Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/q1bv_8.c @ 167:bd3cc4d1df30

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam <cannam@all-day-breakfast.com>
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
166:cbd6d7e562c7 167:bd3cc4d1df30
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 Thu May 24 08:06:15 EDT 2018 */
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_twidsq_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 8 -dif -name q1bv_8 -include dft/simd/q1b.h -sign 1 */
29
30 /*
31 * This function contains 264 FP additions, 192 FP multiplications,
32 * (or, 184 additions, 112 multiplications, 80 fused multiply/add),
33 * 77 stack variables, 1 constants, and 128 memory accesses
34 */
35 #include "dft/simd/q1b.h"
36
37 static void q1bv_8(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT m;
42 R *x;
43 x = ii;
44 for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(16, rs), MAKE_VOLATILE_STRIDE(16, vs)) {
45 V T3, Tv, Te, Tp, T1E, T26, T1P, T20, T2b, T2D, T2m, T2x, T3M, T4e, T3X;
46 V T48, TA, T12, TL, TW, T17, T1z, T1i, T1t, T2I, T3a, T2T, T34, T3f, T3H;
47 V T3q, T3B, Ta, Tw, Tf, Ts, T1L, T27, T1Q, T23, T2i, T2E, T2n, T2A, T3T;
48 V T4f, T3Y, T4b, TH, T13, TM, TZ, T1e, T1A, T1j, T1w, T2P, T3b, T2U, T37;
49 V T3m, T3I, T3r, T3E, T28, T14;
50 {
51 V T1, T2, Tn, Tc, Td, To;
52 T1 = LD(&(x[0]), ms, &(x[0]));
53 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
54 Tn = VADD(T1, T2);
55 Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
56 Td = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
57 To = VADD(Tc, Td);
58 T3 = VSUB(T1, T2);
59 Tv = VADD(Tn, To);
60 Te = VSUB(Tc, Td);
61 Tp = VSUB(Tn, To);
62 }
63 {
64 V T1C, T1D, T1Y, T1N, T1O, T1Z;
65 T1C = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
66 T1D = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
67 T1Y = VADD(T1C, T1D);
68 T1N = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
69 T1O = LD(&(x[WS(vs, 3) + WS(rs, 6)]), ms, &(x[WS(vs, 3)]));
70 T1Z = VADD(T1N, T1O);
71 T1E = VSUB(T1C, T1D);
72 T26 = VADD(T1Y, T1Z);
73 T1P = VSUB(T1N, T1O);
74 T20 = VSUB(T1Y, T1Z);
75 }
76 {
77 V T29, T2a, T2v, T2k, T2l, T2w;
78 T29 = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
79 T2a = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
80 T2v = VADD(T29, T2a);
81 T2k = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
82 T2l = LD(&(x[WS(vs, 4) + WS(rs, 6)]), ms, &(x[WS(vs, 4)]));
83 T2w = VADD(T2k, T2l);
84 T2b = VSUB(T29, T2a);
85 T2D = VADD(T2v, T2w);
86 T2m = VSUB(T2k, T2l);
87 T2x = VSUB(T2v, T2w);
88 }
89 {
90 V T3K, T3L, T46, T3V, T3W, T47;
91 T3K = LD(&(x[WS(vs, 7)]), ms, &(x[WS(vs, 7)]));
92 T3L = LD(&(x[WS(vs, 7) + WS(rs, 4)]), ms, &(x[WS(vs, 7)]));
93 T46 = VADD(T3K, T3L);
94 T3V = LD(&(x[WS(vs, 7) + WS(rs, 2)]), ms, &(x[WS(vs, 7)]));
95 T3W = LD(&(x[WS(vs, 7) + WS(rs, 6)]), ms, &(x[WS(vs, 7)]));
96 T47 = VADD(T3V, T3W);
97 T3M = VSUB(T3K, T3L);
98 T4e = VADD(T46, T47);
99 T3X = VSUB(T3V, T3W);
100 T48 = VSUB(T46, T47);
101 }
102 {
103 V Ty, Tz, TU, TJ, TK, TV;
104 Ty = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
105 Tz = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
106 TU = VADD(Ty, Tz);
107 TJ = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
108 TK = LD(&(x[WS(vs, 1) + WS(rs, 6)]), ms, &(x[WS(vs, 1)]));
109 TV = VADD(TJ, TK);
110 TA = VSUB(Ty, Tz);
111 T12 = VADD(TU, TV);
112 TL = VSUB(TJ, TK);
113 TW = VSUB(TU, TV);
114 }
115 {
116 V T15, T16, T1r, T1g, T1h, T1s;
117 T15 = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
118 T16 = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
119 T1r = VADD(T15, T16);
120 T1g = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
121 T1h = LD(&(x[WS(vs, 2) + WS(rs, 6)]), ms, &(x[WS(vs, 2)]));
122 T1s = VADD(T1g, T1h);
123 T17 = VSUB(T15, T16);
124 T1z = VADD(T1r, T1s);
125 T1i = VSUB(T1g, T1h);
126 T1t = VSUB(T1r, T1s);
127 }
128 {
129 V T2G, T2H, T32, T2R, T2S, T33;
130 T2G = LD(&(x[WS(vs, 5)]), ms, &(x[WS(vs, 5)]));
131 T2H = LD(&(x[WS(vs, 5) + WS(rs, 4)]), ms, &(x[WS(vs, 5)]));
132 T32 = VADD(T2G, T2H);
133 T2R = LD(&(x[WS(vs, 5) + WS(rs, 2)]), ms, &(x[WS(vs, 5)]));
134 T2S = LD(&(x[WS(vs, 5) + WS(rs, 6)]), ms, &(x[WS(vs, 5)]));
135 T33 = VADD(T2R, T2S);
136 T2I = VSUB(T2G, T2H);
137 T3a = VADD(T32, T33);
138 T2T = VSUB(T2R, T2S);
139 T34 = VSUB(T32, T33);
140 }
141 {
142 V T3d, T3e, T3z, T3o, T3p, T3A;
143 T3d = LD(&(x[WS(vs, 6)]), ms, &(x[WS(vs, 6)]));
144 T3e = LD(&(x[WS(vs, 6) + WS(rs, 4)]), ms, &(x[WS(vs, 6)]));
145 T3z = VADD(T3d, T3e);
146 T3o = LD(&(x[WS(vs, 6) + WS(rs, 2)]), ms, &(x[WS(vs, 6)]));
147 T3p = LD(&(x[WS(vs, 6) + WS(rs, 6)]), ms, &(x[WS(vs, 6)]));
148 T3A = VADD(T3o, T3p);
149 T3f = VSUB(T3d, T3e);
150 T3H = VADD(T3z, T3A);
151 T3q = VSUB(T3o, T3p);
