URL https://opencores.org/ocsvn/openrisc/openrisc/trunk

Subversion Repositories openrisc

[/] [openrisc/] [trunk/] [gnu-old/] [gcc-4.2.2/] [gcc/] [config/] [m68k/] [fpgnulib.c] - Blame information for rev 825

Go to most recent revision | Details | Compare with Previous | View Log


/* This is a stripped down version of floatlib.c.  It supplies only those
   functions which exist in libgcc, but for which there is not assembly
   language versions in m68k/lb1sf68.asm.
 
   It also includes simplistic support for extended floats (by working in
   double precision).  You must compile this file again with -DEXTFLOAT
   to get this support.  */
 
/*
** gnulib support for software floating point.
** Copyright (C) 1991 by Pipeline Associates, Inc.  All rights reserved.
** Permission is granted to do *anything* you want with this file,
** commercial or otherwise, provided this message remains intact.  So there!
** I would appreciate receiving any updates/patches/changes that anyone
** makes, and am willing to be the repository for said changes (am I
** making a big mistake?).
**
** Pat Wood
** Pipeline Associates, Inc.
** pipeline!phw@motown.com or
** sun!pipeline!phw or
** uunet!motown!pipeline!phw
**
** 05/01/91 -- V1.0 -- first release to gcc mailing lists
** 05/04/91 -- V1.1 -- added float and double prototypes and return values
**                  -- fixed problems with adding and subtracting zero
**                  -- fixed rounding in truncdfsf2
**                  -- fixed SWAP define and tested on 386
*/
 
/*
** The following are routines that replace the gnulib soft floating point
** routines that are called automatically when -msoft-float is selected.
** The support single and double precision IEEE format, with provisions
** for byte-swapped machines (tested on 386).  Some of the double-precision
** routines work at full precision, but most of the hard ones simply punt
** and call the single precision routines, producing a loss of accuracy.
** long long support is not assumed or included.
** Overall accuracy is close to IEEE (actually 68882) for single-precision
** arithmetic.  I think there may still be a 1 in 1000 chance of a bit
** being rounded the wrong way during a multiply.  I'm not fussy enough to
** bother with it, but if anyone is, knock yourself out.
**
** Efficiency has only been addressed where it was obvious that something
** would make a big difference.  Anyone who wants to do this right for
** best speed should go in and rewrite in assembler.
**
** I have tested this only on a 68030 workstation and 386/ix integrated
** in with -msoft-float.
*/
 
/* the following deal with IEEE single-precision numbers */
#define EXCESS          126L
#define SIGNBIT         0x80000000L
#define HIDDEN          (1L << 23L)
#define SIGN(fp)        ((fp) & SIGNBIT)
#define EXP(fp)         (((fp) >> 23L) & 0xFF)
#define MANT(fp)        (((fp) & 0x7FFFFFL) | HIDDEN)
#define PACK(s,e,m)     ((s) | ((e) << 23L) | (m))
 
/* the following deal with IEEE double-precision numbers */
#define EXCESSD         1022L
#define HIDDEND         (1L << 20L)
#define EXPDBITS        11
#define EXPDMASK        0x7FFL
#define EXPD(fp)        (((fp.l.upper) >> 20L) & 0x7FFL)
#define SIGND(fp)       ((fp.l.upper) & SIGNBIT)
#define MANTD(fp)       (((((fp.l.upper) & 0xFFFFF) | HIDDEND) << 10) | \
                                (fp.l.lower >> 22))
#define MANTDMASK       0xFFFFFL /* mask of upper part */
 
/* the following deal with IEEE extended-precision numbers */
#define EXCESSX         16382L
#define HIDDENX         (1L << 31L)
#define EXPXBITS        15
#define EXPXMASK        0x7FFF
#define EXPX(fp)        (((fp.l.upper) >> 16) & EXPXMASK)
#define SIGNX(fp)       ((fp.l.upper) & SIGNBIT)
#define MANTXMASK       0x7FFFFFFFL /* mask of upper part */
 
union double_long
{
  double d;
  struct {
      long upper;
      unsigned long lower;
    } l;
};
 
union float_long {
  float f;
  long l;
};
 
union long_double_long
{
  long double ld;
  struct
    {
      long upper;
      unsigned long middle;
      unsigned long lower;
    } l;
};
 
