src/longlong.h - coreutils-8.23

Macros defined

Source code

/* longlong.h -- definitions for mixed size 32/64 bit arithmetic.

Copyright 1991-2014 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 http://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
‌#define UMUL_TIME 18
#else /* ! __GNUC__ */
#include <machine/builtins.h>
‌#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
‌#define UDIV_TIME 220
#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 <intrinsics.h>
‌#define UDIV_PREINV_ALWAYS  1
‌#define UDIV_NEEDS_NORMALIZATION 1
‌#define UDIV_TIME 220
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<bl" arithmetically comes out making an actual 0 or 1 in a
   register, which takes an extra cycle.  */
‌#define sub_ddmmss(sh, sl, ah, al, bh, bl)      \
  do {                        \
    UWtype __x;                    \
    __x = (al) - (bl);                \
    if ((al) < (bl))                \
      (sh) = (ah) - (bh) - 1;            \
    else                    \
      (sh) = (ah) - (bh);            \
    (sl) = __x;                    \
  } while (0)
#if defined (__GNUC__) && ! defined (__INTEL_COMPILER)
/* Do both product parts in assembly, since that gives better code with
   all gcc versions.  Some callers will just use the upper part, and in
   that situation we waste an instruction, but not any cycles.  */
‌#define umul_ppmm(ph, pl, m0, m1) \
    __asm__ ("xma.hu %0 = %2, %3, f0\n\txma.l %1 = %2, %3, f0"        \
         : "=&f" (ph), "=f" (pl)                    \
         : "f" (m0), "f" (m1))
‌#define UMUL_TIME 14
‌#define count_leading_zeros(count, x) \
  do {                                    \
    UWtype _x = (x), _y, _a, _c;                    \
    __asm__ ("mux1 %0 = %1, @rev" : "=r" (_y) : "r" (_x));        \
    __asm__ ("czx1.l %0 = %1" : "=r" (_a) : "r" (-_y | _y));        \
    _c = (_a - 1) << 3;                            \
    _x >>= _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 <ia64intrin.h>
‌#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
‌#define UDIV_TIME 220
#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)),                    \
         "rIJ" ((USItype) (bh)),                    \
         "%r" ((USItype) (al)),                    \
         "rIJ" ((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)),                    \
         "rIJ" ((USItype) (bh)),                    \
         "r" ((USItype) (al)),                    \
         "rIJ" ((USItype) (bl)))
#endif

#if defined (__arm__) && !defined (__thumb__) && W_TYPE_SIZE == 32
‌#define add_ssaaaa(sh, sl, ah, al, bh, bl) \
  __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)
‌#define sub_ddmmss(sh, sl, ah, al, bh, bl) \
  do {                                    \
    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))                    \
      {                                    \
    if (__builtin_constant_p (bh))                    \
      __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\trsc\t%0, %3, %2"        \
           : "=r" (sh), "=&r" (sl)                \
           : "rI" (ah), "r" (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)
#if 1 || defined (__arm_m__)    /* `M' series has widening multiply support */
‌#define umul_ppmm(xh, xl, a, b) \
  __asm__ ("umull %0,%1,%2,%3" : "=&r" (xl), "=&r" (xh) : "r" (a), "r" (b))
‌#define UMUL_TIME 5
‌#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
‌#define UDIV_TIME 70
#endif /* LONGLONG_STANDALONE */
#else
‌#define umul_ppmm(xh, xl, a, b) \
  __asm__ ("%@ Inlined umul_ppmm\n"                    \
"    mov    %|r0, %2, lsr #16\n"                    \
"    mov    %|r2, %3, lsr #16\n"                    \
"    bic    %|r1, %2, %|r0, lsl #16\n"                \
"    bic    %|r2, %3, %|r2, lsl #16\n"                \
"    mul    %1, %|r1, %|r2\n"                    \
"    mul    %|r2, %|r0, %|r2\n"                    \
"    mul    %|r1, %0, %|r1\n"                    \
"    mul    %0, %|r0, %0\n"                        \
"    adds    %|r1, %|r2, %|r1\n"                    \
"    addcs    %0, %0, #65536\n"                    \
"    adds    %1, %1, %|r1, lsl #16\n"                \
"    adc    %0, %0, %|r1, lsr #16"                    \
       : "=&r" (xh), "=r" (xl)                    \
       : "r" (a), "r" (b)                        \
       : "r0", "r1", "r2")
‌#define UMUL_TIME 20
#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);
‌#define UDIV_TIME 200
#endif /* LONGLONG_STANDALONE */
#endif
/* This is a bizarre test, but GCC doesn't define any useful common symbol. */
#if defined (__ARM_ARCH_5__)  || defined (__ARM_ARCH_5T__) || \
    defined (__ARM_ARCH_5E__) || defined (__ARM_ARCH_5TE__)|| \
    defined (__ARM_ARCH_6__)  || defined (__ARM_ARCH_6J__) || \
    defined (__ARM_ARCH_6K__) || defined (__ARM_ARCH_6Z__) || \
    defined (__ARM_ARCH_6ZK__)|| defined (__ARM_ARCH_6T2__)|| \
    defined (__ARM_ARCH_6M__) || defined (__ARM_ARCH_7__)  || \
    defined (__ARM_ARCH_7A__) || defined (__ARM_ARCH_7R__) || \
    defined (__ARM_ARCH_7M__) || defined (__ARM_ARCH_7EM__)
‌#define count_leading_zeros(count, x) \
  __asm__ ("clz\t%0, %1" : "=r" (count) : "r" (x))
‌#define COUNT_LEADING_ZEROS_0 32
#endif
#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) \
  __asm__ ("adds\t%1, %x4, %5\n\tadc\t%0, %x2, %x3"            \
       : "=r" (sh), "=&r" (sl)                    \
       : "rZ" (ah), "rZ" (bh), "%r" (al), "rI" (bl) __CLOBBER_CC)
‌#define sub_ddmmss(sh, sl, ah, al, bh, bl) \
  __asm__ ("subs\t%1, %x4, %5\n\tsbc\t%0, %x2, %x3"            \
       : "=r,r" (sh), "=&r,&r" (sl)                    \
       : "rZ,rZ" (ah), "rZ,rZ" (bh), "r,Z" (al), "rI,r" (bl) __CLOBBER_CC)
‌#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)
‌#define count_leading_zeros(count, x) \
  __asm__ ("clz\t%0, %1" : "=r" (count) : "r" (x))
‌#define count_trailing_zeros(count, x) \
  __asm__ ("rbit\t%0, %1\n\tclz\t%0, %0" : "=r" (count) : "r" (x))
‌#define COUNT_LEADING_ZEROS_0 64
#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)
‌#define UMUL_TIME 8
‌#define UDIV_TIME 60
#else
‌#define UMUL_TIME 40
‌#define UDIV_TIME 80
#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