152 T3B = VSUB(T3z, T3A);
153 }
154 {
155 V T6, Tq, T9, Tr;
156 {
157 V T4, T5, T7, T8;
158 T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
159 T5 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
160 T6 = VSUB(T4, T5);
161 Tq = VADD(T4, T5);
162 T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
163 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
164 T9 = VSUB(T7, T8);
165 Tr = VADD(T7, T8);
166 }
167 Ta = VADD(T6, T9);
168 Tw = VADD(Tq, Tr);
169 Tf = VSUB(T6, T9);
170 Ts = VSUB(Tq, Tr);
171 }
172 {
173 V T1H, T21, T1K, T22;
174 {
175 V T1F, T1G, T1I, T1J;
176 T1F = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
177 T1G = LD(&(x[WS(vs, 3) + WS(rs, 5)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
178 T1H = VSUB(T1F, T1G);
179 T21 = VADD(T1F, T1G);
180 T1I = LD(&(x[WS(vs, 3) + WS(rs, 7)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
181 T1J = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
182 T1K = VSUB(T1I, T1J);
183 T22 = VADD(T1I, T1J);
184 }
185 T1L = VADD(T1H, T1K);
186 T27 = VADD(T21, T22);
187 T1Q = VSUB(T1H, T1K);
188 T23 = VSUB(T21, T22);
189 }
190 {
191 V T2e, T2y, T2h, T2z;
192 {
193 V T2c, T2d, T2f, T2g;
194 T2c = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
195 T2d = LD(&(x[WS(vs, 4) + WS(rs, 5)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
196 T2e = VSUB(T2c, T2d);
197 T2y = VADD(T2c, T2d);
198 T2f = LD(&(x[WS(vs, 4) + WS(rs, 7)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
199 T2g = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
200 T2h = VSUB(T2f, T2g);
201 T2z = VADD(T2f, T2g);
202 }
203 T2i = VADD(T2e, T2h);
204 T2E = VADD(T2y, T2z);
205 T2n = VSUB(T2e, T2h);
206 T2A = VSUB(T2y, T2z);
207 }
208 {
209 V T3P, T49, T3S, T4a;
210 {
211 V T3N, T3O, T3Q, T3R;
212 T3N = LD(&(x[WS(vs, 7) + WS(rs, 1)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
213 T3O = LD(&(x[WS(vs, 7) + WS(rs, 5)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
214 T3P = VSUB(T3N, T3O);
215 T49 = VADD(T3N, T3O);
216 T3Q = LD(&(x[WS(vs, 7) + WS(rs, 7)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
217 T3R = LD(&(x[WS(vs, 7) + WS(rs, 3)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
218 T3S = VSUB(T3Q, T3R);
219 T4a = VADD(T3Q, T3R);
220 }
221 T3T = VADD(T3P, T3S);
222 T4f = VADD(T49, T4a);
223 T3Y = VSUB(T3P, T3S);
224 T4b = VSUB(T49, T4a);
225 }
226 {
227 V TD, TX, TG, TY;
228 {
229 V TB, TC, TE, TF;
230 TB = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
231 TC = LD(&(x[WS(vs, 1) + WS(rs, 5)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
232 TD = VSUB(TB, TC);
233 TX = VADD(TB, TC);
234 TE = LD(&(x[WS(vs, 1) + WS(rs, 7)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
235 TF = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
236 TG = VSUB(TE, TF);
237 TY = VADD(TE, TF);
238 }
239 TH = VADD(TD, TG);
240 T13 = VADD(TX, TY);
241 TM = VSUB(TD, TG);
242 TZ = VSUB(TX, TY);
243 }
244 {
245 V T1a, T1u, T1d, T1v;
246 {
247 V T18, T19, T1b, T1c;
248 T18 = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
249 T19 = LD(&(x[WS(vs, 2) + WS(rs, 5)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
250 T1a = VSUB(T18, T19);
251 T1u = VADD(T18, T19);
252 T1b = LD(&(x[WS(vs, 2) + WS(rs, 7)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
253 T1c = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
254 T1d = VSUB(T1b, T1c);
255 T1v = VADD(T1b, T1c);
256 }
257 T1e = VADD(T1a, T1d);
258 T1A = VADD(T1u, T1v);
259 T1j = VSUB(T1a, T1d);
260 T1w = VSUB(T1u, T1v);
261 }
262 {
263 V T2L, T35, T2O, T36;
264 {
265 V T2J, T2K, T2M, T2N;
266 T2J = LD(&(x[WS(vs, 5) + WS(rs, 1)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
267 T2K = LD(&(x[WS(vs, 5) + WS(rs, 5)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
268 T2L = VSUB(T2J, T2K);
269 T35 = VADD(T2J, T2K);
270 T2M = LD(&(x[WS(vs, 5) + WS(rs, 7)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
271 T2N = LD(&(x[WS(vs, 5) + WS(rs, 3)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
272 T2O = VSUB(T2M, T2N);
273 T36 = VADD(T2M, T2N);
274 }
275 T2P = VADD(T2L, T2O);
276 T3b = VADD(T35, T36);
277 T2U = VSUB(T2L, T2O);
278 T37 = VSUB(T35, T36);
279 }
280 {
281 V T3i, T3C, T3l, T3D;
282 {
283 V T3g, T3h, T3j, T3k;
284 T3g = LD(&(x[WS(vs, 6) + WS(rs, 1)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
285 T3h = LD(&(x[WS(vs, 6) + WS(rs, 5)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
286 T3i = VSUB(T3g, T3h);
287 T3C = VADD(T3g, T3h);
288 T3j = LD(&(x[WS(vs, 6) + WS(rs, 7)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
289 T3k = LD(&(x[WS(vs, 6) + WS(rs, 3)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
290 T3l = VSUB(T3j, T3k);
291 T3D = VADD(T3j, T3k);
292 }
293 T3m = VADD(T3i, T3l);
294 T3I = VADD(T3C, T3D);
295 T3r = VSUB(T3i, T3l);
