/src/fftw3/rdft/scalar/r2cb/hb_12.c
Line | Count | Source (jump to first uncovered line) |
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 Sun Sep 8 06:42:11 UTC 2024 */ |
23 | | |
24 | | #include "rdft/codelet-rdft.h" |
25 | | |
26 | | #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) |
27 | | |
28 | | /* Generated by: ../../../genfft/gen_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -dif -name hb_12 -include rdft/scalar/hb.h */ |
29 | | |
30 | | /* |
31 | | * This function contains 118 FP additions, 68 FP multiplications, |
32 | | * (or, 72 additions, 22 multiplications, 46 fused multiply/add), |
33 | | * 47 stack variables, 2 constants, and 48 memory accesses |
34 | | */ |
35 | | #include "rdft/scalar/hb.h" |
36 | | |
37 | | static void hb_12(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms) |
38 | | { |
39 | | DK(KP866025403, +0.866025403784438646763723170752936183471402627); |
40 | | DK(KP500000000, +0.500000000000000000000000000000000000000000000); |
41 | | { |
42 | | INT m; |
43 | | for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) { |
44 | | E T18, T20, T1b, T21, T1s, T2a, T1p, T29, TI, TN, TO, Tb, To, T1f, T23; |
45 | | E T1i, T24, T1z, T2d, T1w, T2c, Tt, Ty, Tz, Tm, TD; |
46 | | { |
47 | | E T1, TE, TM, T6, T4, T1o, TH, T17, TL, T1a, T9, T1r; |
48 | | T1 = cr[0]; |
49 | | TE = ci[WS(rs, 11)]; |
50 | | TM = cr[WS(rs, 6)]; |
51 | | T6 = ci[WS(rs, 5)]; |
52 | | { |
53 | | E T2, T3, TF, TG; |
54 | | T2 = cr[WS(rs, 4)]; |
55 | | T3 = ci[WS(rs, 3)]; |
56 | | T4 = T2 + T3; |
57 | | T1o = T2 - T3; |
58 | | TF = ci[WS(rs, 7)]; |
59 | | TG = cr[WS(rs, 8)]; |
60 | | TH = TF - TG; |
61 | | T17 = TF + TG; |
62 | | } |
63 | | { |
64 | | E TJ, TK, T7, T8; |
65 | | TJ = ci[WS(rs, 9)]; |
66 | | TK = cr[WS(rs, 10)]; |
67 | | TL = TJ - TK; |
68 | | T1a = TJ + TK; |
69 | | T7 = ci[WS(rs, 1)]; |
70 | | T8 = cr[WS(rs, 2)]; |
71 | | T9 = T7 + T8; |
72 | | T1r = T7 - T8; |
73 | | } |
74 | | { |
75 | | E T16, T19, T1q, T1n, T5, Ta; |
76 | | T16 = FNMS(KP500000000, T4, T1); |
77 | | T18 = FNMS(KP866025403, T17, T16); |
78 | | T20 = FMA(KP866025403, T17, T16); |
79 | | T19 = FNMS(KP500000000, T9, T6); |
80 | | T1b = FMA(KP866025403, T1a, T19); |
81 | | T21 = FNMS(KP866025403, T1a, T19); |
82 | | T1q = FMA(KP500000000, TL, TM); |
83 | | T1s = FNMS(KP866025403, T1r, T1q); |
84 | | T2a = FMA(KP866025403, T1r, T1q); |
85 | | T1n = FNMS(KP500000000, TH, TE); |
86 | | T1p = FMA(KP866025403, T1o, T1n); |
