summaryrefslogtreecommitdiffstats
path: root/arch/powerpc/math-emu/op-4.h
blob: c9ae626070daf1d9d6e05e7c9882c5dc5757a092 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
/*
 * Basic four-word fraction declaration and manipulation.
 *
 * When adding quadword support for 32 bit machines, we need
 * to be a little careful as double multiply uses some of these
 * macros: (in op-2.h)
 * _FP_MUL_MEAT_2_wide() uses _FP_FRAC_DECL_4, _FP_FRAC_WORD_4,
 * _FP_FRAC_ADD_4, _FP_FRAC_SRS_4
 * _FP_MUL_MEAT_2_gmp() uses _FP_FRAC_SRS_4 (and should use
 * _FP_FRAC_DECL_4: it appears to be broken and is not used
 * anywhere anyway. )
 *
 * I've now fixed all the macros that were here from the sparc64 code.
 * [*none* of the shift macros were correct!] -- PMM 02/1998
 *
 * The only quadword stuff that remains to be coded is:
 * 1) the conversion to/from ints, which requires
 * that we check (in op-common.h) that the following do the right thing
 * for quadwords: _FP_TO_INT(Q,4,r,X,rsz,rsg), _FP_FROM_INT(Q,4,X,r,rs,rt)
 * 2) multiply, divide and sqrt, which require:
 * _FP_MUL_MEAT_4_*(R,X,Y), _FP_DIV_MEAT_4_*(R,X,Y), _FP_SQRT_MEAT_4(R,S,T,X,q),
 * This also needs _FP_MUL_MEAT_Q and _FP_DIV_MEAT_Q to be defined to
 * some suitable _FP_MUL_MEAT_4_* macros in sfp-machine.h.
 * [we're free to choose whatever FP_MUL_MEAT_4_* macros we need for
 * these; they are used nowhere else. ]
 */

#define _FP_FRAC_DECL_4(X)	_FP_W_TYPE X##_f[4]
#define _FP_FRAC_COPY_4(D,S)			\
  (D##_f[0] = S##_f[0], D##_f[1] = S##_f[1],	\
   D##_f[2] = S##_f[2], D##_f[3] = S##_f[3])
/* The _FP_FRAC_SET_n(X,I) macro is intended for use with another
 * macro such as _FP_ZEROFRAC_n which returns n comma separated values.
 * The result is that we get an expansion of __FP_FRAC_SET_n(X,I0,I1,I2,I3)
 * which just assigns the In values to the array X##_f[].
 * This is why the number of parameters doesn't appear to match
 * at first glance...      -- PMM
 */
#define _FP_FRAC_SET_4(X,I)	__FP_FRAC_SET_4(X, I)
#define _FP_FRAC_HIGH_4(X)	(X##_f[3])
#define _FP_FRAC_LOW_4(X)	(X##_f[0])
#define _FP_FRAC_WORD_4(X,w)	(X##_f[w])

#define _FP_FRAC_SLL_4(X,N)						\
  do {									\
    _FP_I_TYPE _up, _down, _skip, _i;					\
    _skip = (N) / _FP_W_TYPE_SIZE;					\
    _up = (N) % _FP_W_TYPE_SIZE;					\
    _down = _FP_W_TYPE_SIZE - _up;					\
    for (_i = 3; _i > _skip; --_i)					\
      X##_f[_i] = X##_f[_i-_skip] << _up | X##_f[_i-_skip-1] >> _down;	\
/* bugfixed: was X##_f[_i] <<= _up;  -- PMM 02/1998 */                  \
    X##_f[_i] = X##_f[0] << _up; 	                                \
    for (--_i; _i >= 0; --_i)						\
      X##_f[_i] = 0;							\
  } while (0)

/* This one was broken too */
#define _FP_FRAC_SRL_4(X,N)						\
  do {									\
    _FP_I_TYPE _up, _down, _skip, _i;					\
    _skip = (N) / _FP_W_TYPE_SIZE;					\
    _down = (N) % _FP_W_TYPE_SIZE;					\
    _up = _FP_W_TYPE_SIZE - _down;					\
    for (_i = 0; _i < 3-_skip; ++_i)					\
      X##_f[_i] = X##_f[_i+_skip] >> _down | X##_f[_i+_skip+1] << _up;	\
    X##_f[_i] = X##_f[3] >> _down;			         	\
    for (++_i; _i < 4; ++_i)						\
      X##_f[_i] = 0;							\
  } while (0)