296 T3E = VSUB(T3C, T3D);
297 }
298 ST(&(x[0]), VADD(Tv, Tw), ms, &(x[0]));
299 ST(&(x[WS(rs, 2)]), VADD(T1z, T1A), ms, &(x[0]));
300 ST(&(x[WS(rs, 5)]), VADD(T3a, T3b), ms, &(x[WS(rs, 1)]));
301 ST(&(x[WS(rs, 7)]), VADD(T4e, T4f), ms, &(x[WS(rs, 1)]));
302 ST(&(x[WS(rs, 6)]), VADD(T3H, T3I), ms, &(x[0]));
303 ST(&(x[WS(rs, 4)]), VADD(T2D, T2E), ms, &(x[0]));
304 {
305 V Tt, T4c, T2B, T24;
306 ST(&(x[WS(rs, 3)]), VADD(T26, T27), ms, &(x[WS(rs, 1)]));
307 ST(&(x[WS(rs, 1)]), VADD(T12, T13), ms, &(x[WS(rs, 1)]));
308 Tt = BYTW(&(W[TWVL * 10]), VFNMSI(Ts, Tp));
309 ST(&(x[WS(vs, 6)]), Tt, ms, &(x[WS(vs, 6)]));
310 T4c = BYTW(&(W[TWVL * 10]), VFNMSI(T4b, T48));
311 ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
312 T2B = BYTW(&(W[TWVL * 10]), VFNMSI(T2A, T2x));
313 ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 6)]));
314 T24 = BYTW(&(W[TWVL * 10]), VFNMSI(T23, T20));
315 ST(&(x[WS(vs, 6) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
316 }
317 {
318 V T10, T1x, T3F, T38, T1y, Tu;
319 T10 = BYTW(&(W[TWVL * 10]), VFNMSI(TZ, TW));
320 ST(&(x[WS(vs, 6) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
321 T1x = BYTW(&(W[TWVL * 10]), VFNMSI(T1w, T1t));
322 ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 6)]));
323 T3F = BYTW(&(W[TWVL * 10]), VFNMSI(T3E, T3B));
324 ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 6)]));
325 T38 = BYTW(&(W[TWVL * 10]), VFNMSI(T37, T34));
326 ST(&(x[WS(vs, 6) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
327 T1y = BYTW(&(W[TWVL * 2]), VFMAI(T1w, T1t));
328 ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1y, ms, &(x[WS(vs, 2)]));
329 Tu = BYTW(&(W[TWVL * 2]), VFMAI(Ts, Tp));
330 ST(&(x[WS(vs, 2)]), Tu, ms, &(x[WS(vs, 2)]));
331 }
332 {
333 V T2C, T3G, T11, T25, T39, T4d;
334 T2C = BYTW(&(W[TWVL * 2]), VFMAI(T2A, T2x));
335 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T2C, ms, &(x[WS(vs, 2)]));
336 T3G = BYTW(&(W[TWVL * 2]), VFMAI(T3E, T3B));
337 ST(&(x[WS(vs, 2) + WS(rs, 6)]), T3G, ms, &(x[WS(vs, 2)]));
338 T11 = BYTW(&(W[TWVL * 2]), VFMAI(TZ, TW));
339 ST(&(x[WS(vs, 2) + WS(rs, 1)]), T11, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
340 T25 = BYTW(&(W[TWVL * 2]), VFMAI(T23, T20));
341 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T25, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
342 T39 = BYTW(&(W[TWVL * 2]), VFMAI(T37, T34));
343 ST(&(x[WS(vs, 2) + WS(rs, 5)]), T39, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
344 T4d = BYTW(&(W[TWVL * 2]), VFMAI(T4b, T48));
345 ST(&(x[WS(vs, 2) + WS(rs, 7)]), T4d, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
346 }
347 {
348 V Tx, T1B, T3c, T4g, T3J, T2F;
349 Tx = BYTW(&(W[TWVL * 6]), VSUB(Tv, Tw));
350 ST(&(x[WS(vs, 4)]), Tx, ms, &(x[WS(vs, 4)]));
351 T1B = BYTW(&(W[TWVL * 6]), VSUB(T1z, T1A));
352 ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 4)]));
353 T3c = BYTW(&(W[TWVL * 6]), VSUB(T3a, T3b));
354 ST(&(x[WS(vs, 4) + WS(rs, 5)]), T3c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
355 T4g = BYTW(&(W[TWVL * 6]), VSUB(T4e, T4f));
356 ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4g, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
357 T3J = BYTW(&(W[TWVL * 6]), VSUB(T3H, T3I));
358 ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3J, ms, &(x[WS(vs, 4)]));
359 T2F = BYTW(&(W[TWVL * 6]), VSUB(T2D, T2E));
360 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2F, ms, &(x[WS(vs, 4)]));
361 }
362 T28 = BYTW(&(W[TWVL * 6]), VSUB(T26, T27));
363 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T28, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
364 T14 = BYTW(&(W[TWVL * 6]), VSUB(T12, T13));
365 ST(&(x[WS(vs, 4) + WS(rs, 1)]), T14, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
366 {
367 V Th, Ti, Tb, Tg;
368 Tb = VFNMS(LDK(KP707106781), Ta, T3);
369 Tg = VFNMS(LDK(KP707106781), Tf, Te);
370 Th = BYTW(&(W[TWVL * 4]), VFNMSI(Tg, Tb));
371 Ti = BYTW(&(W[TWVL * 8]), VFMAI(Tg, Tb));
372 ST(&(x[WS(vs, 3)]), Th, ms, &(x[WS(vs, 3)]));
373 ST(&(x[WS(vs, 5)]), Ti, ms, &(x[WS(vs, 5)]));
374 }
375 {
376 V T40, T41, T3U, T3Z;
377 T3U = VFNMS(LDK(KP707106781), T3T, T3M);
378 T3Z = VFNMS(LDK(KP707106781), T3Y, T3X);
379 T40 = BYTW(&(W[TWVL * 4]), VFNMSI(T3Z, T3U));
380 T41 = BYTW(&(W[TWVL * 8]), VFMAI(T3Z, T3U));
381 ST(&(x[WS(vs, 3) + WS(rs, 7)]), T40, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
382 ST(&(x[WS(vs, 5) + WS(rs, 7)]), T41, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
383 }
384 {
385 V T2p, T2q, T2j, T2o;
386 T2j = VFNMS(LDK(KP707106781), T2i, T2b);
387 T2o = VFNMS(LDK(KP707106781), T2n, T2m);
388 T2p = BYTW(&(W[TWVL * 4]), VFNMSI(T2o, T2j));
389 T2q = BYTW(&(W[TWVL * 8]), VFMAI(T2o, T2j));
390 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T2p, ms, &(x[WS(vs, 3)]));