87 | | T29 = FNMS(KP866025403, T1o, T1n); |
88 | | TI = TE + TH; |
89 | | TN = TL - TM; |
90 | | TO = TI - TN; |
91 | | T5 = T1 + T4; |
92 | | Ta = T6 + T9; |
93 | | Tb = T5 + Ta; |
94 | | To = T5 - Ta; |
95 | | } |
96 | | } |
97 | | { |
98 | | E Tc, Tp, Tx, Th, Tf, T1v, Ts, T1e, Tw, T1h, Tk, T1y; |
99 | | Tc = cr[WS(rs, 3)]; |
100 | | Tp = ci[WS(rs, 8)]; |
101 | | Tx = cr[WS(rs, 9)]; |
102 | | Th = ci[WS(rs, 2)]; |
103 | | { |
104 | | E Td, Te, Tq, Tr; |
105 | | Td = ci[WS(rs, 4)]; |
106 | | Te = ci[0]; |
107 | | Tf = Td + Te; |
108 | | T1v = Td - Te; |
109 | | Tq = cr[WS(rs, 7)]; |
110 | | Tr = cr[WS(rs, 11)]; |
111 | | Ts = Tq + Tr; |
112 | | T1e = Tq - Tr; |
113 | | } |
114 | | { |
115 | | E Tu, Tv, Ti, Tj; |
116 | | Tu = ci[WS(rs, 10)]; |
117 | | Tv = ci[WS(rs, 6)]; |
118 | | Tw = Tu + Tv; |
119 | | T1h = Tv - Tu; |
120 | | Ti = cr[WS(rs, 1)]; |
121 | | Tj = cr[WS(rs, 5)]; |
122 | | Tk = Ti + Tj; |
123 | | T1y = Ti - Tj; |
124 | | } |
125 | | { |
126 | | E T1d, T1g, T1x, T1u, Tg, Tl; |
127 | | T1d = FNMS(KP500000000, Tf, Tc); |
128 | | T1f = FMA(KP866025403, T1e, T1d); |
129 | | T23 = FNMS(KP866025403, T1e, T1d); |
130 | | T1g = FNMS(KP500000000, Tk, Th); |
131 | | T1i = FMA(KP866025403, T1h, T1g); |
132 | | T24 = FNMS(KP866025403, T1h, T1g); |
133 | | T1x = FMA(KP500000000, Tw, Tx); |
134 | | T1z = FNMS(KP866025403, T1y, T1x); |
135 | | T2d = FMA(KP866025403, T1y, T1x); |
136 | | T1u = FMA(KP500000000, Ts, Tp); |
137 | | T1w = FMA(KP866025403, T1v, T1u); |
138 | | T2c = FNMS(KP866025403, T1v, T1u); |
139 | | Tt = Tp - Ts; |
140 | | Ty = Tw - Tx; |
141 | | Tz = Tt - Ty; |
142 | | Tg = Tc + Tf; |
143 | | Tl = Th + Tk; |
144 | | Tm = Tg + Tl; |
145 | | TD = Tg - Tl; |
146 | | } |
147 | | } |
148 | | cr[0] = Tb + Tm; |
149 | | { |
150 | | E TA, TP, TB, TQ, Tn, TC; |
151 | | TA = To - Tz; |
152 | | TP = TD + TO; |
153 | | Tn = W[16]; |
154 | | TB = Tn * TA; |
155 | | TQ = Tn * TP; |
156 | | TC = W[17]; |
157 | | cr[WS(rs, 9)] = FNMS(TC, TP, TB); |
158 | | ci[WS(rs, 9)] = FMA(TC, TA, TQ); |
159 | | } |
160 | | { |
161 | | E TS, TV, TT, TW, TR, TU; |
162 | | TS = To + Tz; |
163 | | TV = TO - TD; |
164 | | TR = W[4]; |
165 | | TT = TR * TS; |
166 | | TW = TR * TV; |
167 | | TU = W[5]; |
168 | | cr[WS(rs, 3)] = FNMS(TU, TV, TT); |
169 | | ci[WS(rs, 3)] = FMA(TU, TS, TW); |
170 | | } |
171 | | { |
172 | | E T11, T12, T13, TX, TZ, T10, T14, TY; |
173 | | T11 = TI + TN; |
174 | | T12 = Tt + Ty; |