#ifndef EXTFLOAT
 
int
__unordsf2(float a, float b)
{
  union float_long fl;
 
  fl.f = a;
  if (EXP(fl.l) == EXP(~0u) && (MANT(fl.l) & ~HIDDEN) != 0)
    return 1;
  fl.f = b;
  if (EXP(fl.l) == EXP(~0u) && (MANT(fl.l) & ~HIDDEN) != 0)
    return 1;
  return 0;
}
 
int
__unorddf2(double a, double b)
{
  union double_long dl;
 
  dl.d = a;
  if (EXPD(dl) == EXPDMASK
      && ((dl.l.upper & MANTDMASK) != 0 || dl.l.lower != 0))
    return 1;
  dl.d = b;
  if (EXPD(dl) == EXPDMASK
      && ((dl.l.upper & MANTDMASK) != 0 || dl.l.lower != 0))
    return 1;
  return 0;
}
 
/* convert unsigned int to double */
double
__floatunsidf (unsigned long a1)
{
  long exp = 32 + EXCESSD;
  union double_long dl;
 
  if (!a1)
    {
      dl.l.upper = dl.l.lower = 0;
      return dl.d;
    }
 
  while (a1 < 0x2000000L)
    {
      a1 <<= 4;
      exp -= 4;
    }
 
  while (a1 < 0x80000000L)
    {
      a1 <<= 1;
      exp--;
    }
 
  /* pack up and go home */
  dl.l.upper = exp << 20L;
  dl.l.upper |= (a1 >> 11L) & ~HIDDEND;
  dl.l.lower = a1 << 21L;
 
  return dl.d;
}
 
/* convert int to double */
double
__floatsidf (long a1)
{
  long sign = 0, exp = 31 + EXCESSD;
  union double_long dl;
 
  if (!a1)
    {
      dl.l.upper = dl.l.lower = 0;
      return dl.d;
    }
 
  if (a1 < 0)
    {
      sign = SIGNBIT;
      a1 = (long)-(unsigned long)a1;
      if (a1 < 0)
        {
          dl.l.upper = SIGNBIT | ((32 + EXCESSD) << 20L);
          dl.l.lower = 0;
          return dl.d;
        }
    }
 
  while (a1 < 0x1000000L)
    {
      a1 <<= 4;
      exp -= 4;
    }
 
  while (a1 < 0x40000000L)
    {
      a1 <<= 1;
      exp--;
    }
 
  /* pack up and go home */
  dl.l.upper = sign;
  dl.l.upper |= exp << 20L;
  dl.l.upper |= (a1 >> 10L) & ~HIDDEND;
  dl.l.lower = a1 << 22L;
 
  return dl.d;
}
 
/* convert unsigned int to float */
float
__floatunsisf (unsigned long l)
{
  double foo = __floatunsidf (l);
  return foo;
}
 
/* convert int to float */
float
__floatsisf (long l)
{
  double foo = __floatsidf (l);
  return foo;
}
 
/* convert float to double */
double
__extendsfdf2 (float a1)
{
  register union float_long fl1;
  register union double_long dl;
  register long exp;
  register long mant;
 
  fl1.f = a1;
 
  dl.l.upper = SIGN (fl1.l);
  if ((fl1.l & ~SIGNBIT) == 0)
    {
      dl.l.lower = 0;
      return dl.d;
    }
 
  exp = EXP(fl1.l);
  mant = MANT (fl1.l) & ~HIDDEN;
  if (exp == 0)
    {
      /* Denormal.  */
      exp = 1;
      while (!(mant & HIDDEN))
        {
          mant <<= 1;
          exp--;
        }
      mant &= ~HIDDEN;
    }
  exp = exp - EXCESS + EXCESSD;
  dl.l.upper |= exp << 20;
  dl.l.upper |= mant >> 3;
  dl.l.lower = mant << 29;
 
  return dl.d;
}
 
/* convert double to float */
float
__truncdfsf2 (double a1)
{
  register long exp;
  register long mant;
  register union float_long fl;
  register union double_long dl1;
 
  dl1.d = a1;
 
  if ((dl1.l.upper & ~SIGNBIT) == 0 && !dl1.l.lower)
    {
      fl.l = SIGND(dl1);
      return fl.f;
    }
 
  exp = EXPD (dl1) - EXCESSD + EXCESS;
 
  /* shift double mantissa 6 bits so we can round */
  mant = MANTD (dl1) >> 6;
 
  /* Check for underflow and denormals.  */
  if (exp <= 0)
    {
      if (exp < -24)
        mant = 0;
      else
        mant >>= 1 - exp;
      exp = 0;
    }
 
  /* now round and shift down */
  mant += 1;
  mant >>= 1;
 