391 ST(&(x[WS(vs, 5) + WS(rs, 4)]), T2q, ms, &(x[WS(vs, 5)]));
392 }
393 {
394 V T1S, T1T, T1M, T1R;
395 T1M = VFNMS(LDK(KP707106781), T1L, T1E);
396 T1R = VFNMS(LDK(KP707106781), T1Q, T1P);
397 T1S = BYTW(&(W[TWVL * 4]), VFNMSI(T1R, T1M));
398 T1T = BYTW(&(W[TWVL * 8]), VFMAI(T1R, T1M));
399 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1S, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
400 ST(&(x[WS(vs, 5) + WS(rs, 3)]), T1T, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
401 }
402 {
403 V TO, TP, TI, TN;
404 TI = VFNMS(LDK(KP707106781), TH, TA);
405 TN = VFNMS(LDK(KP707106781), TM, TL);
406 TO = BYTW(&(W[TWVL * 4]), VFNMSI(TN, TI));
407 TP = BYTW(&(W[TWVL * 8]), VFMAI(TN, TI));
408 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TO, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
409 ST(&(x[WS(vs, 5) + WS(rs, 1)]), TP, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
410 }
411 {
412 V T1l, T1m, T1f, T1k;
413 T1f = VFNMS(LDK(KP707106781), T1e, T17);
414 T1k = VFNMS(LDK(KP707106781), T1j, T1i);
415 T1l = BYTW(&(W[TWVL * 4]), VFNMSI(T1k, T1f));
416 T1m = BYTW(&(W[TWVL * 8]), VFMAI(T1k, T1f));
417 ST(&(x[WS(vs, 3) + WS(rs, 2)]), T1l, ms, &(x[WS(vs, 3)]));
418 ST(&(x[WS(vs, 5) + WS(rs, 2)]), T1m, ms, &(x[WS(vs, 5)]));
419 }
420 {
421 V T3t, T3u, T3n, T3s;
422 T3n = VFNMS(LDK(KP707106781), T3m, T3f);
423 T3s = VFNMS(LDK(KP707106781), T3r, T3q);
424 T3t = BYTW(&(W[TWVL * 4]), VFNMSI(T3s, T3n));
425 T3u = BYTW(&(W[TWVL * 8]), VFMAI(T3s, T3n));
426 ST(&(x[WS(vs, 3) + WS(rs, 6)]), T3t, ms, &(x[WS(vs, 3)]));
427 ST(&(x[WS(vs, 5) + WS(rs, 6)]), T3u, ms, &(x[WS(vs, 5)]));
428 }
429 {
430 V T2W, T2X, T2Q, T2V;
431 T2Q = VFNMS(LDK(KP707106781), T2P, T2I);
432 T2V = VFNMS(LDK(KP707106781), T2U, T2T);
433 T2W = BYTW(&(W[TWVL * 4]), VFNMSI(T2V, T2Q));
434 T2X = BYTW(&(W[TWVL * 8]), VFMAI(T2V, T2Q));
435 ST(&(x[WS(vs, 3) + WS(rs, 5)]), T2W, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
436 ST(&(x[WS(vs, 5) + WS(rs, 5)]), T2X, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
437 }
438 {
439 V T1p, T1q, T1n, T1o;
440 T1n = VFMA(LDK(KP707106781), T1e, T17);
441 T1o = VFMA(LDK(KP707106781), T1j, T1i);
442 T1p = BYTW(&(W[0]), VFMAI(T1o, T1n));
443 T1q = BYTW(&(W[TWVL * 12]), VFNMSI(T1o, T1n));
444 ST(&(x[WS(vs, 1) + WS(rs, 2)]), T1p, ms, &(x[WS(vs, 1)]));
445 ST(&(x[WS(vs, 7) + WS(rs, 2)]), T1q, ms, &(x[WS(vs, 7)]));
446 }
447 {
448 V Tl, Tm, Tj, Tk;
449 Tj = VFMA(LDK(KP707106781), Ta, T3);
450 Tk = VFMA(LDK(KP707106781), Tf, Te);
451 Tl = BYTW(&(W[0]), VFMAI(Tk, Tj));
452 Tm = BYTW(&(W[TWVL * 12]), VFNMSI(Tk, Tj));
453 ST(&(x[WS(vs, 1)]), Tl, ms, &(x[WS(vs, 1)]));
454 ST(&(x[WS(vs, 7)]), Tm, ms, &(x[WS(vs, 7)]));
455 }
456 {
457 V T2t, T2u, T2r, T2s;
458 T2r = VFMA(LDK(KP707106781), T2i, T2b);
459 T2s = VFMA(LDK(KP707106781), T2n, T2m);
460 T2t = BYTW(&(W[0]), VFMAI(T2s, T2r));
461 T2u = BYTW(&(W[TWVL * 12]), VFNMSI(T2s, T2r));
462 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T2t, ms, &(x[WS(vs, 1)]));
463 ST(&(x[WS(vs, 7) + WS(rs, 4)]), T2u, ms, &(x[WS(vs, 7)]));
464 }
465 {
466 V T3x, T3y, T3v, T3w;
467 T3v = VFMA(LDK(KP707106781), T3m, T3f);
468 T3w = VFMA(LDK(KP707106781), T3r, T3q);
469 T3x = BYTW(&(W[0]), VFMAI(T3w, T3v));
470 T3y = BYTW(&(W[TWVL * 12]), VFNMSI(T3w, T3v));
471 ST(&(x[WS(vs, 1) + WS(rs, 6)]), T3x, ms, &(x[WS(vs, 1)]));
472 ST(&(x[WS(vs, 7) + WS(rs, 6)]), T3y, ms, &(x[WS(vs, 7)]));
473 }
474 {
475 V TS, TT, TQ, TR;
476 TQ = VFMA(LDK(KP707106781), TH, TA);
477 TR = VFMA(LDK(KP707106781), TM, TL);
478 TS = BYTW(&(W[0]), VFMAI(TR, TQ));
479 TT = BYTW(&(W[TWVL * 12]), VFNMSI(TR, TQ));
480 ST(&(x[WS(vs, 1) + WS(rs, 1)]), TS, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
481 ST(&(x[WS(vs, 7) + WS(rs, 1)]), TT, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
482 }
483 {
484 V T1W, T1X, T1U, T1V;
485 T1U = VFMA(LDK(KP707106781), T1L, T1E);
486 T1V = VFMA(LDK(KP707106781), T1Q, T1P);
487 T1W = BYTW(&(W[0]), VFMAI(T1V, T1U));
488 T1X = BYTW(&(W[TWVL * 12]), VFNMSI(T1V, T1U));
489 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1W, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
490 ST(&(x[WS(vs, 7) + WS(rs, 3)]), T1X, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
491 }
492 {
493 V T30, T31, T2Y, T2Z;
494 T2Y = VFMA(LDK(KP707106781), T2P, T2I);
495 T2Z = VFMA(LDK(KP707106781), T2U, T2T);
496 T30 = BYTW(&(W[0]), VFMAI(T2Z, T2Y));
497 T31 = BYTW(&(W[TWVL * 12]), VFNMSI(T2Z, T2Y));
498 ST(&(x[WS(vs, 1) + WS(rs, 5)]), T30, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
499 ST(&(x[WS(vs, 7) + WS(rs, 5)]), T31, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
500 }
501 {
502 V T44, T45, T42, T43;
503 T42 = VFMA(LDK(KP707106781), T3T, T3M);
504 T43 = VFMA(LDK(KP707106781), T3Y, T3X);
505 T44 = BYTW(&(W[0]), VFMAI(T43, T42));
