/src/fftw3/dft/scalar/codelets/n1_12.c
Line | Count | Source |
1 | | /* |
2 | | * Copyright (c) 2003, 2007-14 Matteo Frigo |
3 | | * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology |
4 | | * |
5 | | * This program is free software; you can redistribute it and/or modify |
6 | | * it under the terms of the GNU General Public License as published by |
7 | | * the Free Software Foundation; either version 2 of the License, or |
8 | | * (at your option) any later version. |
9 | | * |
10 | | * This program is distributed in the hope that it will be useful, |
11 | | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
12 | | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
13 | | * GNU General Public License for more details. |
14 | | * |
15 | | * You should have received a copy of the GNU General Public License |
16 | | * along with this program; if not, write to the Free Software |
17 | | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
18 | | * |
19 | | */ |
20 | | |
21 | | /* This file was automatically generated --- DO NOT EDIT */ |
22 | | /* Generated on Sat Nov 15 06:07:43 UTC 2025 */ |
23 | | |
24 | | #include "dft/codelet-dft.h" |
25 | | |
26 | | #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) |
27 | | |
28 | | /* Generated by: ../../../genfft/gen_notw.native -fma -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include dft/scalar/n.h */ |
29 | | |
30 | | /* |
31 | | * This function contains 96 FP additions, 24 FP multiplications, |
32 | | * (or, 72 additions, 0 multiplications, 24 fused multiply/add), |
33 | | * 43 stack variables, 2 constants, and 48 memory accesses |
34 | | */ |
35 | | #include "dft/scalar/n.h" |
36 | | |
37 | | static void n1_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) |
38 | | { |
39 | | DK(KP866025403, +0.866025403784438646763723170752936183471402627); |
40 | | DK(KP500000000, +0.500000000000000000000000000000000000000000000); |
41 | | { |
42 | | INT i; |
43 | | for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(48, is), MAKE_VOLATILE_STRIDE(48, os)) { |
44 | | E T5, TR, TA, Ts, TS, Tz, Ta, TU, TD, Tx, TV, TC, Tg, T1d, TG; |
45 | | E TJ, T1u, T1c, Tl, T1i, TL, TO, T1v, T1h; |
46 | | { |
47 | | E T1, T2, T3, T4; |
48 | | T1 = ri[0]; |
49 | | T2 = ri[WS(is, 4)]; |
50 | | T3 = ri[WS(is, 8)]; |
51 | | T4 = T2 + T3; |
52 | | T5 = T1 + T4; |
53 | | TR = FNMS(KP500000000, T4, T1); |
54 | | TA = T3 - T2; |
55 | | } |
56 | | { |
57 | | E To, Tp, Tq, Tr; |
58 | | To = ii[0]; |
59 | | Tp = ii[WS(is, 4)]; |
60 | | Tq = ii[WS(is, 8)]; |
61 | | Tr = Tp + Tq; |
62 | | Ts = To + Tr; |
63 | | TS = Tp - Tq; |
64 | | Tz = FNMS(KP500000000, Tr, To); |
65 | | } |
66 | | { |
67 | | E T6, T7, T8, T9; |