175 | | T13 = T11 - T12; |
176 | | TY = Tb - Tm; |
177 | | TX = W[10]; |
178 | | TZ = TX * TY; |
179 | | T10 = W[11]; |
180 | | T14 = T10 * TY; |
181 | | ci[0] = T11 + T12; |
182 | | ci[WS(rs, 6)] = FMA(TX, T13, T14); |
183 | | cr[WS(rs, 6)] = FNMS(T10, T13, TZ); |
184 | | } |
185 | | { |
186 | | E T1k, T1E, T1B, T1H; |
187 | | { |
188 | | E T1c, T1j, T1t, T1A; |
189 | | T1c = T18 + T1b; |
190 | | T1j = T1f + T1i; |
191 | | T1k = T1c - T1j; |
192 | | T1E = T1c + T1j; |
193 | | T1t = T1p - T1s; |
194 | | T1A = T1w - T1z; |
195 | | T1B = T1t - T1A; |
196 | | T1H = T1t + T1A; |
197 | | } |
198 | | { |
199 | | E T15, T1l, T1m, T1C; |
200 | | T15 = W[18]; |
201 | | T1l = T15 * T1k; |
202 | | T1m = W[19]; |
203 | | T1C = T1m * T1k; |
204 | | cr[WS(rs, 10)] = FNMS(T1m, T1B, T1l); |
205 | | ci[WS(rs, 10)] = FMA(T15, T1B, T1C); |
206 | | } |
207 | | { |
208 | | E T1D, T1F, T1G, T1I; |
209 | | T1D = W[6]; |
210 | | T1F = T1D * T1E; |
211 | | T1G = W[7]; |
212 | | T1I = T1G * T1E; |
213 | | cr[WS(rs, 4)] = FNMS(T1G, T1H, T1F); |
214 | | ci[WS(rs, 4)] = FMA(T1D, T1H, T1I); |
215 | | } |
216 | | } |
217 | | { |
218 | | E T26, T2i, T2f, T2l; |
219 | | { |
220 | | E T22, T25, T2b, T2e; |
221 | | T22 = T20 + T21; |
222 | | T25 = T23 + T24; |
223 | | T26 = T22 - T25; |
224 | | T2i = T22 + T25; |
225 | | T2b = T29 - T2a; |
226 | | T2e = T2c - T2d; |
227 | | T2f = T2b - T2e; |
228 | | T2l = T2b + T2e; |
229 | | } |
230 | | { |
231 | | E T1Z, T27, T28, T2g; |
232 | | T1Z = W[2]; |
233 | | T27 = T1Z * T26; |
234 | | T28 = W[3]; |
235 | | T2g = T28 * T26; |
236 | | cr[WS(rs, 2)] = FNMS(T28, T2f, T27); |
237 | | ci[WS(rs, 2)] = FMA(T1Z, T2f, T2g); |
238 | | } |
239 | | { |
240 | | E T2h, T2j, T2k, T2m; |
241 | | T2h = W[14]; |
242 | | T2j = T2h * T2i; |
243 | | T2k = W[15]; |
244 | | T2m = T2k * T2i; |
245 | | cr[WS(rs, 8)] = FNMS(T2k, T2l, T2j); |
246 | | ci[WS(rs, 8)] = FMA(T2h, T2l, T2m); |
247 | | } |
248 | | } |
249 | | { |
250 | | E T2q, T2y, T2v, T2B; |
251 | | { |
252 | | E T2o, T2p, T2t, T2u; |
253 | | T2o = T20 - T21; |
254 | | T2p = T2c + T2d; |
255 | | T2q = T2o - T2p; |
256 | | T2y = T2o + T2p; |
257 | | T2t = T29 + T2a; |
258 | | T2u = T23 - T24; |
259 | | T2v = T2t + T2u; |
260 | | T2B = T2t - T2u; |
261 | | } |
262 | | { |
263 | | E T2r, T2w, T2n, T2s; |
264 | | T2n = W[8]; |
265 | | T2r = T2n * T2q; |
266 | | T2w = T2n * T2v; |
267 | | T2s = W[9]; |
268 | | cr[WS(rs, 5)] = FNMS(T2s, T2v, T2r); |