506 T45 = BYTW(&(W[TWVL * 12]), VFNMSI(T43, T42));
507 ST(&(x[WS(vs, 1) + WS(rs, 7)]), T44, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
508 ST(&(x[WS(vs, 7) + WS(rs, 7)]), T45, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
509 }
510 }
511 }
512 VLEAVE();
513 }
514
515 static const tw_instr twinstr[] = {
516 VTW(0, 1),
517 VTW(0, 2),
518 VTW(0, 3),
519 VTW(0, 4),
520 VTW(0, 5),
521 VTW(0, 6),
522 VTW(0, 7),
523 {TW_NEXT, VL, 0}
524 };
525
526 static const ct_desc desc = { 8, XSIMD_STRING("q1bv_8"), twinstr, &GENUS, {184, 112, 80, 0}, 0, 0, 0 };
527
528 void XSIMD(codelet_q1bv_8) (planner *p) {
529 X(kdft_difsq_register) (p, q1bv_8, &desc);
530 }
531 #else
532
533 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -dif -name q1bv_8 -include dft/simd/q1b.h -sign 1 */
534
535 /*
536 * This function contains 264 FP additions, 128 FP multiplications,
537 * (or, 264 additions, 128 multiplications, 0 fused multiply/add),
538 * 77 stack variables, 1 constants, and 128 memory accesses
539 */
540 #include "dft/simd/q1b.h"
541
542 static void q1bv_8(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
543 {
544 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
545 {
546 INT m;
547 R *x;
548 x = ii;
549 for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(16, rs), MAKE_VOLATILE_STRIDE(16, vs)) {
550 V Ta, Tv, Te, Tp, T1L, T26, T1P, T20, T2i, T2D, T2m, T2x, T3T, T4e, T3X;
551 V T48, TH, T12, TL, TW, T1e, T1z, T1i, T1t, T2P, T3a, T2T, T34, T3m, T3H;
552 V T3q, T3B, T7, Tw, Tf, Ts, T1I, T27, T1Q, T23, T2f, T2E, T2n, T2A, T3Q;
553 V T4f, T3Y, T4b, TE, T13, TM, TZ, T1b, T1A, T1j, T1w, T2M, T3b, T2U, T37;
554 V T3j, T3I, T3r, T3E, T28, T14;
555 {
556 V T8, T9, To, Tc, Td, Tn;
557 T8 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
558 T9 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
559 To = VADD(T8, T9);
560 Tc = LD(&(x[0]), ms, &(x[0]));
561 Td = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
562 Tn = VADD(Tc, Td);
563 Ta = VSUB(T8, T9);
564 Tv = VADD(Tn, To);
565 Te = VSUB(Tc, Td);
566 Tp = VSUB(Tn, To);
567 }
568 {
569 V T1J, T1K, T1Z, T1N, T1O, T1Y;
570 T1J = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
571 T1K = LD(&(x[WS(vs, 3) + WS(rs, 6)]), ms, &(x[WS(vs, 3)]));
572 T1Z = VADD(T1J, T1K);
573 T1N = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
574 T1O = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
575 T1Y = VADD(T1N, T1O);
576 T1L = VSUB(T1J, T1K);
577 T26 = VADD(T1Y, T1Z);
578 T1P = VSUB(T1N, T1O);
579 T20 = VSUB(T1Y, T1Z);
580 }
581 {
582 V T2g, T2h, T2w, T2k, T2l, T2v;
583 T2g = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
584 T2h = LD(&(x[WS(vs, 4) + WS(rs, 6)]), ms, &(x[WS(vs, 4)]));
585 T2w = VADD(T2g, T2h);
586 T2k = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
587 T2l = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
588 T2v = VADD(T2k, T2l);
589 T2i = VSUB(T2g, T2h);
590 T2D = VADD(T2v, T2w);
591 T2m = VSUB(T2k, T2l);
592 T2x = VSUB(T2v, T2w);
593 }
594 {
595 V T3R, T3S, T47, T3V, T3W, T46;
596 T3R = LD(&(x[WS(vs, 7) + WS(rs, 2)]), ms, &(x[WS(vs, 7)]));
597 T3S = LD(&(x[WS(vs, 7) + WS(rs, 6)]), ms, &(x[WS(vs, 7)]));
598 T47 = VADD(T3R, T3S);
599 T3V = LD(&(x[WS(vs, 7)]), ms, &(x[WS(vs, 7)]));
600 T3W = LD(&(x[WS(vs, 7) + WS(rs, 4)]), ms, &(x[WS(vs, 7)]));
601 T46 = VADD(T3V, T3W);
602 T3T = VSUB(T3R, T3S);
603 T4e = VADD(T46, T47);
604 T3X = VSUB(T3V, T3W);
605 T48 = VSUB(T46, T47);
606 }
607 {
608 V TF, TG, TV, TJ, TK, TU;
609 TF = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
610 TG = LD(&(x[WS(vs, 1) + WS(rs, 6)]), ms, &(x[WS(vs, 1)]));
611 TV = VADD(TF, TG);
612 TJ = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
613 TK = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
614 TU = VADD(TJ, TK);
615 TH = VSUB(TF, TG);
616 T12 = VADD(TU, TV);
617 TL = VSUB(TJ, TK);
618 TW = VSUB(TU, TV);
619 }
620 {
621 V T1c, T1d, T1s, T1g, T1h, T1r;
622 T1c = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
623 T1d = LD(&(x[WS(vs, 2) + WS(rs, 6)]), ms, &(x[WS(vs, 2)]));
624 T1s = VADD(T1c, T1d);
625 T1g = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
626 T1h = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
627 T1r = VADD(T1g, T1h);
628 T1e = VSUB(T1c, T1d);
629 T1z = VADD(T1r, T1s);
630 T1i = VSUB(T1g, T1h);
631 T1t = VSUB(T1r, T1s);
632 }
633 {
634 V T2N, T2O, T33, T2R, T2S, T32;
635 T2N = LD(&(x[WS(vs, 5) + WS(rs, 2)]), ms, &(x[WS(vs, 5)]));
636 T2O = LD(&(x[WS(vs, 5) + WS(rs, 6)]), ms, &(x[WS(vs, 5)]));
637 T33 = VADD(T2N, T2O);
638 T2R = LD(&(x[WS(vs, 5)]), ms, &(x[WS(vs, 5)]));
639 T2S = LD(&(x[WS(vs, 5) + WS(rs, 4)]), ms, &(x[WS(vs, 5)]));
640 T32 = VADD(T2R, T2S);
641 T2P = VSUB(T2N, T2O);
642 T3a = VADD(T32, T33);
643 T2T = VSUB(T2R, T2S);
644 T34 = VSUB(T32, T33);
