Subversion Repositories QNX 8.QNX8 LLVM/Clang compiler suite

Rev

Details | Last modification | View Log | RSS feed

Rev Author Line No. Line
14 pmbaty 1
/*===-- __clang_cuda_complex_builtins - CUDA impls of runtime complex fns ---===
2
 *
3
 * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4
 * See https://llvm.org/LICENSE.txt for license information.
5
 * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6
 *
7
 *===-----------------------------------------------------------------------===
8
 */
9
 
10
#ifndef __CLANG_CUDA_COMPLEX_BUILTINS
11
#define __CLANG_CUDA_COMPLEX_BUILTINS
12
 
13
// This header defines __muldc3, __mulsc3, __divdc3, and __divsc3.  These are
14
// libgcc functions that clang assumes are available when compiling c99 complex
15
// operations.  (These implementations come from libc++, and have been modified
16
// to work with CUDA and OpenMP target offloading [in C and C++ mode].)
17
 
18
#pragma push_macro("__DEVICE__")
19
#if defined(__OPENMP_NVPTX__) || defined(__OPENMP_AMDGCN__)
20
#pragma omp declare target
21
#define __DEVICE__ __attribute__((noinline, nothrow, cold, weak))
22
#else
23
#define __DEVICE__ __device__ inline
24
#endif
25
 
26
// To make the algorithms available for C and C++ in CUDA and OpenMP we select
27
// different but equivalent function versions. TODO: For OpenMP we currently
28
// select the native builtins as the overload support for templates is lacking.
29
#if !defined(__OPENMP_NVPTX__) && !defined(__OPENMP_AMDGCN__)
30
#define _ISNANd std::isnan
31
#define _ISNANf std::isnan
32
#define _ISINFd std::isinf
33
#define _ISINFf std::isinf
34
#define _ISFINITEd std::isfinite
35
#define _ISFINITEf std::isfinite
36
#define _COPYSIGNd std::copysign
37
#define _COPYSIGNf std::copysign
38
#define _SCALBNd std::scalbn
39
#define _SCALBNf std::scalbn
40
#define _ABSd std::abs
41
#define _ABSf std::abs
42
#define _LOGBd std::logb
43
#define _LOGBf std::logb
44
// Rather than pulling in std::max from algorithm everytime, use available ::max.
45
#define _fmaxd max
46
#define _fmaxf max
47
#else
48
#ifdef __AMDGCN__
49
#define _ISNANd __ocml_isnan_f64
50
#define _ISNANf __ocml_isnan_f32
51
#define _ISINFd __ocml_isinf_f64
52
#define _ISINFf __ocml_isinf_f32
53
#define _ISFINITEd __ocml_isfinite_f64
54
#define _ISFINITEf __ocml_isfinite_f32
55
#define _COPYSIGNd __ocml_copysign_f64
56
#define _COPYSIGNf __ocml_copysign_f32
57
#define _SCALBNd __ocml_scalbn_f64
58
#define _SCALBNf __ocml_scalbn_f32
59
#define _ABSd __ocml_fabs_f64
60
#define _ABSf __ocml_fabs_f32
61
#define _LOGBd __ocml_logb_f64
62
#define _LOGBf __ocml_logb_f32
63
#define _fmaxd __ocml_fmax_f64
64
#define _fmaxf __ocml_fmax_f32
65
#else
66
#define _ISNANd __nv_isnand
67
#define _ISNANf __nv_isnanf
68
#define _ISINFd __nv_isinfd
69
#define _ISINFf __nv_isinff
70
#define _ISFINITEd __nv_isfinited
71
#define _ISFINITEf __nv_finitef
72
#define _COPYSIGNd __nv_copysign
73
#define _COPYSIGNf __nv_copysignf
74
#define _SCALBNd __nv_scalbn
75
#define _SCALBNf __nv_scalbnf
76
#define _ABSd __nv_fabs
77
#define _ABSf __nv_fabsf
78
#define _LOGBd __nv_logb
79
#define _LOGBf __nv_logbf
80
#define _fmaxd __nv_fmax
81
#define _fmaxf __nv_fmaxf
82
#endif
83
#endif
84
 