68 | | T6 = ri[WS(is, 6)]; |
69 | | T7 = ri[WS(is, 10)]; |
70 | | T8 = ri[WS(is, 2)]; |
71 | | T9 = T7 + T8; |
72 | | Ta = T6 + T9; |
73 | | TU = FNMS(KP500000000, T9, T6); |
74 | | TD = T8 - T7; |
75 | | } |
76 | | { |
77 | | E Tt, Tu, Tv, Tw; |
78 | | Tt = ii[WS(is, 6)]; |
79 | | Tu = ii[WS(is, 10)]; |
80 | | Tv = ii[WS(is, 2)]; |
81 | | Tw = Tu + Tv; |
82 | | Tx = Tt + Tw; |
83 | | TV = Tu - Tv; |
84 | | TC = FNMS(KP500000000, Tw, Tt); |
85 | | } |
86 | | { |
87 | | E Tc, Td, Te, Tf; |
88 | | Tc = ri[WS(is, 3)]; |
89 | | Td = ri[WS(is, 7)]; |
90 | | Te = ri[WS(is, 11)]; |
91 | | Tf = Td + Te; |
92 | | Tg = Tc + Tf; |
93 | | T1d = Te - Td; |
94 | | TG = FNMS(KP500000000, Tf, Tc); |
95 | | } |
96 | | { |
97 | | E T1a, TH, TI, T1b; |
98 | | T1a = ii[WS(is, 3)]; |
99 | | TH = ii[WS(is, 7)]; |
100 | | TI = ii[WS(is, 11)]; |
101 | | T1b = TH + TI; |
102 | | TJ = TH - TI; |
103 | | T1u = T1a + T1b; |
104 | | T1c = FNMS(KP500000000, T1b, T1a); |
105 | | } |
106 | | { |
107 | | E Th, Ti, Tj, Tk; |
108 | | Th = ri[WS(is, 9)]; |
109 | | Ti = ri[WS(is, 1)]; |
110 | | Tj = ri[WS(is, 5)]; |
111 | | Tk = Ti + Tj; |
112 | | Tl = Th + Tk; |
113 | | T1i = Tj - Ti; |
114 | | TL = FNMS(KP500000000, Tk, Th); |
115 | | } |
116 | | { |
117 | | E T1f, TM, TN, T1g; |
118 | | T1f = ii[WS(is, 9)]; |
119 | | TM = ii[WS(is, 1)]; |
120 | | TN = ii[WS(is, 5)]; |
121 | | T1g = TM + TN; |
122 | | TO = TM - TN; |
123 | | T1v = T1f + T1g; |
124 | | T1h = FNMS(KP500000000, T1g, T1f); |
125 | | } |
126 | | { |
127 | | E Tb, Tm, T1t, T1w; |
128 | | Tb = T5 + Ta; |
129 | | Tm = Tg + Tl; |
130 | | ro[WS(os, 6)] = Tb - Tm; |
131 | | ro[0] = Tb + Tm; |
132 | | { |
133 | | E T1x, T1y, Tn, Ty; |
134 | | T1x = Ts + Tx; |
135 | | T1y = T1u + T1v; |
136 | | io[WS(os, 6)] = T1x - T1y; |
137 | | io[0] = T1x + T1y; |
138 | | Tn = Tg - Tl; |
139 | | Ty = Ts - Tx; |
140 | | io[WS(os, 3)] = Tn + Ty; |
141 | | io[WS(os, 9)] = Ty - Tn; |
142 | | } |
143 | | T1t = T5 - Ta; |
144 | | T1w = T1u - T1v; |
145 | | ro[WS(os, 3)] = T1t - T1w; |
146 | | ro[WS(os, 9)] = T1t + T1w; |
147 | | { |
148 | | E T11, T1l, T1k, T1m, T14, T18, T17, T19; |
149 | | { |
150 | | E TZ, T10, T1e, T1j; |
151 | | TZ = FMA(KP866025403, TA, Tz); |
152 | | T10 = FMA(KP866025403, TD, TC); |
153 | | T11 = TZ - T10; |
154 | | T1l = TZ + T10; |
155 | | T1e = FMA(KP866025403, T1d, T1c); |
156 | | T1j = FMA(KP866025403, T1i, T1h); |
157 | | T1k = T1e - T1j; |
158 | | T1m = T1e + T1j; |
159 | | } |
160 | | { |
161 | | E T12, T13, T15, T16; |
162 | | T12 = FMA(KP866025403, TJ, TG); |
163 | | T13 = FMA(KP866025403, TO, TL); |
164 | | T14 = T12 - T13; |
165 | | T18 = T12 + T13; |
166 | | T15 = FMA(KP866025403, TS, TR); |
167 | | T16 = FMA(KP866025403, TV, TU); |