  /* did the round overflow? */
  if (mant & 0xFF000000L)
    {
      mant >>= 1;
      exp++;
    }
 
  mant &= ~HIDDEN;
 
  /* pack up and go home */
  fl.l = PACK (SIGND (dl1), exp, mant);
  return (fl.f);
}
 
/* convert double to int */
long
__fixdfsi (double a1)
{
  register union double_long dl1;
  register long exp;
  register long l;
 
  dl1.d = a1;
 
  if (!dl1.l.upper && !dl1.l.lower)
    return 0;
 
  exp = EXPD (dl1) - EXCESSD - 31;
  l = MANTD (dl1);
 
  if (exp > 0)
    {
      /* Return largest integer.  */
      return SIGND (dl1) ? 0x80000000L : 0x7fffffffL;
    }
 
  if (exp <= -32)
    return 0;
 
  /* shift down until exp = 0 */
  if (exp < 0)
    l >>= -exp;
 
  return (SIGND (dl1) ? -l : l);
}
 
/* convert float to int */
long
__fixsfsi (float a1)
{
  double foo = a1;
  return __fixdfsi (foo);
}
 
#else /* EXTFLOAT */
 
/* Primitive extended precision floating point support.
 
   We assume all numbers are normalized, don't do any rounding, etc.  */
 
/* Prototypes for the above in case we use them.  */
double __floatunsidf (unsigned long);
double __floatsidf (long);
float __floatsisf (long);
double __extendsfdf2 (float);
float __truncdfsf2 (double);
long __fixdfsi (double);
long __fixsfsi (float);
 
int
__unordxf2(long double a, long double b)
{
  union long_double_long ldl;
 
  ldl.ld = a;
  if (EXPX(ldl) == EXPXMASK
      && ((ldl.l.middle & MANTXMASK) != 0 || ldl.l.lower != 0))
    return 1;
  ldl.ld = b;
  if (EXPX(ldl) == EXPXMASK
      && ((ldl.l.middle & MANTXMASK) != 0 || ldl.l.lower != 0))
    return 1;
  return 0;
}
 
/* convert double to long double */
long double
__extenddfxf2 (double d)
{
  register union double_long dl;
  register union long_double_long ldl;
  register long exp;
 
  dl.d = d;
  /*printf ("dfxf in: %g\n", d);*/
 
  ldl.l.upper = SIGND (dl);
  if ((dl.l.upper & ~SIGNBIT) == 0 && !dl.l.lower)
    {
      ldl.l.middle = 0;
      ldl.l.lower = 0;
      return ldl.ld;
    }
 
  exp = EXPD (dl) - EXCESSD + EXCESSX;
  ldl.l.upper |= exp << 16;
  ldl.l.middle = HIDDENX;
  /* 31-20: # mantissa bits in ldl.l.middle - # mantissa bits in dl.l.upper */
  ldl.l.middle |= (dl.l.upper & MANTDMASK) << (31 - 20);
  /* 1+20: explicit-integer-bit + # mantissa bits in dl.l.upper */
  ldl.l.middle |= dl.l.lower >> (1 + 20);
  /* 32 - 21: # bits of dl.l.lower in ldl.l.middle */
  ldl.l.lower = dl.l.lower << (32 - 21);
 
  /*printf ("dfxf out: %s\n", dumpxf (ldl.ld));*/
  return ldl.ld;
}
 
/* convert long double to double */
double
__truncxfdf2 (long double ld)
{
  register long exp;
  register union double_long dl;
  register union long_double_long ldl;
 
  ldl.ld = ld;
  /*printf ("xfdf in: %s\n", dumpxf (ld));*/
 
  dl.l.upper = SIGNX (ldl);
  if ((ldl.l.upper & ~SIGNBIT) == 0 && !ldl.l.middle && !ldl.l.lower)
    {
      dl.l.lower = 0;
      return dl.d;
    }
 
  exp = EXPX (ldl) - EXCESSX + EXCESSD;
  /* ??? quick and dirty: keep `exp' sane */
  if (exp >= EXPDMASK)
    exp = EXPDMASK - 1;
  dl.l.upper |= exp << (32 - (EXPDBITS + 1));
  /* +1-1: add one for sign bit, but take one off for explicit-integer-bit */
  dl.l.upper |= (ldl.l.middle & MANTXMASK) >> (EXPDBITS + 1 - 1);
  dl.l.lower = (ldl.l.middle & MANTXMASK) << (32 - (EXPDBITS + 1 - 1));
  dl.l.lower |= ldl.l.lower >> (EXPDBITS + 1 - 1);
 