#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;                                \
    ASSERT ((n) != 0);                            \
    __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 */

#ifndef UMUL_TIME
‌#define UMUL_TIME 10
#endif
#ifndef UDIV_TIME
‌#define UDIV_TIME 40
#endif
#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)))
‌#define umul_ppmm(w1, w0, u, v) \
  __asm__ ("mulq %3"                            \
       : "=a" (w0), "=d" (w1)                    \
       : "%0" ((UDItype)(u)), "rm" ((UDItype)(v)))
‌#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)))
/* bsrq destination must be a 64-bit register, hence UDItype for __cbtmp. */
‌#define count_leading_zeros(count, x)                    \
  do {                                    \
    UDItype __cbtmp;                            \
    ASSERT ((x) != 0);                            \
    __asm__ ("bsrq %1,%0" : "=r" (__cbtmp) : "rm" ((UDItype)(x)));    \
    (count) = __cbtmp ^ 63;                        \
  } while (0)
/* bsfq destination must be a 64-bit register, "%q0" forces this in case
   count is only an int. */
‌#define count_trailing_zeros(count, x)                    \
  do {                                    \
    ASSERT ((x) != 0);                            \
    __asm__ ("bsfq %1,%q0" : "=r" (count) : "rm" ((UDItype)(x)));    \
  } while (0)
#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 UMUL_TIME 45
‌#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 UDIV_TIME 90
‌#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)
‌#define UMUL_TIME 100
‌#define UDIV_TIME 400
#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; })
‌#define UMUL_TIME 5
‌#define UDIV_TIME 25
#else
‌#define UMUL_TIME 17
‌#define UDIV_TIME 150
#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)
‌#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
‌#define UMUL_TIME 10
‌#define UDIV_TIME 100
#endif /* __mips */

