1 #include <stdio.h>
2 #include <complex.h>
3
4 #ifndef PREF
5 #define PREF c
6 #endif
7
8 #define FX(N,P) P ## _ ## N
9 #define MK(N,P) FX(P,N)
10
11 #define N 32
12 #define TYPE double
13
14 // ------ FMA
15
16 // Complex FMA instructions rotating the result
17
18 __attribute__((noinline,noipa))
19 void MK(fma0, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
20 {
21 for (int i=0; i < N; i++)
22 c[i] += a[i] * b[i];
23 }
24
25 __attribute__((noinline,noipa))
26 void MK(fma90, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
27 {
28 for (int i=0; i < N; i++)
29 c[i] += a[i] * b[i] * I;
30 }
31
32 __attribute__((noinline,noipa))
33 void MK(fma180, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
34 {
35 for (int i=0; i < N; i++)
36 c[i] += a[i] * b[i] * I * I;
37 }
38
39 __attribute__((noinline,noipa))
40 void MK(fma270, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
41 {
42 for (int i=0; i < N; i++)
43 c[i] += a[i] * b[i] * I * I * I;
44 }
45
46 // Complex FMA instructions rotating the second parameter.
47
48
49 __attribute__((noinline,noipa))
50 void MK(fma0_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
51 {
52 for (int i=0; i < N; i++)
53 c[i] += a[i] * b[i];
54 }
55
56 __attribute__((noinline,noipa))
57 void MK(fma90_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
58 {
59 for (int i=0; i < N; i++)
60 c[i] += a[i] * (b[i] * I);
61 }
62
63 __attribute__((noinline,noipa))
64 void MK(fma180_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
65 {
66 for (int i=0; i < N; i++)
67 c[i] += a[i] * (b[i] * I * I);
68 }
69
70 __attribute__((noinline,noipa))
71 void MK(fma270_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
72 {
73 for (int i=0; i < N; i++)
74 c[i] += a[i] * (b[i] * I * I * I);
75 }
76
77 // Complex FMA instructions with conjucated values.
78
79
80 __attribute__((noinline,noipa))
81 void MK(fma_conj_first, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
82 {
83 for (int i=0; i < N; i++)
84 c[i] += conj (a[i]) * b[i];
85 }
86
87 __attribute__((noinline,noipa))
88 void MK(fma_conj_second, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
89 {
90 for (int i=0; i < N; i++)
91 c[i] += a[i] * conj (b[i]);
92 }
93
94 __attribute__((noinline,noipa))
95 void MK(fma_conj_both, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
96 {
97 for (int i=0; i < N; i++)
98 c[i] += conj (a[i]) * conj (b[i]);
99 }
100
101 // ----- FMS
102
103 // Complex FMS instructions rotating the result
104
105 __attribute__((noinline,noipa))
106 void MK(fms0, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
107 {
108 for (int i=0; i < N; i++)
109 c[i] -= a[i] * b[i];
110 }
111
112 __attribute__((noinline,noipa))
113 void MK(fms90, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
114 {
115 for (int i=0; i < N; i++)
116 c[i] -= a[i] * b[i] * I;
117 }
118
119 __attribute__((noinline,noipa))
120 void MK(fms180, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
121 {
122 for (int i=0; i < N; i++)
123 c[i] -= a[i] * b[i] * I * I;
124 }
125
126 __attribute__((noinline,noipa))
127 void MK(fms270, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
128 {
129 for (int i=0; i < N; i++)
130 c[i] -= a[i] * b[i] * I * I * I;
131 }
132
133 // Complex FMS instructions rotating the second parameter.
134
135 __attribute__((noinline,noipa))
136 void MK(fms0_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
137 {
138 for (int i=0; i < N; i++)
139 c[i] -= a[i] * b[i];
140 }
141
142 __attribute__((noinline,noipa))
143 void MK(fms90_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
144 {
145 for (int i=0; i < N; i++)
146 c[i] -= a[i] * (b[i] * I);
147 }
148
149 __attribute__((noinline,noipa))
150 void MK(fms180_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
151 {
152 for (int i=0; i < N; i++)
153 c[i] -= a[i] * (b[i] * I * I);
154 }
155
156 __attribute__((noinline,noipa))
157 void MK(fms270_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
158 {
159 for (int i=0; i < N; i++)
160 c[i] -= a[i] * (b[i] * I * I * I);
161 }
162
163 // Complex FMS instructions with conjucated values.