  /*printf ("xfdf out: %g\n", dl.d);*/
  return dl.d;
}
 
/* convert a float to a long double */
long double
__extendsfxf2 (float f)
{
  long double foo = __extenddfxf2 (__extendsfdf2 (f));
  return foo;
}
 
/* convert a long double to a float */
float
__truncxfsf2 (long double ld)
{
  float foo = __truncdfsf2 (__truncxfdf2 (ld));
  return foo;
}
 
/* convert an int to a long double */
long double
__floatsixf (long l)
{
  double foo = __floatsidf (l);
  return foo;
}
 
/* convert an unsigned int to a long double */
long double
__floatunsixf (unsigned long l)
{
  double foo = __floatunsidf (l);
  return foo;
}
 
/* convert a long double to an int */
long
__fixxfsi (long double ld)
{
  long foo = __fixdfsi ((double) ld);
  return foo;
}
 
/* The remaining provide crude math support by working in double precision.  */
 
long double
__addxf3 (long double x1, long double x2)
{
  return (double) x1 + (double) x2;
}
 
long double
__subxf3 (long double x1, long double x2)
{
  return (double) x1 - (double) x2;
}
 
long double
__mulxf3 (long double x1, long double x2)
{
  return (double) x1 * (double) x2;
}
 
long double
__divxf3 (long double x1, long double x2)
{
  return (double) x1 / (double) x2;
}
 
long double
__negxf2 (long double x1)
{
  return - (double) x1;
}
 
long
__cmpxf2 (long double x1, long double x2)
{
  return __cmpdf2 ((double) x1, (double) x2);
}
 
long
__eqxf2 (long double x1, long double x2)
{
  return __cmpdf2 ((double) x1, (double) x2);
}
 
long
__nexf2 (long double x1, long double x2)
{
  return __cmpdf2 ((double) x1, (double) x2);
}
 
long
__ltxf2 (long double x1, long double x2)
{
  return __cmpdf2 ((double) x1, (double) x2);
}
 
long
__lexf2 (long double x1, long double x2)
{
  return __cmpdf2 ((double) x1, (double) x2);
}
 
long
__gtxf2 (long double x1, long double x2)
{
  return __cmpdf2 ((double) x1, (double) x2);
}
 
long
__gexf2 (long double x1, long double x2)
{
  return __cmpdf2 ((double) x1, (double) x2);
}
 
#endif /* EXTFLOAT */

Browse

Tools

Subversion Repositories openrisc

[/] [openrisc/] [trunk/] [gnu-old/] [gcc-4.2.2/] [gcc/] [config/] [m68k/] [fpgnulib.c] - Blame information for rev 825

