mpn_umul_ppmm \ - && ! defined (LONGLONG_STANDALONE) -#define umul_ppmm(wh, wl, u, v) \ - do { \ - UWtype __umul_ppmm__p0; \ - (wh) = mpn_umul_ppmm (&__umul_ppmm__p0, (UWtype) (u), (UWtype) (v));\ - (wl) = __umul_ppmm__p0; \ - } while (0) -#endif - -#define mpn_umul_ppmm_r __MPN(umul_ppmm_r) -extern __longlong_h_C UWtype mpn_umul_ppmm_r (UWtype, UWtype, UWtype *); - -#if ! defined (umul_ppmm) && HAVE_NATIVE_mpn_umul_ppmm_r \ - && ! defined (LONGLONG_STANDALONE) -#define umul_ppmm(wh, wl, u, v) \ - do { \ - UWtype __umul_p0; \ - (wh) = mpn_umul_ppmm_r ((UWtype) (u), (UWtype) (v), &__umul_p0); \ - (wl) = __umul_p0; \ - } while (0) -#endif - -#define mpn_udiv_qrnnd __MPN(udiv_qrnnd) -extern __longlong_h_C UWtype mpn_udiv_qrnnd (UWtype *, UWtype, UWtype, UWtype); - -#if ! defined (udiv_qrnnd) && HAVE_NATIVE_mpn_udiv_qrnnd \ - && ! defined (LONGLONG_STANDALONE) -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { \ - UWtype __udiv_qrnnd_r; \ - (q) = mpn_udiv_qrnnd (&__udiv_qrnnd_r, \ - (UWtype) (n1), (UWtype) (n0), (UWtype) d); \ - (r) = __udiv_qrnnd_r; \ - } while (0) -#endif - -#define mpn_udiv_qrnnd_r __MPN(udiv_qrnnd_r) -extern __longlong_h_C UWtype mpn_udiv_qrnnd_r (UWtype, UWtype, UWtype, UWtype *); - -#if ! defined (udiv_qrnnd) && HAVE_NATIVE_mpn_udiv_qrnnd_r \ - && ! defined (LONGLONG_STANDALONE) -#define udiv_qrnnd(q, r, n1, n0, d) \ - do { \ - UWtype __udiv_qrnnd_r; \ - (q) = mpn_udiv_qrnnd_r ((UWtype) (n1), (UWtype) (n0), (UWtype) d, \ - &__udiv_qrnnd_r); \ - (r) = __udiv_qrnnd_r; \ - } while (0) -#endif - - -/* If this machine has no inline assembler, use C macros. */ - -#if !defined (add_ssaaaa) -#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ - do { \ - UWtype __x; \ - __x = (al) + (bl); \ - (sh) = (ah) + (bh) + (__x < (al)); \ - (sl) = __x; \ - } while (0) -#endif - -#if !defined (sub_ddmmss) -#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ - do { \ - UWtype __x; \ - __x = (al) - (bl); \ - (sh) = (ah) - (bh) - ((al) < (bl)); \ - (sl) = __x; \ - } while (0) -#endif - -/* If we lack umul_ppmm but have smul_ppmm, define umul_ppmm in terms of - smul_ppmm. */ -#if !defined (umul_ppmm) && defined (smul_ppmm) -#define umul_ppmm(w1, w0, u, v) \ - do { \ - UWtype __w1; \ - UWtype __xm0 = (u), __xm1 = (v); \ - smul_ppmm (__w1, w0, __xm0, __xm1); \ - (w1) = __w1 + (-(__xm0 >> (W_TYPE_SIZE - 1)) & __xm1) \ - + (-(__xm1 >> (W_TYPE_SIZE - 1)) & __xm0); \ - } while (0) -#endif - -/* If we still don't have umul_ppmm, define it using plain C. - - For reference, when this code is used for squaring (ie. u and v identical - expressions), gcc recognises __x1 and __x2 are the same and generates 3 - multiplies, not 4. The