645 }
646 {
647 V T3k, T3l, T3A, T3o, T3p, T3z;
648 T3k = LD(&(x[WS(vs, 6) + WS(rs, 2)]), ms, &(x[WS(vs, 6)]));
649 T3l = LD(&(x[WS(vs, 6) + WS(rs, 6)]), ms, &(x[WS(vs, 6)]));
650 T3A = VADD(T3k, T3l);
651 T3o = LD(&(x[WS(vs, 6)]), ms, &(x[WS(vs, 6)]));
652 T3p = LD(&(x[WS(vs, 6) + WS(rs, 4)]), ms, &(x[WS(vs, 6)]));
653 T3z = VADD(T3o, T3p);
654 T3m = VSUB(T3k, T3l);
655 T3H = VADD(T3z, T3A);
656 T3q = VSUB(T3o, T3p);
657 T3B = VSUB(T3z, T3A);
658 }
659 {
660 V T3, Tq, T6, Tr;
661 {
662 V T1, T2, T4, T5;
663 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
664 T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
665 T3 = VSUB(T1, T2);
666 Tq = VADD(T1, T2);
667 T4 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
668 T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
669 T6 = VSUB(T4, T5);
670 Tr = VADD(T4, T5);
671 }
672 T7 = VMUL(LDK(KP707106781), VSUB(T3, T6));
673 Tw = VADD(Tq, Tr);
674 Tf = VMUL(LDK(KP707106781), VADD(T3, T6));
675 Ts = VBYI(VSUB(Tq, Tr));
676 }
677 {
678 V T1E, T21, T1H, T22;
679 {
680 V T1C, T1D, T1F, T1G;
681 T1C = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
682 T1D = LD(&(x[WS(vs, 3) + WS(rs, 5)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
683 T1E = VSUB(T1C, T1D);
684 T21 = VADD(T1C, T1D);
685 T1F = LD(&(x[WS(vs, 3) + WS(rs, 7)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
686 T1G = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
687 T1H = VSUB(T1F, T1G);
688 T22 = VADD(T1F, T1G);
689 }
690 T1I = VMUL(LDK(KP707106781), VSUB(T1E, T1H));
691 T27 = VADD(T21, T22);
692 T1Q = VMUL(LDK(KP707106781), VADD(T1E, T1H));
693 T23 = VBYI(VSUB(T21, T22));
694 }
695 {
696 V T2b, T2y, T2e, T2z;
697 {
698 V T29, T2a, T2c, T2d;
699 T29 = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
700 T2a = LD(&(x[WS(vs, 4) + WS(rs, 5)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
701 T2b = VSUB(T29, T2a);
702 T2y = VADD(T29, T2a);
703 T2c = LD(&(x[WS(vs, 4) + WS(rs, 7)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
704 T2d = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
705 T2e = VSUB(T2c, T2d);
706 T2z = VADD(T2c, T2d);
707 }
708 T2f = VMUL(LDK(KP707106781), VSUB(T2b, T2e));
709 T2E = VADD(T2y, T2z);
710 T2n = VMUL(LDK(KP707106781), VADD(T2b, T2e));
711 T2A = VBYI(VSUB(T2y, T2z));
712 }
713 {
714 V T3M, T49, T3P, T4a;
715 {
716 V T3K, T3L, T3N, T3O;
717 T3K = LD(&(x[WS(vs, 7) + WS(rs, 1)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
718 T3L = LD(&(x[WS(vs, 7) + WS(rs, 5)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
719 T3M = VSUB(T3K, T3L);
720 T49 = VADD(T3K, T3L);
721 T3N = LD(&(x[WS(vs, 7) + WS(rs, 7)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
722 T3O = LD(&(x[WS(vs, 7) + WS(rs, 3)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
723 T3P = VSUB(T3N, T3O);
724 T4a = VADD(T3N, T3O);
725 }
726 T3Q = VMUL(LDK(KP707106781), VSUB(T3M, T3P));
727 T4f = VADD(T49, T4a);
728 T3Y = VMUL(LDK(KP707106781), VADD(T3M, T3P));
729 T4b = VBYI(VSUB(T49, T4a));
730 }
731 {
732 V TA, TX, TD, TY;
733 {
734 V Ty, Tz, TB, TC;
735 Ty = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
736 Tz = LD(&(x[WS(vs, 1) + WS(rs, 5)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
737 TA = VSUB(Ty, Tz);
738 TX = VADD(Ty, Tz);
739 TB = LD(&(x[WS(vs, 1) + WS(rs, 7)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
740 TC = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
741 TD = VSUB(TB, TC);
742 TY = VADD(TB, TC);
743 }
744 TE = VMUL(LDK(KP707106781), VSUB(TA, TD));
745 T13 = VADD(TX, TY);
746 TM = VMUL(LDK(KP707106781), VADD(TA, TD));
747 TZ = VBYI(VSUB(TX, TY));
748 }
749 {
750 V T17, T1u, T1a, T1v;
751 {
752 V T15, T16, T18, T19;
753 T15 = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
754 T16 = LD(&(x[WS(vs, 2) + WS(rs, 5)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
755 T17 = VSUB(T15, T16);
756 T1u = VADD(T15, T16);
757 T18 = LD(&(x[WS(vs, 2) + WS(rs, 7)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
758 T19 = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
759 T1a = VSUB(T18, T19);
760 T1v = VADD(T18, T19);
761 }
762 T1b = VMUL(LDK(KP707106781), VSUB(T17, T1a));
763 T1A = VADD(T1u, T1v);
764 T1j = VMUL(LDK(KP707106781), VADD(T17, T1a));
765 T1w = VBYI(VSUB(T1u, T1v));
766 }
767 {
768 V T2I, T35, T2L, T36;
769 {
770 V T2G, T2H, T2J, T2K;
771 T2G = LD(&(x[WS(vs, 5) + WS(rs, 1)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
772 T2H = LD(&(x[WS(vs, 5) + WS(rs, 5)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
773 T2I = VSUB(T2G, T2H);
774 T35 = VADD(T2G, T2H);
775 T2J = LD(&(x[WS(vs, 5) + WS(rs, 7)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
776 T2K = LD(&(x[WS(vs, 5) + WS(rs, 3)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
777 T2L = VSUB(T2J, T2K);