/* Right shift with sticky-lsb.
 * What this actually means is that we do a standard right-shift,
 * but that if any of the bits that fall off the right hand side
 * were one then we always set the LSbit.
 */
#define _FP_FRAC_SRS_4(X,N,size)					\
  do {									\
    _FP_I_TYPE _up, _down, _skip, _i;					\
    _FP_W_TYPE _s;							\
    _skip = (N) / _FP_W_TYPE_SIZE;					\
    _down = (N) % _FP_W_TYPE_SIZE;					\
    _up = _FP_W_TYPE_SIZE - _down;					\
    for (_s = _i = 0; _i < _skip; ++_i)					\
      _s |= X##_f[_i];							\
    _s |= X##_f[_i] << _up;						\
/* s is now != 0 if we want to set the LSbit */                         \
    for (_i = 0; _i < 3-_skip; ++_i)					\
      X##_f[_i] = X##_f[_i+_skip] >> _down | X##_f[_i+_skip+1] << _up;	\
    X##_f[_i] = X##_f[3] >> _down;					\
    for (++_i; _i < 4; ++_i)						\
      X##_f[_i] = 0;							\
    /* don't fix the LSB until the very end when we're sure f[0] is stable */ \
    X##_f[0] |= (_s != 0);                                              \
  } while (0)

#define _FP_FRAC_ADD_4(R,X,Y)						\
  __FP_FRAC_ADD_4(R##_f[3], R##_f[2], R##_f[1], R##_f[0],		\
		  X##_f[3], X##_f[2], X##_f[1], X##_f[0],		\
		  Y##_f[3], Y##_f[2], Y##_f[1], Y##_f[0])

#define _FP_FRAC_SUB_4(R,X,Y)                                           \
  __FP_FRAC_SUB_4(R##_f[3], R##_f[2], R##_f[1], R##_f[0],		\
		  X##_f[3], X##_f[2], X##_f[1], X##_f[0],		\
		  Y##_f[3], Y##_f[2], Y##_f[1], Y##_f[0])

#define _FP_FRAC_ADDI_4(X,I)                                            \
  __FP_FRAC_ADDI_4(X##_f[3], X##_f[2], X##_f[1], X##_f[0], I)

#define _FP_ZEROFRAC_4  0,0,0,0
#define _FP_MINFRAC_4   0,0,0,1

#define _FP_FRAC_ZEROP_4(X)     ((X##_f[0] | X##_f[1] | X##_f[2] | X##_f[3]) == 0)
#define _FP_FRAC_NEGP_4(X)      ((_FP_WS_TYPE)X##_f[3] < 0)
#define _FP_FRAC_OVERP_4(fs,X)  (X##_f[0] & _FP_OVERFLOW_##fs)

#define _FP_FRAC_EQ_4(X,Y)                              \
 (X##_f[0] == Y##_f[0] && X##_f[1] == Y##_f[1]          \
  && X##_f[2] == Y##_f[2] && X##_f[3] == Y##_f[3])

#define _FP_FRAC_GT_4(X,Y)                              \
 (X##_f[3] > Y##_f[3] ||                                \
  (X##_f[3] == Y##_f[3] && (X##_f[2] > Y##_f[2] ||      \
   (X##_f[2] == Y##_f[2] && (X##_f[1] > Y##_f[1] ||     \
    (X##_f[1] == Y##_f[1] && X##_f[0] > Y##_f[0])       \
   ))                                                   \
  ))                                                    \
 )

#define _FP_FRAC_GE_4(X,Y)                              \
 (X##_f[3] > Y##_f[3] ||                                \
  (X##_f[3] == Y##_f[3] && (X##_f[2] > Y##_f[2] ||      \
   (X##_f[2] == Y##_f[2] && (X##_f[1] > Y##_f[1] ||     \
    (X##_f[1] == Y##_f[1] && X##_f[0] >= Y##_f[0])      \
   ))                                                   \
  ))                                                    \
 )