#if (defined (__mips) && __mips >= 3) && W_TYPE_SIZE == 64
#if __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)
‌#define umul_ppmm(w1, w0, u, v) \
  __asm__ ("dmultu %2,%3" : "=l" (w0), "=h" (w1) : "d" (u), "d" (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" (u), "d" (v))
#endif
‌#define UMUL_TIME 20
‌#define UDIV_TIME 140
#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));    \
    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));    \
    else                                \
      __asm__ ("add%I5c %1,%4,%5\n\tadde %0,%2,%3"            \
         : "=r" (sh), "=&r" (sl)                    \
         : "r" (ah), "r" (bh), "%r" (al), "rI" (bl));        \
  } 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));\
    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));\
    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));\
    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));\
    else                                \
      __asm__ ("subf%I4c %1,%5,%4\n\tsubfe %0,%3,%2"            \
           : "=r" (sh), "=&r" (sl)                    \
           : "r" (ah), "r" (bh), "rI" (al), "r" (bl));        \
  } 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 UMUL_TIME 15
‌#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)
‌#define SMUL_TIME 14
‌#define UDIV_TIME 120
#else
‌#define UMUL_TIME 8
‌#define smul_ppmm(xh, xl, m0, m1) \
  __asm__ ("mul %0,%2,%3" : "=r" (xh), "=q" (xl) : "r" (m0), "r" (m1))
‌#define SMUL_TIME 4
‌#define sdiv_qrnnd(q, r, nh, nl, d) \
  __asm__ ("div %0,%2,%4" : "=r" (q), "=q" (r) : "r" (nh), "1" (nl), "r" (d))
‌#define UDIV_TIME 100
#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" (ah), "%r" (al), "rI" (bl));    \
    else if (__builtin_constant_p (bh) && (bh) == ~(UDItype) 0)        \
      __asm__ ("add%I4c %1,%3,%4\n\taddme %0,%2"            \
         : "=r" (sh), "=&r" (sl) : "r" (ah), "%r" (al), "rI" (bl));    \
    else                                \
      __asm__ ("add%I5c %1,%4,%5\n\tadde %0,%2,%3"            \
         : "=r" (sh), "=&r" (sl)                    \
         : "r" (ah), "r" (bh), "%r" (al), "rI" (bl));        \
  } 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" (bh), "rI" (al), "*rI" (-bl)); \
    else if (__builtin_constant_p (ah) && (ah) == ~(UDItype) 0)    \
      __asm__ ("addic %1,%3,%4\n\tsubfme %0,%2"            \
           : "=r" (sh), "=&r" (sl) : "r" (bh), "rI" (al), "*rI" (-bl)); \
    else if (__builtin_constant_p (bh) && (bh) == 0)        \
      __asm__ ("addic %1,%3,%4\n\taddme %0,%2"            \
           : "=r" (sh), "=&r" (sl) : "r" (ah), "rI" (al), "*rI" (-bl)); \
    else if (__builtin_constant_p (bh) && (bh) == ~(UDItype) 0)    \
      __asm__ ("addic %1,%3,%4\n\taddze %0,%2"            \
           : "=r" (sh), "=&r" (sl) : "r" (ah), "rI" (al), "*rI" (-bl)); \
    else                                \
      __asm__ ("addic %1,%4,%5\n\tsubfe %0,%3,%2"            \
           : "=r" (sh), "=&r" (sl)                \
           : "r" (ah), "r" (bh), "rI" (al), "*rI" (-bl));    \
    } else {                                \
    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)); \
    else if (__builtin_constant_p (ah) && (ah) == ~(UDItype) 0)    \
      __asm__ ("subf%I3c %1,%4,%3\n\tsubfme %0,%2"            \
           : "=r" (sh), "=&r" (sl) : "r" (bh), "rI" (al), "r" (bl)); \
    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)); \
    else if (__builtin_constant_p (bh) && (bh) == ~(UDItype) 0)    \
      __asm__ ("subf%I3c %1,%4,%3\n\taddze %0,%2"            \
           : "=r" (sh), "=&r" (sl) : "r" (ah), "rI" (al), "r" (bl)); \
    else                                \
      __asm__ ("subf%I4c %1,%5,%4\n\tsubfe %0,%3,%2"        \
           : "=r" (sh), "=&r" (sl)                \
           : "r" (ah), "r" (bh), "rI" (al), "r" (bl));        \
    }                                    \
  } 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 0 && __GMP_GNUC_PREREQ (4,4) /* Disable, this results in libcalls! */
#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 UMUL_TIME 15
‌#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)
‌#define SMUL_TIME 14  /* ??? */
‌#define UDIV_TIME 120 /* ??? */
#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 UMUL_TIME 20
‌#define UDIV_TIME 200
‌#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 (__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")
‌#define UMUL_TIME 5
#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))
‌#define UMUL_TIME 5

#if HAVE_HOST_CPU_supersparc
‌#define UDIV_TIME 60        /* SuperSPARC timing */
#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)
‌#define UDIV_TIME 25
#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 UMUL_TIME 5
‌#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 UDIV_TIME 37
‌#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)
‌#define UMUL_TIME 39        /* 39 instructions */
#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);
#ifndef UDIV_TIME
‌#define UDIV_TIME 140
#endif
#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" (ah), "rI" (bh), "%rJ" (al), "rI" (bl),        \
        "%rJ" ((al) >> 32), "rI" ((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" (ah), "rI" (bh), "rJ" (al), "rI" (bl),            \
        "rJ" ((al) >> 32), "rI" ((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" (ah), "rJ" (bh), "%rJ" (al), "rI" (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))
#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) \
  {                                    \
    UDWtype __ll = __umulsidi3 (m0, m1);                \
    ph = (UWtype) (__ll >> W_TYPE_SIZE);                \
    pl = (UWtype) __ll;                            \
  }
#endif

#if !defined (__umulsidi3)
‌#define __umulsidi3(u, v) \
  ({UWtype __hi, __lo;                            \
    umul_ppmm (__hi, __lo, u, v);                    \
    ((UDWtype) __hi << W_TYPE_SIZE) | __lo; })
#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 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 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 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 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)
‌#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

/* Give defaults for UMUL_TIME and UDIV_TIME.  */
#ifndef UMUL_TIME
‌#define UMUL_TIME 1
#endif

#ifndef UDIV_TIME
‌#define UDIV_TIME UMUL_TIME
#endif