269 | | ci[WS(rs, 5)] = FMA(T2s, T2q, T2w); |
270 | | } |
271 | | { |
272 | | E T2z, T2C, T2x, T2A; |
273 | | T2x = W[20]; |
274 | | T2z = T2x * T2y; |
275 | | T2C = T2x * T2B; |
276 | | T2A = W[21]; |
277 | | cr[WS(rs, 11)] = FNMS(T2A, T2B, T2z); |
278 | | ci[WS(rs, 11)] = FMA(T2A, T2y, T2C); |
279 | | } |
280 | | } |
281 | | { |
282 | | E T1M, T1U, T1R, T1X; |
283 | | { |
284 | | E T1K, T1L, T1P, T1Q; |
285 | | T1K = T18 - T1b; |
286 | | T1L = T1w + T1z; |
287 | | T1M = T1K - T1L; |
288 | | T1U = T1K + T1L; |
289 | | T1P = T1p + T1s; |
290 | | T1Q = T1f - T1i; |
291 | | T1R = T1P + T1Q; |
292 | | T1X = T1P - T1Q; |
293 | | } |
294 | | { |
295 | | E T1N, T1S, T1J, T1O; |
296 | | T1J = W[0]; |
297 | | T1N = T1J * T1M; |
298 | | T1S = T1J * T1R; |
299 | | T1O = W[1]; |
300 | | cr[WS(rs, 1)] = FNMS(T1O, T1R, T1N); |
301 | | ci[WS(rs, 1)] = FMA(T1O, T1M, T1S); |
302 | | } |
303 | | { |
304 | | E T1V, T1Y, T1T, T1W; |
305 | | T1T = W[12]; |
306 | | T1V = T1T * T1U; |
307 | | T1Y = T1T * T1X; |
308 | | T1W = W[13]; |
309 | | cr[WS(rs, 7)] = FNMS(T1W, T1X, T1V); |
310 | | ci[WS(rs, 7)] = FMA(T1W, T1U, T1Y); |
311 | | } |
312 | | } |
313 | | } |
314 | | } |
315 | | } |
316 | | |
317 | | static const tw_instr twinstr[] = { |
318 | | { TW_FULL, 1, 12 }, |
319 | | { TW_NEXT, 1, 0 } |
320 | | }; |
321 | | |
322 | | static const hc2hc_desc desc = { 12, "hb_12", twinstr, &GENUS, { 72, 22, 46, 0 } }; |
323 | | |
324 | | void X(codelet_hb_12) (planner *p) { |
325 | | X(khc2hc_register) (p, hb_12, &desc); |
326 | | } |
327 | | #else |
328 | | |
329 | | /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 12 -dif -name hb_12 -include rdft/scalar/hb.h */ |
330 | | |
331 | | /* |
332 | | * This function contains 118 FP additions, 60 FP multiplications, |
333 | | * (or, 88 additions, 30 multiplications, 30 fused multiply/add), |
334 | | * 39 stack variables, 2 constants, and 48 memory accesses |
335 | | */ |
336 | | #include "rdft/scalar/hb.h" |
337 | | |
338 | | static void hb_12(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms) |
339 | 0 | { |
340 | 0 | DK(KP500000000, +0.500000000000000000000000000000000000000000000); |
341 | 0 | DK(KP866025403, +0.866025403784438646763723170752936183471402627); |
342 | 0 | { |
343 | 0 | INT m; |
344 | 0 | for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) { |
345 | 0 | E T5, TH, T12, T1M, T1i, T1U, Tg, Tt, T19, T1X, T1p, T1P, Ta, TM, T15; |
346 | 0 | E T1N, T1l, T1V, Tl, Ty, T1c, T1Y, T1s, T1Q; |