85
#if defined(__cplusplus)
86
extern "C" {
87
#endif
88
 
89
__DEVICE__ double _Complex __muldc3(double __a, double __b, double __c,
90
                                    double __d) {
91
  double __ac = __a * __c;
92
  double __bd = __b * __d;
93
  double __ad = __a * __d;
94
  double __bc = __b * __c;
95
  double _Complex z;
96
  __real__(z) = __ac - __bd;
97
  __imag__(z) = __ad + __bc;
98
  if (_ISNANd(__real__(z)) && _ISNANd(__imag__(z))) {
99
    int __recalc = 0;
100
    if (_ISINFd(__a) || _ISINFd(__b)) {
101
      __a = _COPYSIGNd(_ISINFd(__a) ? 1 : 0, __a);
102
      __b = _COPYSIGNd(_ISINFd(__b) ? 1 : 0, __b);
103
      if (_ISNANd(__c))
104
        __c = _COPYSIGNd(0, __c);
105
      if (_ISNANd(__d))
106
        __d = _COPYSIGNd(0, __d);
107
      __recalc = 1;
108
    }
109
    if (_ISINFd(__c) || _ISINFd(__d)) {
110
      __c = _COPYSIGNd(_ISINFd(__c) ? 1 : 0, __c);
111
      __d = _COPYSIGNd(_ISINFd(__d) ? 1 : 0, __d);
112
      if (_ISNANd(__a))
113
        __a = _COPYSIGNd(0, __a);
114
      if (_ISNANd(__b))
115
        __b = _COPYSIGNd(0, __b);
116
      __recalc = 1;
117
    }
118
    if (!__recalc &&
119
        (_ISINFd(__ac) || _ISINFd(__bd) || _ISINFd(__ad) || _ISINFd(__bc))) {
120
      if (_ISNANd(__a))
121
        __a = _COPYSIGNd(0, __a);
122
      if (_ISNANd(__b))
123
        __b = _COPYSIGNd(0, __b);
124
      if (_ISNANd(__c))
125
        __c = _COPYSIGNd(0, __c);
126
      if (_ISNANd(__d))
127
        __d = _COPYSIGNd(0, __d);
128
      __recalc = 1;
129
    }
130
    if (__recalc) {
131
      // Can't use std::numeric_limits<double>::infinity() -- that doesn't have
132
      // a device overload (and isn't constexpr before C++11, naturally).
133
      __real__(z) = __builtin_huge_val() * (__a * __c - __b * __d);
134
      __imag__(z) = __builtin_huge_val() * (__a * __d + __b * __c);
135
    }
136
  }
137
  return z;
138
}
139
 
140
__DEVICE__ float _Complex __mulsc3(float __a, float __b, float __c, float __d) {
141
  float __ac = __a * __c;
142
  float __bd = __b * __d;
143
  float __ad = __a * __d;
144
  float __bc = __b * __c;
145
  float _Complex z;
146
  __real__(z) = __ac - __bd;
147
  __imag__(z) = __ad + __bc;
148
  if (_ISNANf(__real__(z)) && _ISNANf(__imag__(z))) {
149
    int __recalc = 0;
150
    if (_ISINFf(__a) || _ISINFf(__b)) {
151
      __a = _COPYSIGNf(_ISINFf(__a) ? 1 : 0, __a);
152
      __b = _COPYSIGNf(_ISINFf(__b) ? 1 : 0, __b);
153
      if (_ISNANf(__c))
154
        __c = _COPYSIGNf(0, __c);
155
      if (_ISNANf(__d))
156
        __d = _COPYSIGNf(0, __d);
157
      __recalc = 1;
158
    }
159
    if (_ISINFf(__c) || _ISINFf(__d)) {
160
      __c = _COPYSIGNf(_ISINFf(__c) ? 1 : 0, __c);
161
      __d = _COPYSIGNf(_ISINFf(__d) ? 1 : 0, __d);
162
      if (_ISNANf(__a))
163
        __a = _COPYSIGNf(0, __a);
164
      if (_ISNANf(__b))
165
        __b = _COPYSIGNf(0, __b);
166
      __recalc = 1;
167
    }
168
    if (!__recalc &&
169
        (_ISINFf(__ac) || _ISINFf(__bd) || _ISINFf(__ad) || _ISINFf(__bc))) {
170
      if (_ISNANf(__a))
171
        __a = _COPYSIGNf(0, __a);
172
      if (_ISNANf(__b))
173
        __b = _COPYSIGNf(0, __b);
174
      if (_ISNANf(__c))
175
        __c = _COPYSIGNf(0, __c);
176
      if (_ISNANf(__d))
177
        __d = _COPYSIGNf(0, __d);
178
      __recalc = 1;
179
    }
180
    if (__recalc) {
181
      __real__(z) = __builtin_huge_valf() * (__a * __c - __b * __d);
182
      __imag__(z) = __builtin_huge_valf() * (__a * __d + __b * __c);
183
    }
184
  }
185
  return z;
186
}
187
 