Line No.	Rev	Author	Line
1	38	julius	`/* This is a stripped down version of floatlib.c. It supplies only those`
2			`functions which exist in libgcc, but for which there is not assembly`
3			`language versions in m68k/lb1sf68.asm.`
4
5			`It also includes simplistic support for extended floats (by working in`
6			`double precision). You must compile this file again with -DEXTFLOAT`
7			`to get this support. */`
8
9			`/*`
10			`** gnulib support for software floating point.`
11			`** Copyright (C) 1991 by Pipeline Associates, Inc. All rights reserved.`
12			`** Permission is granted to do anything you want with this file,`
13			`** commercial or otherwise, provided this message remains intact. So there!`
14			`** I would appreciate receiving any updates/patches/changes that anyone`
15			`** makes, and am willing to be the repository for said changes (am I`
16			`** making a big mistake?).`
17			`**`
18			`** Pat Wood`
19			`** Pipeline Associates, Inc.`
20			`** pipeline!phw@motown.com or`
21			`** sun!pipeline!phw or`
22			`** uunet!motown!pipeline!phw`
23			`**`
24			`** 05/01/91 -- V1.0 -- first release to gcc mailing lists`
25			`** 05/04/91 -- V1.1 -- added float and double prototypes and return values`
26			`** -- fixed problems with adding and subtracting zero`
27			`** -- fixed rounding in truncdfsf2`
28			`** -- fixed SWAP define and tested on 386`
29			`*/`
30
31			`/*`
32			`** The following are routines that replace the gnulib soft floating point`
33			`** routines that are called automatically when -msoft-float is selected.`
34			`** The support single and double precision IEEE format, with provisions`
35			`** for byte-swapped machines (tested on 386). Some of the double-precision`
36			`** routines work at full precision, but most of the hard ones simply punt`
37			`** and call the single precision routines, producing a loss of accuracy.`
38			`** long long support is not assumed or included.`
39			`** Overall accuracy is close to IEEE (actually 68882) for single-precision`
40			`** arithmetic. I think there may still be a 1 in 1000 chance of a bit`
41			`** being rounded the wrong way during a multiply. I'm not fussy enough to`
42			`** bother with it, but if anyone is, knock yourself out.`
43			`**`
44			`** Efficiency has only been addressed where it was obvious that something`
45			`** would make a big difference. Anyone who wants to do this right for`
46			`** best speed should go in and rewrite in assembler.`
47			`**`
48			`** I have tested this only on a 68030 workstation and 386/ix integrated`
49			`** in with -msoft-float.`
50			`*/`
51
52			`/* the following deal with IEEE single-precision numbers */`
53			`#define EXCESS 126L`
54			`#define SIGNBIT 0x80000000L`
55			`#define HIDDEN (1L << 23L)`
56			`#define SIGN(fp) ((fp) & SIGNBIT)`
57			`#define EXP(fp) (((fp) >> 23L) & 0xFF)`
58			`#define MANT(fp) (((fp) & 0x7FFFFFL) \| HIDDEN)`
59			`#define PACK(s,e,m) ((s) \| ((e) << 23L) \| (m))`
60
61			`/* the following deal with IEEE double-precision numbers */`
62			`#define EXCESSD 1022L`
63			`#define HIDDEND (1L << 20L)`
64			`#define EXPDBITS 11`
65			`#define EXPDMASK 0x7FFL`
66			`#define EXPD(fp) (((fp.l.upper) >> 20L) & 0x7FFL)`
67			`#define SIGND(fp) ((fp.l.upper) & SIGNBIT)`
68			`#define MANTD(fp) (((((fp.l.upper) & 0xFFFFF) \| HIDDEND) << 10) \| \`
69			`(fp.l.lower >> 22))`
70			`#define MANTDMASK 0xFFFFFL /* mask of upper part */`
71
72			`/* the following deal with IEEE extended-precision numbers */`
73			`#define EXCESSX 16382L`
74			`#define HIDDENX (1L << 31L)`
75			`#define EXPXBITS 15`
76			`#define EXPXMASK 0x7FFF`
77			`#define EXPX(fp) (((fp.l.upper) >> 16) & EXPXMASK)`
78			`#define SIGNX(fp) ((fp.l.upper) & SIGNBIT)`
79			`#define MANTXMASK 0x7FFFFFFFL /* mask of upper part */`
80
81			`union double_long`
82			`{`