347 | 0 | { |
348 | 0 | E T1, TD, T4, T1g, TG, T11, T10, T1h; |
349 | 0 | T1 = cr[0]; |
350 | 0 | TD = ci[WS(rs, 11)]; |
351 | 0 | { |
352 | 0 | E T2, T3, TE, TF; |
353 | 0 | T2 = cr[WS(rs, 4)]; |
354 | 0 | T3 = ci[WS(rs, 3)]; |
355 | 0 | T4 = T2 + T3; |
356 | 0 | T1g = KP866025403 * (T2 - T3); |
357 | 0 | TE = ci[WS(rs, 7)]; |
358 | 0 | TF = cr[WS(rs, 8)]; |
359 | 0 | TG = TE - TF; |
360 | 0 | T11 = KP866025403 * (TE + TF); |
361 | 0 | } |
362 | 0 | T5 = T1 + T4; |
363 | 0 | TH = TD + TG; |
364 | 0 | T10 = FNMS(KP500000000, T4, T1); |
365 | 0 | T12 = T10 - T11; |
366 | 0 | T1M = T10 + T11; |
367 | 0 | T1h = FNMS(KP500000000, TG, TD); |
368 | 0 | T1i = T1g + T1h; |
369 | 0 | T1U = T1h - T1g; |
370 | 0 | } |
371 | 0 | { |
372 | 0 | E Tc, Tp, Tf, T17, Ts, T1o, T18, T1n; |
373 | 0 | Tc = cr[WS(rs, 3)]; |
374 | 0 | Tp = ci[WS(rs, 8)]; |
375 | 0 | { |
376 | 0 | E Td, Te, Tq, Tr; |
377 | 0 | Td = ci[WS(rs, 4)]; |
378 | 0 | Te = ci[0]; |
379 | 0 | Tf = Td + Te; |
380 | 0 | T17 = KP866025403 * (Td - Te); |
381 | 0 | Tq = cr[WS(rs, 7)]; |
382 | 0 | Tr = cr[WS(rs, 11)]; |
383 | 0 | Ts = Tq + Tr; |
384 | 0 | T1o = KP866025403 * (Tq - Tr); |
385 | 0 | } |
386 | 0 | Tg = Tc + Tf; |
387 | 0 | Tt = Tp - Ts; |
388 | 0 | T18 = FMA(KP500000000, Ts, Tp); |
389 | 0 | T19 = T17 + T18; |
390 | 0 | T1X = T18 - T17; |
391 | 0 | T1n = FNMS(KP500000000, Tf, Tc); |
392 | 0 | T1p = T1n + T1o; |
393 | 0 | T1P = T1n - T1o; |
394 | 0 | } |
395 | 0 | { |
396 | 0 | E T6, TL, T9, T1j, TK, T14, T13, T1k; |
397 | 0 | T6 = ci[WS(rs, 5)]; |
398 | 0 | TL = cr[WS(rs, 6)]; |
399 | 0 | { |
400 | 0 | E T7, T8, TI, TJ; |
401 | 0 | T7 = ci[WS(rs, 1)]; |
402 | 0 | T8 = cr[WS(rs, 2)]; |
403 | 0 | T9 = T7 + T8; |
404 | 0 | T1j = KP866025403 * (T7 - T8); |
405 | 0 | TI = ci[WS(rs, 9)]; |
406 | 0 | TJ = cr[WS(rs, 10)]; |
407 | 0 | TK = TI - TJ; |
408 | 0 | T14 = KP866025403 * (TI + TJ); |
409 | 0 | } |
410 | 0 | Ta = T6 + T9; |
411 | 0 | TM = TK - TL; |
412 | 0 | T13 = FNMS(KP500000000, T9, T6); |
413 | 0 | T15 = T13 + T14; |
414 | 0 | T1N = T13 - T14; |
415 | 0 | T1k = FMA(KP500000000, TK, TL); |
416 | 0 | T1l = T1j - T1k; |
417 | 0 | T1V = T1j + T1k; |
418 | 0 | } |
419 | 0 | { |
420 | 0 | E Th, Tx, Tk, T1a, Tw, T1r, T1b, T1q; |
421 | 0 | Th = ci[WS(rs, 2)]; |
422 | 0 | Tx = cr[WS(rs, 9)]; |
423 | 0 | { |
424 | 0 | E Ti, Tj, Tu, Tv; |
425 | 0 | Ti = cr[WS(rs, 1)]; |
426 | 0 | Tj = cr[WS(rs, 5)]; |
427 | 0 | Tk = Ti + Tj; |
428 | 0 | T1a = KP866025403 * (Ti - Tj); |