168 | | T17 = T15 + T16; |
169 | | T19 = T15 - T16; |
170 | | } |
171 | | io[WS(os, 1)] = T11 - T14; |
172 | | ro[WS(os, 1)] = T19 + T1k; |
173 | | io[WS(os, 7)] = T11 + T14; |
174 | | ro[WS(os, 7)] = T19 - T1k; |
175 | | ro[WS(os, 10)] = T17 - T18; |
176 | | io[WS(os, 10)] = T1l - T1m; |
177 | | ro[WS(os, 4)] = T17 + T18; |
178 | | io[WS(os, 4)] = T1l + T1m; |
179 | | } |
180 | | { |
181 | | E TF, T1r, T1q, T1s, TQ, TY, TX, T1n; |
182 | | { |
183 | | E TB, TE, T1o, T1p; |
184 | | TB = FNMS(KP866025403, TA, Tz); |
185 | | TE = FNMS(KP866025403, TD, TC); |
186 | | TF = TB - TE; |
187 | | T1r = TB + TE; |
188 | | T1o = FNMS(KP866025403, T1d, T1c); |
189 | | T1p = FNMS(KP866025403, T1i, T1h); |
190 | | T1q = T1o - T1p; |
191 | | T1s = T1o + T1p; |
192 | | } |
193 | | { |
194 | | E TK, TP, TT, TW; |
195 | | TK = FNMS(KP866025403, TJ, TG); |
196 | | TP = FNMS(KP866025403, TO, TL); |
197 | | TQ = TK - TP; |
198 | | TY = TK + TP; |
199 | | TT = FNMS(KP866025403, TS, TR); |
200 | | TW = FNMS(KP866025403, TV, TU); |
201 | | TX = TT + TW; |
202 | | T1n = TT - TW; |
203 | | } |
204 | | io[WS(os, 5)] = TF - TQ; |
205 | | ro[WS(os, 5)] = T1n + T1q; |
206 | | io[WS(os, 11)] = TF + TQ; |
207 | | ro[WS(os, 11)] = T1n - T1q; |
208 | | ro[WS(os, 2)] = TX - TY; |
209 | | io[WS(os, 2)] = T1r - T1s; |
210 | | ro[WS(os, 8)] = TX + TY; |
211 | | io[WS(os, 8)] = T1r + T1s; |
212 | | } |
213 | | } |
214 | | } |
215 | | } |
216 | | } |
217 | | |
218 | | static const kdft_desc desc = { 12, "n1_12", { 72, 0, 24, 0 }, &GENUS, 0, 0, 0, 0 }; |
219 | | |
220 | | void X(codelet_n1_12) (planner *p) { X(kdft_register) (p, n1_12, &desc); |
221 | | } |
222 | | |
223 | | #else |
224 | | |
225 | | /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include dft/scalar/n.h */ |
226 | | |
227 | | /* |
228 | | * This function contains 96 FP additions, 16 FP multiplications, |
229 | | * (or, 88 additions, 8 multiplications, 8 fused multiply/add), |
230 | | * 43 stack variables, 2 constants, and 48 memory accesses |
231 | | */ |
232 | | #include "dft/scalar/n.h" |
233 | | |
234 | | static void n1_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) |
235 | 191 | { |
236 | 191 | DK(KP866025403, +0.866025403784438646763723170752936183471402627); |
237 | 191 | DK(KP500000000, +0.500000000000000000000000000000000000000000000); |
238 | 191 | { |
239 | 191 | INT i; |
240 | 1.09k | for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(48, is), MAKE_VOLATILE_STRIDE(48, os)) { |
241 | 904 | E T5, TR, TA, Ts, TS, Tz, Ta, TU, TD, Tx, TV, TC, Tg, T1a, TG; |
242 | 904 | E TJ, T1u, T1d, Tl, T1f, TL, TO, T1v, T1i; |
243 | 904 | { |
244 | 904 | E T1, T2, T3, T4; |
245 | 904 | T1 = ri[0]; |