subsequent additions could be optimized a bit, - but the only place GMP currently uses such a square is mpn_sqr_basecase, - and chips obliged to use this generic C umul will have plenty of worse - performance problems than a couple of extra instructions on the diagonal - of sqr_basecase. */ - -#if !defined (umul_ppmm) -#define umul_ppmm(w1, w0, u, v) \ - do { \ - UWtype __x0, __x1, __x2, __x3; \ - UHWtype __ul, __vl, __uh, __vh; \ - UWtype __u = (u), __v = (v); \ - \ - __ul = __ll_lowpart (__u); \ - __uh = __ll_highpart (__u); \ - __vl = __ll_lowpart (__v); \ - __vh = __ll_highpart (__v); \ - \ - __x0 = (UWtype) __ul * __vl; \ - __x1 = (UWtype) __ul * __vh; \ - __x2 = (UWtype) __uh * __vl; \ - __x3 = (UWtype) __uh * __vh; \ - \ - __x1 += __ll_highpart (__x0);/* this can't give carry */ \ - __x1 += __x2; /* but this indeed can */ \ - if (__x1 < __x2) /* did we get it? */ \ - __x3 += __ll_B; /* yes, add it in the proper pos. */ \ - \ - (w1) = __x3 + __ll_highpart (__x1); \ - (w0) = (__x1 << W_TYPE_SIZE/2) + __ll_lowpart (__x0); \ - } while (0) -#endif - -/* If we don't have smul_ppmm, define it using umul_ppmm (which surely will - exist in one form or another. */ -#if !defined (smul_ppmm) -#define smul_ppmm(w1, w0, u, v) \ - do { \ - UWtype __w1; \ - UWtype __xm0 = (u), __xm1 = (v); \ - umul_ppmm (__w1, w0, __xm0, __xm1); \ - (w1) = __w1 - (-(__xm0 >> (W_TYPE_SIZE - 1)) & __xm1) \ - - (-(__xm1 >> (W_TYPE_SIZE - 1)) & __xm0); \ - } while (0) -#endif - -/* Define this unconditionally, so it can be used for debugging. */ -#define __udiv_qrnnd_c(q, r, n1, n0, d) \ - do { \ - UWtype __d1, __d0, __q1, __q0, __r1, __r0, __m; \ - \ - ASSERT ((d) != 0); \ - ASSERT ((n1) < (d)); \ - \ - __d1 = __ll_highpart (d); \ - __d0 = __ll_lowpart (d); \ - \ - __q1 = (n1) / __d1; \ - __r1 = (n1) - __q1 * __d1; \ - __m = __q1 * __d0; \ - __r1 = __r1 * __ll_B | __ll_highpart (n0); \ - if (__r1 < __m) \ - { \ - __q1--, __r1 += (d); \ - if (__r1 >= (d)) /* i.e. we didn't get carry when adding to __r1 */\ - if (__r1 < __m) \ - __q1--, __r1 += (d); \ - } \ - __r1 -= __m; \ - \ - __q0 = __r1 / __d1; \ - __r0 = __r1 - __q0 * __d1; \ - __m = __q0 * __d0; \ - __r0 = __r0 * __ll_B | __ll_lowpart (n0); \ - if (__r0 < __m) \ - { \ - __q0--, __r0 += (d); \ - if (__r0 >= (d)) \ - if (__r0 < __m) \ - __q0--, __r0 += (d); \ - } \ - __r0 -= __m; \ - \ - (q) = __q1 * __ll_B | __q0; \ - (r) = __r0; \ - } while (0) - -/* If the processor has no udiv_qrnnd but sdiv_qrnnd, go through - __udiv_w_sdiv (defined in libgcc or elsewhere). */ -#if !defined (udiv_qrnnd) && defined (sdiv_qrnnd) \ - && ! defined (LONGLONG_STANDALONE) -#define udiv_qrnnd(q, r, nh, nl, d) \ - do { \ - UWtype __r; \ - (q) = __MPN(udiv_w_sdiv) (&__r, nh, nl, d); \ - (r) = __r; \ - } while (0) -__GMP_DECLSPEC UWtype __MPN(udiv_w_sdiv) (UWtype *, UWtype, UWtype, UWtype); -#endif - -/* If udiv_qrnnd was not defined for this processor, use __udiv_qrnnd_c. */ -#if !defined (udiv_qrnnd) -#define UDIV_NEEDS_NORMALIZATION 1 -#define udiv_qrnnd __udiv_qrnnd_c -#endif - -#if !defined (count_leading_zeros) -#define count_leading_zeros(count, x) \ - do { \ - UWtype __xr = (x); \ - UWtype __a; \ - \ - if (W_TYPE_SIZE == 32) \ - { \ - __a = __xr < ((UWtype) 1 << 2*__BITS4) \ - ? (__xr < ((UWtype) 1 << __BITS4) ? 1 : __BITS4 + 1) \ - : (__xr < ((UWtype) 1 << 3*__BITS4) ? 2*__BITS4 + 1 \ - : 3*__BITS4 + 1); \ - } \ - else \ - { \ - for (__a = W_TYPE_SIZE - 8; __a > 0; __a -= 8) \ - if (((__xr >> __a) & 0xff) != 0) \ - break; \ - ++__a; \ - } \ - \ - (count) = W_TYPE_SIZE + 1 - __a - __clz_tab[__xr >> __a]; \ - } while (0) -/* This version gives a well-defined value for zero. */ -#define COUNT_LEADING_ZEROS_0 (W_TYPE_SIZE - 1) -#define COUNT_LEADING_ZEROS_NEED_CLZ_TAB -#define COUNT_LEADING_ZEROS_SLOW -#endif - -/* clz_tab needed by mpn/x86/pentium/mod_1.asm in a fat binary */ -#if HAVE_HOST_CPU_FAMILY_x86 && WANT_FAT_BINARY -#define COUNT_LEADING_ZEROS_NEED_CLZ_TAB -#endif - -#ifdef COUNT_LEADING_ZEROS_NEED_CLZ_TAB -extern const unsigned char __GMP_DECLSPEC __clz_tab[129]; -#endif - -#if !defined (count_trailing_zeros) -#if !defined (COUNT_LEADING_ZEROS_SLOW) -/* Define count_trailing_zeros using an asm count_leading_zeros. */ -#define count_trailing_zeros(count, x) \ - do { \ - UWtype __ctz_x = (x); \ - UWtype __ctz_c; \ - ASSERT (__ctz_x != 0); \ - count_leading_zeros (__ctz_c, __ctz_x & -__ctz_x); \ - (count) = W_TYPE_SIZE - 1 - __ctz_c; \ - } while (0) -#else -/* Define count_trailing_zeros in plain C, assuming small counts are common. - We use clz_tab without ado, since the C count_leading_zeros above will have - pulled it in. */ -#define count_trailing_zeros(count, x) \ - do { \ - UWtype __ctz_x = (x); \ - int __ctz_c; \ - \ - if (LIKELY ((__ctz_x & 0xff) != 0)) \ - (count) = __clz_tab[__ctz_x & -__ctz_x] - 2; \ - else \ - { \ - for (__ctz_c = 8 - 2; __ctz_c < W_TYPE_SIZE - 2; __ctz_c += 8) \ - { \ - __ctz_x >>= 8; \ - if (LIKELY ((__ctz_x & 0xff) != 0)) \ - break; \ - } \ - \ - (count) = __ctz_c + __clz_tab[__ctz_x & -__ctz_x]; \ - } \ - } while (0) -#endif -#endif - -#ifndef UDIV_NEEDS_NORMALIZATION -#define UDIV_NEEDS_NORMALIZATION 0 -#endif - -/* Whether udiv_qrnnd is actually implemented with udiv_qrnnd_preinv, and - that hence the latter should always be used. */ -#ifndef UDIV_PREINV_ALWAYS -#define UDIV_PREINV_ALWAYS 0 -#endif -- 2.26.1