429 | 0 | Tu = ci[WS(rs, 10)]; |
430 | 0 | Tv = ci[WS(rs, 6)]; |
431 | 0 | Tw = Tu + Tv; |
432 | 0 | T1r = KP866025403 * (Tv - Tu); |
433 | 0 | } |
434 | 0 | Tl = Th + Tk; |
435 | 0 | Ty = Tw - Tx; |
436 | 0 | T1b = FMA(KP500000000, Tw, Tx); |
437 | 0 | T1c = T1a - T1b; |
438 | 0 | T1Y = T1a + T1b; |
439 | 0 | T1q = FNMS(KP500000000, Tk, Th); |
440 | 0 | T1s = T1q + T1r; |
441 | 0 | T1Q = T1q - T1r; |
442 | 0 | } |
443 | 0 | { |
444 | 0 | E Tb, Tm, TU, TW, TX, TY, TT, TV; |
445 | 0 | Tb = T5 + Ta; |
446 | 0 | Tm = Tg + Tl; |
447 | 0 | TU = Tb - Tm; |
448 | 0 | TW = TH + TM; |
449 | 0 | TX = Tt + Ty; |
450 | 0 | TY = TW - TX; |
451 | 0 | cr[0] = Tb + Tm; |
452 | 0 | ci[0] = TW + TX; |
453 | 0 | TT = W[10]; |
454 | 0 | TV = W[11]; |
455 | 0 | cr[WS(rs, 6)] = FNMS(TV, TY, TT * TU); |
456 | 0 | ci[WS(rs, 6)] = FMA(TV, TU, TT * TY); |
457 | 0 | } |
458 | 0 | { |
459 | 0 | E TA, TQ, TO, TS; |
460 | 0 | { |
461 | 0 | E To, Tz, TC, TN; |
462 | 0 | To = T5 - Ta; |
463 | 0 | Tz = Tt - Ty; |
464 | 0 | TA = To - Tz; |
465 | 0 | TQ = To + Tz; |
466 | 0 | TC = Tg - Tl; |
467 | 0 | TN = TH - TM; |
468 | 0 | TO = TC + TN; |
469 | 0 | TS = TN - TC; |
470 | 0 | } |
471 | 0 | { |
472 | 0 | E Tn, TB, TP, TR; |
473 | 0 | Tn = W[16]; |
474 | 0 | TB = W[17]; |
475 | 0 | cr[WS(rs, 9)] = FNMS(TB, TO, Tn * TA); |
476 | 0 | ci[WS(rs, 9)] = FMA(Tn, TO, TB * TA); |
477 | 0 | TP = W[4]; |
478 | 0 | TR = W[5]; |
479 | 0 | cr[WS(rs, 3)] = FNMS(TR, TS, TP * TQ); |
480 | 0 | ci[WS(rs, 3)] = FMA(TP, TS, TR * TQ); |
481 | 0 | } |
482 | 0 | } |
483 | 0 | { |
484 | 0 | E T28, T2e, T2c, T2g; |
485 | 0 | { |
486 | 0 | E T26, T27, T2a, T2b; |
487 | 0 | T26 = T1M - T1N; |
488 | 0 | T27 = T1X + T1Y; |
489 | 0 | T28 = T26 - T27; |
490 | 0 | T2e = T26 + T27; |
491 | 0 | T2a = T1U + T1V; |
492 | 0 | T2b = T1P - T1Q; |
493 | 0 | T2c = T2a + T2b; |
494 | 0 | T2g = T2a - T2b; |
495 | 0 | } |
496 | 0 | { |
497 | 0 | E T25, T29, T2d, T2f; |
498 | 0 | T25 = W[8]; |
499 | 0 | T29 = W[9]; |
500 | 0 | cr[WS(rs, 5)] = FNMS(T29, T2c, T25 * T28); |
501 | 0 | ci[WS(rs, 5)] = FMA(T25, T2c, T29 * T28); |
502 | 0 | T2d = W[20]; |
503 | 0 | T2f = W[21]; |
504 | 0 | cr[WS(rs, 11)] = FNMS(T2f, T2g, T2d * T2e); |
505 | 0 | ci[WS(rs, 11)] = FMA(T2d, T2g, T2f * T2e); |
506 | 0 | } |
507 | 0 | } |
508 | 0 | { |
509 | 0 | E T1S, T22, T20, T24; |
510 | 0 | { |
511 | 0 | E T1O, T1R, T1W, T1Z; |
512 | 0 | T1O = T1M + T1N; |
513 | 0 | T1R = T1P + T1Q; |