83			`double d;`
84			`struct {`
85			`long upper;`
86			`unsigned long lower;`
87			`} l;`
88			`};`
89
90			`union float_long {`
91			`float f;`
92			`long l;`
93			`};`
94
95			`union long_double_long`
96			`{`
97			`long double ld;`
98			`struct`
99			`{`
100			`long upper;`
101			`unsigned long middle;`
102			`unsigned long lower;`
103			`} l;`
104			`};`
105
106			`#ifndef EXTFLOAT`
107
108			`int`
109			`__unordsf2(float a, float b)`
110			`{`
111			`union float_long fl;`
112
113			`fl.f = a;`
114			`if (EXP(fl.l) == EXP(~0u) && (MANT(fl.l) & ~HIDDEN) != 0)`
115			`return 1;`
116			`fl.f = b;`
117			`if (EXP(fl.l) == EXP(~0u) && (MANT(fl.l) & ~HIDDEN) != 0)`
118			`return 1;`
119			`return 0;`
120			`}`
121
122			`int`
123			`__unorddf2(double a, double b)`
124			`{`
125			`union double_long dl;`
126
127			`dl.d = a;`
128			`if (EXPD(dl) == EXPDMASK`
129			`&& ((dl.l.upper & MANTDMASK) != 0 \|\| dl.l.lower != 0))`
130			`return 1;`
131			`dl.d = b;`
132			`if (EXPD(dl) == EXPDMASK`
133			`&& ((dl.l.upper & MANTDMASK) != 0 \|\| dl.l.lower != 0))`
134			`return 1;`
135			`return 0;`
136			`}`
137
138			`/* convert unsigned int to double */`
139			`double`
140			`__floatunsidf (unsigned long a1)`
141			`{`
142			`long exp = 32 + EXCESSD;`
143			`union double_long dl;`
144
145			`if (!a1)`
146			`{`
147			`dl.l.upper = dl.l.lower = 0;`
148			`return dl.d;`
149			`}`
150
151			`while (a1 < 0x2000000L)`
152			`{`
153			`a1 <<= 4;`
154			`exp -= 4;`
155			`}`
156
157			`while (a1 < 0x80000000L)`
158			`{`
159			`a1 <<= 1;`
160			`exp--;`
161			`}`
162
163			`/* pack up and go home */`
164			`dl.l.upper = exp << 20L;`
165			`dl.l.upper \|= (a1 >> 11L) & ~HIDDEND;`
166			`dl.l.lower = a1 << 21L;`
167
168			`return dl.d;`
169			`}`
170
171			`/* convert int to double */`
172			`double`
173			`__floatsidf (long a1)`
174			`{`
175			`long sign = 0, exp = 31 + EXCESSD;`
176			`union double_long dl;`
177
178			`if (!a1)`
179			`{`
180			`dl.l.upper = dl.l.lower = 0;`
181			`return dl.d;`
182			`}`
183
184			`if (a1 < 0)`
185			`{`
186			`sign = SIGNBIT;`
187			`a1 = (long)-(unsigned long)a1;`
188			`if (a1 < 0)`
189			`{`
190			`dl.l.upper = SIGNBIT \| ((32 + EXCESSD) << 20L);`
191			`dl.l.lower = 0;`
192			`return dl.d;`
193			`}`
194			`}`
195
196			`while (a1 < 0x1000000L)`
197			`{`
198			`a1 <<= 4;`
199			`exp -= 4;`
200			`}`
201
202			`while (a1 < 0x40000000L)`
203			`{`
204			`a1 <<= 1;`
205			`exp--;`
206			`}`
207
208			`/* pack up and go home */`
209			`dl.l.upper = sign;`
210			`dl.l.upper \|= exp << 20L;`
211			`dl.l.upper \|= (a1 >> 10L) & ~HIDDEND;`
212			`dl.l.lower = a1 << 22L;`
213
214			`return dl.d;`
215			`}`
216
217			`/* convert unsigned int to float */`
218			`float`
219			`__floatunsisf (unsigned long l)`
220			`{`
221			`double foo = __floatunsidf (l);`
222			`return foo;`
223			`}`
224
225			`/* convert int to float */`
226			`float`
227			`__floatsisf (long l)`
228			`{`
229			`double foo = __floatsidf (l);`
230			`return foo;`
231			`}`
232
233			`/* convert float to double */`
234			`double`
235			`__extendsfdf2 (float a1)`
236			`{`
237			`register union float_long fl1;`
238			`register union double_long dl;`
239			`register long exp;`
240			`register long mant;`
241
242			`fl1.f = a1;`
243
244			`dl.l.upper = SIGN (fl1.l);`
245			`if ((fl1.l & ~SIGNBIT) == 0)`
246			`{`
247			`dl.l.lower = 0;`
248			`return dl.d;`
249			`}`
250
251			`exp = EXP(fl1.l);`
252			`mant = MANT (fl1.l) & ~HIDDEN;`
253			`if (exp == 0)`
254			`{`
255			`/* Denormal. */`
256			`exp = 1;`
257			`while (!(mant & HIDDEN))`
258			`{`
259			`mant <<= 1;`
260			`exp--;`
261			`}`
262			`mant &= ~HIDDEN;`
263			`}`
264			`exp = exp - EXCESS + EXCESSD;`
265			`dl.l.upper \|= exp << 20;`
266			`dl.l.upper \|= mant >> 3;`
267			`dl.l.lower = mant << 29;`
268