164
165 __attribute__((noinline,noipa))
166 void MK(fms_conj_first, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
167 {
168 for (int i=0; i < N; i++)
169 c[i] -= conj (a[i]) * b[i];
170 }
171
172 __attribute__((noinline,noipa))
173 void MK(fms_conj_second, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
174 {
175 for (int i=0; i < N; i++)
176 c[i] -= a[i] * conj (b[i]);
177 }
178
179 __attribute__((noinline,noipa))
180 void MK(fms_conj_both, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
181 {
182 for (int i=0; i < N; i++)
183 c[i] -= conj (a[i]) * conj (b[i]);
184 }
185
186
187 // ----- MUL
188
189 // Complex MUL instructions rotating the result
190
191 __attribute__((noinline,noipa))
192 void MK(mul0, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
193 {
194 for (int i=0; i < N; i++)
195 c[i] = a[i] * b[i];
196 }
197
198 __attribute__((noinline,noipa))
199 void MK(mul90, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
200 {
201 for (int i=0; i < N; i++)
202 c[i] = a[i] * b[i] * I;
203 }
204
205 __attribute__((noinline,noipa))
206 void MK(mul180, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
207 {
208 for (int i=0; i < N; i++)
209 c[i] = a[i] * b[i] * I * I;
210 }
211
212 __attribute__((noinline,noipa))
213 void MK(mul270, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
214 {
215 for (int i=0; i < N; i++)
216 c[i] = a[i] * b[i] * I * I * I;
217 }
218
219 // Complex MUL instructions rotating the second parameter.
220
221 __attribute__((noinline,noipa))
222 void MK(mul0_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
223 {
224 for (int i=0; i < N; i++)
225 c[i] = a[i] * b[i];
226 }
227
228 __attribute__((noinline,noipa))
229 void MK(mul90_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
230 {
231 for (int i=0; i < N; i++)
232 c[i] = a[i] * (b[i] * I);
233 }
234
235 __attribute__((noinline,noipa))
236 void MK(mul180_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
237 {
238 for (int i=0; i < N; i++)
239 c[i] = a[i] * (b[i] * I * I);
240 }
241
242 __attribute__((noinline,noipa))
243 void MK(mul270_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
244 {
245 for (int i=0; i < N; i++)
246 c[i] = a[i] * (b[i] * I * I * I);
247 }
248
249 // Complex FMS instructions with conjucated values.
250
251 __attribute__((noinline,noipa))
252 void MK(mul_conj_first, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
253 {
254 for (int i=0; i < N; i++)
255 c[i] = conj (a[i]) * b[i];
256 }
257
258 __attribute__((noinline,noipa))
259 void MK(mul_conj_second, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
260 {
261 for (int i=0; i < N; i++)
262 c[i] = a[i] * conj (b[i]);
263 }
264
265 __attribute__((noinline,noipa))
266 void MK(mul_conj_both, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
267 {
268 for (int i=0; i < N; i++)
269 c[i] = conj (a[i]) * conj (b[i]);
270 }
271
272
273 // ----- ADD
274
275 // Complex ADD instructions rotating the result
276
277 __attribute__((noinline,noipa))
278 void MK(add0, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
279 {
280 for (int i=0; i < N; i++)
281 c[i] = a[i] + b[i];
282 }
283
284 __attribute__((noinline,noipa))
285 void MK(add90, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
286 {
287 for (int i=0; i < N; i++)
288 c[i] = (a[i] + b[i]) * I;
289 }
290
291 __attribute__((noinline,noipa))
292 void MK(add180, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
293 {
294 for (int i=0; i < N; i++)
295 c[i] = (a[i] + b[i]) * I * I;
296 }
297
298 __attribute__((noinline,noipa))
299 void MK(add270, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
300 {
301 for (int i=0; i < N; i++)
302 c[i] = (a[i] + b[i]) * I * I * I;
303 }
304
305 // Complex ADD instructions rotating the second parameter.
306
307 __attribute__((noinline,noipa))
308 void MK(add0_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
309 {
310 for (int i=0; i < N; i++)
311 c[i] = a[i] + b[i];
312 }
313
314 __attribute__((noinline,noipa))
315 void MK(add90_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
316 {
317 for (int i=0; i < N; i++)
318 c[i] = a[i] + (b[i] * I);
319 }
320
321 __attribute__((noinline,noipa))
322 void MK(add180_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
323 {
324 for (int i=0; i < N; i++)
325 c[i] = a[i] + (b[i] * I * I);
326 }
327
328 __attribute__((noinline,noipa))
329 void MK(add270_snd, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
330 {
331 for (int i=0; i < N; i++)
332 c[i] = a[i] + (b[i] * I * I * I);
333 }
334
335 // Complex ADD instructions with conjucated values.
336
337 __attribute__((noinline,noipa))
338 void MK(add_conj_first, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
339 {
340 for (int i=0; i < N; i++)
341 c[i] = conj (a[i]) + b[i];
342 }
343
344 __attribute__((noinline,noipa))
345 void MK(add_conj_second, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
346 {
347 for (int i=0; i < N; i++)
348 c[i] = a[i] + conj (b[i]);
349 }
350
351 __attribute__((noinline,noipa))
352 void MK(add_conj_both, PREF) (TYPE complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
353 {
354 for (int i=0; i < N; i++)
355 c[i] = conj (a[i]) + conj (b[i]);
356 }
357
358