514 | 0 | T1S = T1O - T1R; |
515 | 0 | T22 = T1O + T1R; |
516 | 0 | T1W = T1U - T1V; |
517 | 0 | T1Z = T1X - T1Y; |
518 | 0 | T20 = T1W - T1Z; |
519 | 0 | T24 = T1W + T1Z; |
520 | 0 | } |
521 | 0 | { |
522 | 0 | E T1L, T1T, T21, T23; |
523 | 0 | T1L = W[2]; |
524 | 0 | T1T = W[3]; |
525 | 0 | cr[WS(rs, 2)] = FNMS(T1T, T20, T1L * T1S); |
526 | 0 | ci[WS(rs, 2)] = FMA(T1T, T1S, T1L * T20); |
527 | 0 | T21 = W[14]; |
528 | 0 | T23 = W[15]; |
529 | 0 | cr[WS(rs, 8)] = FNMS(T23, T24, T21 * T22); |
530 | 0 | ci[WS(rs, 8)] = FMA(T23, T22, T21 * T24); |
531 | 0 | } |
532 | 0 | } |
533 | 0 | { |
534 | 0 | E T1C, T1I, T1G, T1K; |
535 | 0 | { |
536 | 0 | E T1A, T1B, T1E, T1F; |
537 | 0 | T1A = T12 + T15; |
538 | 0 | T1B = T1p + T1s; |
539 | 0 | T1C = T1A - T1B; |
540 | 0 | T1I = T1A + T1B; |
541 | 0 | T1E = T1i + T1l; |
542 | 0 | T1F = T19 + T1c; |
543 | 0 | T1G = T1E - T1F; |
544 | 0 | T1K = T1E + T1F; |
545 | 0 | } |
546 | 0 | { |
547 | 0 | E T1z, T1D, T1H, T1J; |
548 | 0 | T1z = W[18]; |
549 | 0 | T1D = W[19]; |
550 | 0 | cr[WS(rs, 10)] = FNMS(T1D, T1G, T1z * T1C); |
551 | 0 | ci[WS(rs, 10)] = FMA(T1D, T1C, T1z * T1G); |
552 | 0 | T1H = W[6]; |
553 | 0 | T1J = W[7]; |
554 | 0 | cr[WS(rs, 4)] = FNMS(T1J, T1K, T1H * T1I); |
555 | 0 | ci[WS(rs, 4)] = FMA(T1J, T1I, T1H * T1K); |
556 | 0 | } |
557 | 0 | } |
558 | 0 | { |
559 | 0 | E T1e, T1w, T1u, T1y; |
560 | 0 | { |
561 | 0 | E T16, T1d, T1m, T1t; |
562 | 0 | T16 = T12 - T15; |
563 | 0 | T1d = T19 - T1c; |
564 | 0 | T1e = T16 - T1d; |
565 | 0 | T1w = T16 + T1d; |
566 | 0 | T1m = T1i - T1l; |
567 | 0 | T1t = T1p - T1s; |
568 | 0 | T1u = T1m + T1t; |
569 | 0 | T1y = T1m - T1t; |
570 | 0 | } |
571 | 0 | { |
572 | 0 | E TZ, T1f, T1v, T1x; |
573 | 0 | TZ = W[0]; |
574 | 0 | T1f = W[1]; |
575 | 0 | cr[WS(rs, 1)] = FNMS(T1f, T1u, TZ * T1e); |
576 | 0 | ci[WS(rs, 1)] = FMA(TZ, T1u, T1f * T1e); |
577 | 0 | T1v = W[12]; |
578 | 0 | T1x = W[13]; |
579 | 0 | cr[WS(rs, 7)] = FNMS(T1x, T1y, T1v * T1w); |
580 | 0 | ci[WS(rs, 7)] = FMA(T1v, T1y, T1x * T1w); |
581 | 0 | } |
582 | 0 | } |
583 | 0 | } |
584 | 0 | } |
585 | 0 | } |
586 | | |
587 | | static const tw_instr twinstr[] = { |
588 | | { TW_FULL, 1, 12 }, |
589 | | { TW_NEXT, 1, 0 } |
590 | | }; |
591 | | |
592 | | static const hc2hc_desc desc = { 12, "hb_12", twinstr, &GENUS, { 88, 30, 30, 0 } }; |
593 | | |
594 | 1 | void X(codelet_hb_12) (planner *p) { |
595 | 1 | X(khc2hc_register) (p, hb_12, &desc); |
596 | 1 | } |
597 | | #endif |