246 | 904 | T2 = ri[WS(is, 4)]; |
247 | 904 | T3 = ri[WS(is, 8)]; |
248 | 904 | T4 = T2 + T3; |
249 | 904 | T5 = T1 + T4; |
250 | 904 | TR = FNMS(KP500000000, T4, T1); |
251 | 904 | TA = KP866025403 * (T3 - T2); |
252 | 904 | } |
253 | 904 | { |
254 | 904 | E To, Tp, Tq, Tr; |
255 | 904 | To = ii[0]; |
256 | 904 | Tp = ii[WS(is, 4)]; |
257 | 904 | Tq = ii[WS(is, 8)]; |
258 | 904 | Tr = Tp + Tq; |
259 | 904 | Ts = To + Tr; |
260 | 904 | TS = KP866025403 * (Tp - Tq); |
261 | 904 | Tz = FNMS(KP500000000, Tr, To); |
262 | 904 | } |
263 | 904 | { |
264 | 904 | E T6, T7, T8, T9; |
265 | 904 | T6 = ri[WS(is, 6)]; |
266 | 904 | T7 = ri[WS(is, 10)]; |
267 | 904 | T8 = ri[WS(is, 2)]; |
268 | 904 | T9 = T7 + T8; |
269 | 904 | Ta = T6 + T9; |
270 | 904 | TU = FNMS(KP500000000, T9, T6); |
271 | 904 | TD = KP866025403 * (T8 - T7); |
272 | 904 | } |
273 | 904 | { |
274 | 904 | E Tt, Tu, Tv, Tw; |
275 | 904 | Tt = ii[WS(is, 6)]; |
276 | 904 | Tu = ii[WS(is, 10)]; |
277 | 904 | Tv = ii[WS(is, 2)]; |
278 | 904 | Tw = Tu + Tv; |
279 | 904 | Tx = Tt + Tw; |
280 | 904 | TV = KP866025403 * (Tu - Tv); |
281 | 904 | TC = FNMS(KP500000000, Tw, Tt); |
282 | 904 | } |
283 | 904 | { |
284 | 904 | E Tc, Td, Te, Tf; |
285 | 904 | Tc = ri[WS(is, 3)]; |
286 | 904 | Td = ri[WS(is, 7)]; |
287 | 904 | Te = ri[WS(is, 11)]; |
288 | 904 | Tf = Td + Te; |
289 | 904 | Tg = Tc + Tf; |
290 | 904 | T1a = KP866025403 * (Te - Td); |
291 | 904 | TG = FNMS(KP500000000, Tf, Tc); |
292 | 904 | } |
293 | 904 | { |
294 | 904 | E T1b, TH, TI, T1c; |
295 | 904 | T1b = ii[WS(is, 3)]; |
296 | 904 | TH = ii[WS(is, 7)]; |
297 | 904 | TI = ii[WS(is, 11)]; |
298 | 904 | T1c = TH + TI; |
299 | 904 | TJ = KP866025403 * (TH - TI); |
300 | 904 | T1u = T1b + T1c; |
301 | 904 | T1d = FNMS(KP500000000, T1c, T1b); |
302 | 904 | } |
303 | 904 | { |
304 | 904 | E Th, Ti, Tj, Tk; |
305 | 904 | Th = ri[WS(is, 9)]; |
306 | 904 | Ti = ri[WS(is, 1)]; |
307 | 904 | Tj = ri[WS(is, 5)]; |
308 | 904 | Tk = Ti + Tj; |
309 | 904 | Tl = Th + Tk; |
310 | 904 | T1f = KP866025403 * (Tj - Ti); |
311 | 904 | TL = FNMS(KP500000000, Tk, Th); |
312 | 904 | } |
313 | 904 | { |
314 | 904 | E T1g, TM, TN, T1h; |
315 | 904 | T1g = ii[WS(is, 9)]; |
316 | 904 | TM = ii[WS(is, 1)]; |
317 | 904 | TN = ii[WS(is, 5)]; |
318 | 904 | T1h = TM + TN; |
319 | 904 | TO = KP866025403 * (TM - TN); |
320 | 904 | T1v = T1g + T1h; |
321 | 904 | T1i = FNMS(KP500000000, T1h, T1g); |
322 | 904 | } |
323 | 904 | { |
324 | 904 | E Tb, Tm, T1t, T1w; |
325 | 904 | Tb = T5 + Ta; |
326 | 904 | Tm = Tg + Tl; |
327 | 904 | ro[WS(os, 6)] = Tb - Tm; |
328 | 904 | ro[0] = Tb + Tm; |
329 | 904 | { |
330 | 904 | E T1x, T1y, Tn, Ty; |
331 | 904 | T1x = Ts + Tx; |
332 | 904 | T1y = T1u + T1v; |