#define _FP_FRAC_CLZ_4(R,X)             \
  do {                                  \
    if (X##_f[3])                       \
    {                                   \
        __FP_CLZ(R,X##_f[3]);           \
    }                                   \
    else if (X##_f[2])                  \
    {                                   \
        __FP_CLZ(R,X##_f[2]);           \
        R += _FP_W_TYPE_SIZE;           \
    }                                   \
    else if (X##_f[1])                  \
    {                                   \
        __FP_CLZ(R,X##_f[2]);           \
        R += _FP_W_TYPE_SIZE*2;         \
    }                                   \
    else                                \
    {                                   \
        __FP_CLZ(R,X##_f[0]);           \
        R += _FP_W_TYPE_SIZE*3;         \
    }                                   \
  } while(0)


#define _FP_UNPACK_RAW_4(fs, X, val)                            \
  do {                                                          \
    union _FP_UNION_##fs _flo; _flo.flt = (val);        	\
    X##_f[0] = _flo.bits.frac0;                                 \
    X##_f[1] = _flo.bits.frac1;                                 \
    X##_f[2] = _flo.bits.frac2;                                 \
    X##_f[3] = _flo.bits.frac3;                                 \
    X##_e  = _flo.bits.exp;                                     \
    X##_s  = _flo.bits.sign;                                    \
  } while (0)

#define _FP_PACK_RAW_4(fs, val, X)                              \
  do {                                                          \
    union _FP_UNION_##fs _flo;					\
    _flo.bits.frac0 = X##_f[0];                                 \
    _flo.bits.frac1 = X##_f[1];                                 \
    _flo.bits.frac2 = X##_f[2];                                 \
    _flo.bits.frac3 = X##_f[3];                                 \
    _flo.bits.exp   = X##_e;                                    \
    _flo.bits.sign  = X##_s;                                    \
    (val) = _flo.flt;                                   	\
  } while (0)


/*
 * Internals
 */

#define __FP_FRAC_SET_4(X,I3,I2,I1,I0)					\
  (X##_f[3] = I3, X##_f[2] = I2, X##_f[1] = I1, X##_f[0] = I0)

#ifndef __FP_FRAC_ADD_4
#define __FP_FRAC_ADD_4(r3,r2,r1,r0,x3,x2,x1,x0,y3,y2,y1,y0)	\
  do {								\
    int _c1, _c2, _c3;						\
    r0 = x0 + y0;						\
    _c1 = r0 < x0;						\
    r1 = x1 + y1;						\
    _c2 = r1 < x1;						\
    r1 += _c1;							\
    _c2 |= r1 < _c1;						\
    r2 = x2 + y2;						\
    _c3 = r2 < x2;						\
    r2 += _c2;							\
    _c3 |= r2 < _c2;						\
    r3 = x3 + y3 + _c3;						\
  } while (0)
#endif

#ifndef __FP_FRAC_SUB_4
#define __FP_FRAC_SUB_4(r3,r2,r1,r0,x3,x2,x1,x0,y3,y2,y1,y0)	\
  do {								\
    int _c1, _c2, _c3;						\
    r0 = x0 - y0;						\
    _c1 = r0 > x0;						\
    r1 = x1 - y1;						\
    _c2 = r1 > x1;						\
    r1 -= _c1;							\
    _c2 |= r1 > _c1;						\
    r2 = x2 - y2;						\
    _c3 = r2 > x2;						\
    r2 -= _c2;							\
    _c3 |= r2 > _c2;						\
    r3 = x3 - y3 - _c3;						\
  } while (0)
#endif

#ifndef __FP_FRAC_ADDI_4
/* I always wanted to be a lisp programmer :-> */
#define __FP_FRAC_ADDI_4(x3,x2,x1,x0,i)                                 \
  (x3 += ((x2 += ((x1 += ((x0 += i) < x0)) < x1) < x2)))
#endif