188
__DEVICE__ double _Complex __divdc3(double __a, double __b, double __c,
189
                                    double __d) {
190
  int __ilogbw = 0;
191
  // Can't use std::max, because that's defined in <algorithm>, and we don't
192
  // want to pull that in for every compile.  The CUDA headers define
193
  // ::max(float, float) and ::max(double, double), which is sufficient for us.
194
  double __logbw = _LOGBd(_fmaxd(_ABSd(__c), _ABSd(__d)));
195
  if (_ISFINITEd(__logbw)) {
196
    __ilogbw = (int)__logbw;
197
    __c = _SCALBNd(__c, -__ilogbw);
198
    __d = _SCALBNd(__d, -__ilogbw);
199
  }
200
  double __denom = __c * __c + __d * __d;
201
  double _Complex z;
202
  __real__(z) = _SCALBNd((__a * __c + __b * __d) / __denom, -__ilogbw);
203
  __imag__(z) = _SCALBNd((__b * __c - __a * __d) / __denom, -__ilogbw);
204
  if (_ISNANd(__real__(z)) && _ISNANd(__imag__(z))) {
205
    if ((__denom == 0.0) && (!_ISNANd(__a) || !_ISNANd(__b))) {
206
      __real__(z) = _COPYSIGNd(__builtin_huge_val(), __c) * __a;
207
      __imag__(z) = _COPYSIGNd(__builtin_huge_val(), __c) * __b;
208
    } else if ((_ISINFd(__a) || _ISINFd(__b)) && _ISFINITEd(__c) &&
209
               _ISFINITEd(__d)) {
210
      __a = _COPYSIGNd(_ISINFd(__a) ? 1.0 : 0.0, __a);
211
      __b = _COPYSIGNd(_ISINFd(__b) ? 1.0 : 0.0, __b);
212
      __real__(z) = __builtin_huge_val() * (__a * __c + __b * __d);
213
      __imag__(z) = __builtin_huge_val() * (__b * __c - __a * __d);
214
    } else if (_ISINFd(__logbw) && __logbw > 0.0 && _ISFINITEd(__a) &&
215
               _ISFINITEd(__b)) {
216
      __c = _COPYSIGNd(_ISINFd(__c) ? 1.0 : 0.0, __c);
217
      __d = _COPYSIGNd(_ISINFd(__d) ? 1.0 : 0.0, __d);
218
      __real__(z) = 0.0 * (__a * __c + __b * __d);
219
      __imag__(z) = 0.0 * (__b * __c - __a * __d);
220
    }
221
  }
222
  return z;
223
}
224
 
225
__DEVICE__ float _Complex __divsc3(float __a, float __b, float __c, float __d) {
226
  int __ilogbw = 0;
227
  float __logbw = _LOGBf(_fmaxf(_ABSf(__c), _ABSf(__d)));
228
  if (_ISFINITEf(__logbw)) {
229
    __ilogbw = (int)__logbw;
230
    __c = _SCALBNf(__c, -__ilogbw);
231
    __d = _SCALBNf(__d, -__ilogbw);
232
  }
233
  float __denom = __c * __c + __d * __d;
234
  float _Complex z;
235
  __real__(z) = _SCALBNf((__a * __c + __b * __d) / __denom, -__ilogbw);
236
  __imag__(z) = _SCALBNf((__b * __c - __a * __d) / __denom, -__ilogbw);
237
  if (_ISNANf(__real__(z)) && _ISNANf(__imag__(z))) {
238
    if ((__denom == 0) && (!_ISNANf(__a) || !_ISNANf(__b))) {
239
      __real__(z) = _COPYSIGNf(__builtin_huge_valf(), __c) * __a;
240
      __imag__(z) = _COPYSIGNf(__builtin_huge_valf(), __c) * __b;
241
    } else if ((_ISINFf(__a) || _ISINFf(__b)) && _ISFINITEf(__c) &&
242
               _ISFINITEf(__d)) {
243
      __a = _COPYSIGNf(_ISINFf(__a) ? 1 : 0, __a);
244
      __b = _COPYSIGNf(_ISINFf(__b) ? 1 : 0, __b);
245
      __real__(z) = __builtin_huge_valf() * (__a * __c + __b * __d);
246
      __imag__(z) = __builtin_huge_valf() * (__b * __c - __a * __d);
247
    } else if (_ISINFf(__logbw) && __logbw > 0 && _ISFINITEf(__a) &&
248
               _ISFINITEf(__b)) {
249
      __c = _COPYSIGNf(_ISINFf(__c) ? 1 : 0, __c);
250
      __d = _COPYSIGNf(_ISINFf(__d) ? 1 : 0, __d);
251
      __real__(z) = 0 * (__a * __c + __b * __d);
252
      __imag__(z) = 0 * (__b * __c - __a * __d);
253
    }
254
  }
255
  return z;
256
}
257
 
258
#if defined(__cplusplus)
259
} // extern "C"
260
#endif
261
 
262
#undef _ISNANd
263
#undef _ISNANf
264
#undef _ISINFd
265
#undef _ISINFf
266
#undef _COPYSIGNd
267
#undef _COPYSIGNf
268
#undef _ISFINITEd
269
#undef _ISFINITEf
270
#undef _SCALBNd
271
#undef _SCALBNf
272
#undef _ABSd
273
#undef _ABSf
274
#undef _LOGBd
275
#undef _LOGBf
276
#undef _fmaxd
277
#undef _fmaxf
278
 
279
#if defined(__OPENMP_NVPTX__) || defined(__OPENMP_AMDGCN__)
280
#pragma omp end declare target
281
#endif
282
 
283
#pragma pop_macro("__DEVICE__")
284
 
285
#endif // __CLANG_CUDA_COMPLEX_BUILTINS