333 | 904 | io[WS(os, 6)] = T1x - T1y; |
334 | 904 | io[0] = T1x + T1y; |
335 | 904 | Tn = Tg - Tl; |
336 | 904 | Ty = Ts - Tx; |
337 | 904 | io[WS(os, 3)] = Tn + Ty; |
338 | 904 | io[WS(os, 9)] = Ty - Tn; |
339 | 904 | } |
340 | 904 | T1t = T5 - Ta; |
341 | 904 | T1w = T1u - T1v; |
342 | 904 | ro[WS(os, 3)] = T1t - T1w; |
343 | 904 | ro[WS(os, 9)] = T1t + T1w; |
344 | 904 | { |
345 | 904 | E T11, T1l, T1k, T1m, T14, T18, T17, T19; |
346 | 904 | { |
347 | 904 | E TZ, T10, T1e, T1j; |
348 | 904 | TZ = TA + Tz; |
349 | 904 | T10 = TD + TC; |
350 | 904 | T11 = TZ - T10; |
351 | 904 | T1l = TZ + T10; |
352 | 904 | T1e = T1a + T1d; |
353 | 904 | T1j = T1f + T1i; |
354 | 904 | T1k = T1e - T1j; |
355 | 904 | T1m = T1e + T1j; |
356 | 904 | } |
357 | 904 | { |
358 | 904 | E T12, T13, T15, T16; |
359 | 904 | T12 = TG + TJ; |
360 | 904 | T13 = TL + TO; |
361 | 904 | T14 = T12 - T13; |
362 | 904 | T18 = T12 + T13; |
363 | 904 | T15 = TR + TS; |
364 | 904 | T16 = TU + TV; |
365 | 904 | T17 = T15 + T16; |
366 | 904 | T19 = T15 - T16; |
367 | 904 | } |
368 | 904 | io[WS(os, 1)] = T11 - T14; |
369 | 904 | ro[WS(os, 1)] = T19 + T1k; |
370 | 904 | io[WS(os, 7)] = T11 + T14; |
371 | 904 | ro[WS(os, 7)] = T19 - T1k; |
372 | 904 | ro[WS(os, 10)] = T17 - T18; |
373 | 904 | io[WS(os, 10)] = T1l - T1m; |
374 | 904 | ro[WS(os, 4)] = T17 + T18; |
375 | 904 | io[WS(os, 4)] = T1l + T1m; |
376 | 904 | } |
377 | 904 | { |
378 | 904 | E TF, T1r, T1q, T1s, TQ, TY, TX, T1n; |
379 | 904 | { |
380 | 904 | E TB, TE, T1o, T1p; |
381 | 904 | TB = Tz - TA; |
382 | 904 | TE = TC - TD; |
383 | 904 | TF = TB - TE; |
384 | 904 | T1r = TB + TE; |
385 | 904 | T1o = T1d - T1a; |
386 | 904 | T1p = T1i - T1f; |
387 | 904 | T1q = T1o - T1p; |
388 | 904 | T1s = T1o + T1p; |
389 | 904 | } |
390 | 904 | { |
391 | 904 | E TK, TP, TT, TW; |
392 | 904 | TK = TG - TJ; |
393 | 904 | TP = TL - TO; |
394 | 904 | TQ = TK - TP; |
395 | 904 | TY = TK + TP; |
396 | 904 | TT = TR - TS; |
397 | 904 | TW = TU - TV; |
398 | 904 | TX = TT + TW; |
399 | 904 | T1n = TT - TW; |
400 | 904 | } |
401 | 904 | io[WS(os, 5)] = TF - TQ; |
402 | 904 | ro[WS(os, 5)] = T1n + T1q; |
403 | 904 | io[WS(os, 11)] = TF + TQ; |
404 | 904 | ro[WS(os, 11)] = T1n - T1q; |
405 | 904 | ro[WS(os, 2)] = TX - TY; |
406 | 904 | io[WS(os, 2)] = T1r - T1s; |
407 | 904 | ro[WS(os, 8)] = TX + TY; |
408 | 904 | io[WS(os, 8)] = T1r + T1s; |
409 | 904 | } |
410 | 904 | } |
411 | 904 | } |
412 | 191 | } |
413 | 191 | } |
414 | | |
415 | | static const kdft_desc desc = { 12, "n1_12", { 88, 8, 8, 0 }, &GENUS, 0, 0, 0, 0 }; |
416 | | |
417 | 1 | void X(codelet_n1_12) (planner *p) { X(kdft_register) (p, n1_12, &desc); |
418 | 1 | } |
419 | | |
420 | | #endif |