/* Convert FP values between word sizes. This appears to be more
 * complicated than I'd have expected it to be, so these might be
 * wrong... These macros are in any case somewhat bogus because they
 * use information about what various FRAC_n variables look like
 * internally [eg, that 2 word vars are X_f0 and x_f1]. But so do
 * the ones in op-2.h and op-1.h.
 */
#define _FP_FRAC_CONV_1_4(dfs, sfs, D, S)                               \
   do {                                                                 \
     _FP_FRAC_SRS_4(S, (_FP_WFRACBITS_##sfs - _FP_WFRACBITS_##dfs),     \
                        _FP_WFRACBITS_##sfs);                           \
     D##_f = S##_f[0];                                                   \
  } while (0)

#define _FP_FRAC_CONV_2_4(dfs, sfs, D, S)                               \
   do {                                                                 \
     _FP_FRAC_SRS_4(S, (_FP_WFRACBITS_##sfs - _FP_WFRACBITS_##dfs),     \
                        _FP_WFRACBITS_##sfs);                           \
     D##_f0 = S##_f[0];                                                  \
     D##_f1 = S##_f[1];                                                  \
  } while (0)

/* Assembly/disassembly for converting to/from integral types.
 * No shifting or overflow handled here.
 */
/* Put the FP value X into r, which is an integer of size rsize. */
#define _FP_FRAC_ASSEMBLE_4(r, X, rsize)                                \
  do {                                                                  \
    if (rsize <= _FP_W_TYPE_SIZE)                                       \
      r = X##_f[0];                                                     \
    else if (rsize <= 2*_FP_W_TYPE_SIZE)                                \
    {                                                                   \
      r = X##_f[1];                                                     \
      r <<= _FP_W_TYPE_SIZE;                                            \
      r += X##_f[0];                                                    \
    }                                                                   \
    else                                                                \
    {                                                                   \
      /* I'm feeling lazy so we deal with int == 3words (implausible)*/ \
      /* and int == 4words as a single case.                         */ \
      r = X##_f[3];                                                     \
      r <<= _FP_W_TYPE_SIZE;                                            \
      r += X##_f[2];                                                    \
      r <<= _FP_W_TYPE_SIZE;                                            \
      r += X##_f[1];                                                    \
      r <<= _FP_W_TYPE_SIZE;                                            \
      r += X##_f[0];                                                    \
    }                                                                   \
  } while (0)

/* "No disassemble Number Five!" */
/* move an integer of size rsize into X's fractional part. We rely on
 * the _f[] array consisting of words of size _FP_W_TYPE_SIZE to avoid
 * having to mask the values we store into it.
 */
#define _FP_FRAC_DISASSEMBLE_4(X, r, rsize)                             \
  do {                                                                  \
    X##_f[0] = r;                                                       \
    X##_f[1] = (rsize <= _FP_W_TYPE_SIZE ? 0 : r >> _FP_W_TYPE_SIZE);   \
    X##_f[2] = (rsize <= 2*_FP_W_TYPE_SIZE ? 0 : r >> 2*_FP_W_TYPE_SIZE); \
    X##_f[3] = (rsize <= 3*_FP_W_TYPE_SIZE ? 0 : r >> 3*_FP_W_TYPE_SIZE); \
  } while (0)

#define _FP_FRAC_CONV_4_1(dfs, sfs, D, S)                               \
   do {                                                                 \
     D##_f[0] = S##_f;                                                  \
     D##_f[1] = D##_f[2] = D##_f[3] = 0;                                \
     _FP_FRAC_SLL_4(D, (_FP_WFRACBITS_##dfs - _FP_WFRACBITS_##sfs));    \
   } while (0)

#define _FP_FRAC_CONV_4_2(dfs, sfs, D, S)                               \
   do {                                                                 \
     D##_f[0] = S##_f0;                                                 \
     D##_f[1] = S##_f1;                                                 \
     D##_f[2] = D##_f[3] = 0;                                           \
     _FP_FRAC_SLL_4(D, (_FP_WFRACBITS_##dfs - _FP_WFRACBITS_##sfs));    \
   } while (0)

/* FIXME! This has to be written */
#define _FP_SQRT_MEAT_4(R, S, T, X, q)
OpenPOWER on IntegriCloud