778 T36 = VADD(T2J, T2K);
779 }
780 T2M = VMUL(LDK(KP707106781), VSUB(T2I, T2L));
781 T3b = VADD(T35, T36);
782 T2U = VMUL(LDK(KP707106781), VADD(T2I, T2L));
783 T37 = VBYI(VSUB(T35, T36));
784 }
785 {
786 V T3f, T3C, T3i, T3D;
787 {
788 V T3d, T3e, T3g, T3h;
789 T3d = LD(&(x[WS(vs, 6) + WS(rs, 1)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
790 T3e = LD(&(x[WS(vs, 6) + WS(rs, 5)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
791 T3f = VSUB(T3d, T3e);
792 T3C = VADD(T3d, T3e);
793 T3g = LD(&(x[WS(vs, 6) + WS(rs, 7)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
794 T3h = LD(&(x[WS(vs, 6) + WS(rs, 3)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
795 T3i = VSUB(T3g, T3h);
796 T3D = VADD(T3g, T3h);
797 }
798 T3j = VMUL(LDK(KP707106781), VSUB(T3f, T3i));
799 T3I = VADD(T3C, T3D);
800 T3r = VMUL(LDK(KP707106781), VADD(T3f, T3i));
801 T3E = VBYI(VSUB(T3C, T3D));
802 }
803 ST(&(x[0]), VADD(Tv, Tw), ms, &(x[0]));
804 ST(&(x[WS(rs, 2)]), VADD(T1z, T1A), ms, &(x[0]));
805 ST(&(x[WS(rs, 5)]), VADD(T3a, T3b), ms, &(x[WS(rs, 1)]));
806 ST(&(x[WS(rs, 7)]), VADD(T4e, T4f), ms, &(x[WS(rs, 1)]));
807 ST(&(x[WS(rs, 6)]), VADD(T3H, T3I), ms, &(x[0]));
808 ST(&(x[WS(rs, 4)]), VADD(T2D, T2E), ms, &(x[0]));
809 {
810 V Tt, T4c, T2B, T24;
811 ST(&(x[WS(rs, 3)]), VADD(T26, T27), ms, &(x[WS(rs, 1)]));
812 ST(&(x[WS(rs, 1)]), VADD(T12, T13), ms, &(x[WS(rs, 1)]));
813 Tt = BYTW(&(W[TWVL * 10]), VSUB(Tp, Ts));
814 ST(&(x[WS(vs, 6)]), Tt, ms, &(x[WS(vs, 6)]));
815 T4c = BYTW(&(W[TWVL * 10]), VSUB(T48, T4b));
816 ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
817 T2B = BYTW(&(W[TWVL * 10]), VSUB(T2x, T2A));
818 ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 6)]));
819 T24 = BYTW(&(W[TWVL * 10]), VSUB(T20, T23));
820 ST(&(x[WS(vs, 6) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
821 }
822 {
823 V T10, T1x, T3F, T38, T1y, Tu;
824 T10 = BYTW(&(W[TWVL * 10]), VSUB(TW, TZ));
825 ST(&(x[WS(vs, 6) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
826 T1x = BYTW(&(W[TWVL * 10]), VSUB(T1t, T1w));
827 ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 6)]));
828 T3F = BYTW(&(W[TWVL * 10]), VSUB(T3B, T3E));
829 ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 6)]));
830 T38 = BYTW(&(W[TWVL * 10]), VSUB(T34, T37));
831 ST(&(x[WS(vs, 6) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
832 T1y = BYTW(&(W[TWVL * 2]), VADD(T1t, T1w));
833 ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1y, ms, &(x[WS(vs, 2)]));
834 Tu = BYTW(&(W[TWVL * 2]), VADD(Tp, Ts));
835 ST(&(x[WS(vs, 2)]), Tu, ms, &(x[WS(vs, 2)]));
836 }
837 {
838 V T2C, T3G, T11, T25, T39, T4d;
839 T2C = BYTW(&(W[TWVL * 2]), VADD(T2x, T2A));
840 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T2C, ms, &(x[WS(vs, 2)]));
841 T3G = BYTW(&(W[TWVL * 2]), VADD(T3B, T3E));
842 ST(&(x[WS(vs, 2) + WS(rs, 6)]), T3G, ms, &(x[WS(vs, 2)]));
843 T11 = BYTW(&(W[TWVL * 2]), VADD(TW, TZ));
844 ST(&(x[WS(vs, 2) + WS(rs, 1)]), T11, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
845 T25 = BYTW(&(W[TWVL * 2]), VADD(T20, T23));
846 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T25, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
847 T39 = BYTW(&(W[TWVL * 2]), VADD(T34, T37));
848 ST(&(x[WS(vs, 2) + WS(rs, 5)]), T39, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
849 T4d = BYTW(&(W[TWVL * 2]), VADD(T48, T4b));
850 ST(&(x[WS(vs, 2) + WS(rs, 7)]), T4d, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
851 }
852 {
853 V Tx, T1B, T3c, T4g, T3J, T2F;
854 Tx = BYTW(&(W[TWVL * 6]), VSUB(Tv, Tw));
855 ST(&(x[WS(vs, 4)]), Tx, ms, &(x[WS(vs, 4)]));
856 T1B = BYTW(&(W[TWVL * 6]), VSUB(T1z, T1A));
857 ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 4)]));
858 T3c = BYTW(&(W[TWVL * 6]), VSUB(T3a, T3b));
859 ST(&(x[WS(vs, 4) + WS(rs, 5)]), T3c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
860 T4g = BYTW(&(W[TWVL * 6]), VSUB(T4e, T4f));
861 ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4g, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
862 T3J = BYTW(&(W[TWVL * 6]), VSUB(T3H, T3I));
863 ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3J, ms, &(x[WS(vs, 4)]));
864 T2F = BYTW(&(W[TWVL * 6]), VSUB(T2D, T2E));
865 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2F, ms, &(x[WS(vs, 4)]));
866 }
867 T28 = BYTW(&(W[TWVL * 6]), VSUB(T26, T27));
868 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T28, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
869 T14 = BYTW(&(W[TWVL * 6]), VSUB(T12, T13));
870 ST(&(x[WS(vs, 4) + WS(rs, 1)]), T14, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
871 {
872 V Th, Ti, Tb, Tg;
873 Tb = VBYI(VSUB(T7, Ta));
874 Tg = VSUB(Te, Tf);
875 Th = BYTW(&(W[TWVL * 4]), VADD(Tb, Tg));
876 Ti = BYTW(&(W[TWVL * 8]), VSUB(Tg, Tb));