269			`return dl.d;`
270			`}`
271
272			`/* convert double to float */`
273			`float`
274			`__truncdfsf2 (double a1)`
275			`{`
276			`register long exp;`
277			`register long mant;`
278			`register union float_long fl;`
279			`register union double_long dl1;`
280
281			`dl1.d = a1;`
282
283			`if ((dl1.l.upper & ~SIGNBIT) == 0 && !dl1.l.lower)`
284			`{`
285			`fl.l = SIGND(dl1);`
286			`return fl.f;`
287			`}`
288
289			`exp = EXPD (dl1) - EXCESSD + EXCESS;`
290
291			`/* shift double mantissa 6 bits so we can round */`
292			`mant = MANTD (dl1) >> 6;`
293
294			`/* Check for underflow and denormals. */`
295			`if (exp <= 0)`
296			`{`
297			`if (exp < -24)`
298			`mant = 0;`
299			`else`
300			`mant >>= 1 - exp;`
301			`exp = 0;`
302			`}`
303
304			`/* now round and shift down */`
305			`mant += 1;`
306			`mant >>= 1;`
307
308			`/* did the round overflow? */`
309			`if (mant & 0xFF000000L)`
310			`{`
311			`mant >>= 1;`
312			`exp++;`
313			`}`
314
315			`mant &= ~HIDDEN;`
316
317			`/* pack up and go home */`
318			`fl.l = PACK (SIGND (dl1), exp, mant);`
319			`return (fl.f);`
320			`}`
321
322			`/* convert double to int */`
323			`long`
324			`__fixdfsi (double a1)`
325			`{`
326			`register union double_long dl1;`
327			`register long exp;`
328			`register long l;`
329
330			`dl1.d = a1;`
331
332			`if (!dl1.l.upper && !dl1.l.lower)`
333			`return 0;`
334
335			`exp = EXPD (dl1) - EXCESSD - 31;`
336			`l = MANTD (dl1);`
337
338			`if (exp > 0)`
339			`{`
340			`/* Return largest integer. */`
341			`return SIGND (dl1) ? 0x80000000L : 0x7fffffffL;`
342			`}`
343
344			`if (exp <= -32)`
345			`return 0;`
346
347			`/* shift down until exp = 0 */`
348			`if (exp < 0)`
349			`l >>= -exp;`
350
351			`return (SIGND (dl1) ? -l : l);`
352			`}`
353
354			`/* convert float to int */`
355			`long`
356			`__fixsfsi (float a1)`
357			`{`
358			`double foo = a1;`
359			`return __fixdfsi (foo);`
360			`}`
361
362			`#else /* EXTFLOAT */`
363
364			`/* Primitive extended precision floating point support.`
365
366			`We assume all numbers are normalized, don't do any rounding, etc. */`
367
368			`/* Prototypes for the above in case we use them. */`
369			`double __floatunsidf (unsigned long);`
370			`double __floatsidf (long);`
371			`float __floatsisf (long);`
372			`double __extendsfdf2 (float);`
373			`float __truncdfsf2 (double);`
374			`long __fixdfsi (double);`
375			`long __fixsfsi (float);`
376
377			`int`
378			`__unordxf2(long double a, long double b)`
379			`{`
380			`union long_double_long ldl;`
381
382			`ldl.ld = a;`
383			`if (EXPX(ldl) == EXPXMASK`
384			`&& ((ldl.l.middle & MANTXMASK) != 0 \|\| ldl.l.lower != 0))`
385			`return 1;`
386			`ldl.ld = b;`
387			`if (EXPX(ldl) == EXPXMASK`
388			`&& ((ldl.l.middle & MANTXMASK) != 0 \|\| ldl.l.lower != 0))`
389			`return 1;`
390			`return 0;`
391			`}`
392
393			`/* convert double to long double */`
394			`long double`
395			`__extenddfxf2 (double d)`
396			`{`
397			`register union double_long dl;`
398			`register union long_double_long ldl;`
399			`register long exp;`
400
401			`dl.d = d;`
402			`/printf ("dfxf in: %g\n", d);/`
403
404			`ldl.l.upper = SIGND (dl);`
405			`if ((dl.l.upper & ~SIGNBIT) == 0 && !dl.l.lower)`
406			`{`
407			`ldl.l.middle = 0;`
408			`ldl.l.lower = 0;`
409			`return ldl.ld;`
410			`}`
411
412			`exp = EXPD (dl) - EXCESSD + EXCESSX;`
413			`ldl.l.upper \|= exp << 16;`
414			`ldl.l.middle = HIDDENX;`
415			`/* 31-20: # mantissa bits in ldl.l.middle - # mantissa bits in dl.l.upper */`
416			`ldl.l.middle \|= (dl.l.upper & MANTDMASK) << (31 - 20);`
417			`/* 1+20: explicit-integer-bit + # mantissa bits in dl.l.upper */`
418			`ldl.l.middle \|= dl.l.lower >> (1 + 20);`
419			`/* 32 - 21: # bits of dl.l.lower in ldl.l.middle */`
420			`ldl.l.lower = dl.l.lower << (32 - 21);`
421
422			`/printf ("dfxf out: %s\n", dumpxf (ldl.ld));/`
423			`return ldl.ld;`
424			`}`
425
426			`/* convert long double to double */`
427			`double`
428			`__truncxfdf2 (long double ld)`
429			`{`
430			`register long exp;`
431			`register union double_long dl;`
432			`register union long_double_long ldl;`
433
434			`ldl.ld = ld;`
435			`/printf ("xfdf in: %s\n", dumpxf (ld));/`
436
437			`dl.l.upper = SIGNX (ldl);`
438			`if ((ldl.l.upper & ~SIGNBIT) == 0 && !ldl.l.middle && !ldl.l.lower)`
439			`{`
440			`dl.l.lower = 0;`
441			`return dl.d;`
442			`}`
443
444			`exp = EXPX (ldl) - EXCESSX + EXCESSD;`
445			/* ??? quick and dirty: keep `exp' sane */
446			`if (exp >= EXPDMASK)`
447			`exp = EXPDMASK - 1;`
448			`dl.l.upper \|= exp << (32 - (EXPDBITS + 1));`
449			`/* +1-1: add one for sign bit, but take one off for explicit-integer-bit */`
450			`dl.l.upper \|= (ldl.l.middle & MANTXMASK) >> (EXPDBITS + 1 - 1);`
451			`dl.l.lower = (ldl.l.middle & MANTXMASK) << (32 - (EXPDBITS + 1 - 1));`
452			`dl.l.lower \|= ldl.l.lower >> (EXPDBITS + 1 - 1);`
453
454			`/printf ("xfdf out: %g\n", dl.d);/`
455			`return dl.d;`
456			`}`
457
458			`/* convert a float to a long double */`
459			`long double`
460			`__extendsfxf2 (float f)`
461			`{`
462			`long double foo = __extenddfxf2 (__extendsfdf2 (f));`
463			`return foo;`
464			`}`
465
466			`/* convert a long double to a float */`
467			`float`
468			`__truncxfsf2 (long double ld)`
469			`{`
470			`float foo = __truncdfsf2 (__truncxfdf2 (ld));`
471			`return foo;`
472			`}`
473
474			`/* convert an int to a long double */`
475			`long double`
476			`__floatsixf (long l)`
477			`{`
478			`double foo = __floatsidf (l);`
479			`return foo;`
480			`}`
481
482			`/* convert an unsigned int to a long double */`
483			`long double`
484			`__floatunsixf (unsigned long l)`
485			`{`
486			`double foo = __floatunsidf (l);`
487			`return foo;`
488			`}`
489
490			`/* convert a long double to an int */`
491			`long`
492			`__fixxfsi (long double ld)`
493			`{`
494			`long foo = __fixdfsi ((double) ld);`
495			`return foo;`
496			`}`
497
498			`/* The remaining provide crude math support by working in double precision. */`
499
500			`long double`
501			`__addxf3 (long double x1, long double x2)`
502			`{`
503			`return (double) x1 + (double) x2;`
504			`}`
505
506			`long double`
507			`__subxf3 (long double x1, long double x2)`
508			`{`
509			`return (double) x1 - (double) x2;`
510			`}`
511
512			`long double`
513			`__mulxf3 (long double x1, long double x2)`
514			`{`
515			`return (double) x1 * (double) x2;`
516			`}`
517
518			`long double`
519			`__divxf3 (long double x1, long double x2)`
520			`{`
521			`return (double) x1 / (double) x2;`
522			`}`
523
524			`long double`
525			`__negxf2 (long double x1)`
526			`{`
527			`return - (double) x1;`
528			`}`
529
530			`long`
531			`__cmpxf2 (long double x1, long double x2)`
532			`{`
533			`return __cmpdf2 ((double) x1, (double) x2);`
534			`}`
535
536			`long`
537			`__eqxf2 (long double x1, long double x2)`
538			`{`
539			`return __cmpdf2 ((double) x1, (double) x2);`
540			`}`
541
542			`long`
543			`__nexf2 (long double x1, long double x2)`
544			`{`
545			`return __cmpdf2 ((double) x1, (double) x2);`
546			`}`
547
548			`long`
549			`__ltxf2 (long double x1, long double x2)`
550			`{`
551			`return __cmpdf2 ((double) x1, (double) x2);`
552			`}`
553
554			`long`
555			`__lexf2 (long double x1, long double x2)`
556			`{`
557			`return __cmpdf2 ((double) x1, (double) x2);`
558			`}`
559
560			`long`
561			`__gtxf2 (long double x1, long double x2)`
562			`{`
563			`return __cmpdf2 ((double) x1, (double) x2);`
564			`}`
565
566			`long`
567			`__gexf2 (long double x1, long double x2)`
568			`{`
569			`return __cmpdf2 ((double) x1, (double) x2);`
570			`}`
571
572			`#endif /* EXTFLOAT */`