877 ST(&(x[WS(vs, 3)]), Th, ms, &(x[WS(vs, 3)]));
878 ST(&(x[WS(vs, 5)]), Ti, ms, &(x[WS(vs, 5)]));
879 }
880 {
881 V T40, T41, T3U, T3Z;
882 T3U = VBYI(VSUB(T3Q, T3T));
883 T3Z = VSUB(T3X, T3Y);
884 T40 = BYTW(&(W[TWVL * 4]), VADD(T3U, T3Z));
885 T41 = BYTW(&(W[TWVL * 8]), VSUB(T3Z, T3U));
886 ST(&(x[WS(vs, 3) + WS(rs, 7)]), T40, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
887 ST(&(x[WS(vs, 5) + WS(rs, 7)]), T41, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
888 }
889 {
890 V T2p, T2q, T2j, T2o;
891 T2j = VBYI(VSUB(T2f, T2i));
892 T2o = VSUB(T2m, T2n);
893 T2p = BYTW(&(W[TWVL * 4]), VADD(T2j, T2o));
894 T2q = BYTW(&(W[TWVL * 8]), VSUB(T2o, T2j));
895 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T2p, ms, &(x[WS(vs, 3)]));
896 ST(&(x[WS(vs, 5) + WS(rs, 4)]), T2q, ms, &(x[WS(vs, 5)]));
897 }
898 {
899 V T1S, T1T, T1M, T1R;
900 T1M = VBYI(VSUB(T1I, T1L));
901 T1R = VSUB(T1P, T1Q);
902 T1S = BYTW(&(W[TWVL * 4]), VADD(T1M, T1R));
903 T1T = BYTW(&(W[TWVL * 8]), VSUB(T1R, T1M));
904 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1S, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
905 ST(&(x[WS(vs, 5) + WS(rs, 3)]), T1T, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
906 }
907 {
908 V TO, TP, TI, TN;
909 TI = VBYI(VSUB(TE, TH));
910 TN = VSUB(TL, TM);
911 TO = BYTW(&(W[TWVL * 4]), VADD(TI, TN));
912 TP = BYTW(&(W[TWVL * 8]), VSUB(TN, TI));
913 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TO, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
914 ST(&(x[WS(vs, 5) + WS(rs, 1)]), TP, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
915 }
916 {
917 V T1l, T1m, T1f, T1k;
918 T1f = VBYI(VSUB(T1b, T1e));
919 T1k = VSUB(T1i, T1j);
920 T1l = BYTW(&(W[TWVL * 4]), VADD(T1f, T1k));
921 T1m = BYTW(&(W[TWVL * 8]), VSUB(T1k, T1f));
922 ST(&(x[WS(vs, 3) + WS(rs, 2)]), T1l, ms, &(x[WS(vs, 3)]));
923 ST(&(x[WS(vs, 5) + WS(rs, 2)]), T1m, ms, &(x[WS(vs, 5)]));
924 }
925 {
926 V T3t, T3u, T3n, T3s;
927 T3n = VBYI(VSUB(T3j, T3m));
928 T3s = VSUB(T3q, T3r);
929 T3t = BYTW(&(W[TWVL * 4]), VADD(T3n, T3s));
930 T3u = BYTW(&(W[TWVL * 8]), VSUB(T3s, T3n));
931 ST(&(x[WS(vs, 3) + WS(rs, 6)]), T3t, ms, &(x[WS(vs, 3)]));
932 ST(&(x[WS(vs, 5) + WS(rs, 6)]), T3u, ms, &(x[WS(vs, 5)]));
933 }
934 {
935 V T2W, T2X, T2Q, T2V;
936 T2Q = VBYI(VSUB(T2M, T2P));
937 T2V = VSUB(T2T, T2U);
938 T2W = BYTW(&(W[TWVL * 4]), VADD(T2Q, T2V));
939 T2X = BYTW(&(W[TWVL * 8]), VSUB(T2V, T2Q));
940 ST(&(x[WS(vs, 3) + WS(rs, 5)]), T2W, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
941 ST(&(x[WS(vs, 5) + WS(rs, 5)]), T2X, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
942 }
943 {
944 V T1p, T1q, T1n, T1o;
945 T1n = VBYI(VADD(T1e, T1b));
946 T1o = VADD(T1i, T1j);
947 T1p = BYTW(&(W[0]), VADD(T1n, T1o));
948 T1q = BYTW(&(W[TWVL * 12]), VSUB(T1o, T1n));
949 ST(&(x[WS(vs, 1) + WS(rs, 2)]), T1p, ms, &(x[WS(vs, 1)]));
950 ST(&(x[WS(vs, 7) + WS(rs, 2)]), T1q, ms, &(x[WS(vs, 7)]));
951 }
952 {
953 V Tl, Tm, Tj, Tk;
954 Tj = VBYI(VADD(Ta, T7));
955 Tk = VADD(Te, Tf);
956 Tl = BYTW(&(W[0]), VADD(Tj, Tk));
957 Tm = BYTW(&(W[TWVL * 12]), VSUB(Tk, Tj));
958 ST(&(x[WS(vs, 1)]), Tl, ms, &(x[WS(vs, 1)]));
959 ST(&(x[WS(vs, 7)]), Tm, ms, &(x[WS(vs, 7)]));
960 }
961 {
962 V T2t, T2u, T2r, T2s;
963 T2r = VBYI(VADD(T2i, T2f));
964 T2s = VADD(T2m, T2n);
965 T2t = BYTW(&(W[0]), VADD(T2r, T2s));
966 T2u = BYTW(&(W[TWVL * 12]), VSUB(T2s, T2r));
967 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T2t, ms, &(x[WS(vs, 1)]));
968 ST(&(x[WS(vs, 7) + WS(rs, 4)]), T2u, ms, &(x[WS(vs, 7)]));
969 }
970 {
971 V T3x, T3y, T3v, T3w;
972 T3v = VBYI(VADD(T3m, T3j));
973 T3w = VADD(T3q, T3r);
974 T3x = BYTW(&(W[0]), VADD(T3v, T3w));
975 T3y = BYTW(&(W[TWVL * 12]), VSUB(T3w, T3v));
976 ST(&(x[WS(vs, 1) + WS(rs, 6)]), T3x, ms, &(x[WS(vs, 1)]));
977 ST(&(x[WS(vs, 7) + WS(rs, 6)]), T3y, ms, &(x[WS(vs, 7)]));
978 }
979 {
980 V TS, TT, TQ, TR;
981 TQ = VBYI(VADD(TH, TE));
982 TR = VADD(TL, TM);
983 TS = BYTW(&(W[0]), VADD(TQ, TR));
984 TT = BYTW(&(W[TWVL * 12]), VSUB(TR, TQ));
985 ST(&(x[WS(vs, 1) + WS(rs, 1)]), TS, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
986 ST(&(x[WS(vs, 7) + WS(rs, 1)]), TT, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
987 }
988 {
989 V T1W, T1X, T1U, T1V;
990 T1U = VBYI(VADD(T1L, T1I));
991 T1V = VADD(T1P, T1Q);
992 T1W = BYTW(&(W[0]), VADD(T1U, T1V));
993 T1X = BYTW(&(W[TWVL * 12]), VSUB(T1V, T1U));
994 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1W, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
995 ST(&(x[WS(vs, 7) + WS(rs, 3)]), T1X, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
996 }
997 {
998 V T30, T31, T2Y, T2Z;
999 T2Y = VBYI(VADD(T2P, T2M));
1000 T2Z = VADD(T2T, T2U);
1001 T30 = BYTW(&(W[0]), VADD(T2Y, T2Z));
1002 T31 = BYTW(&(W[TWVL * 12]), VSUB(T2Z, T2Y));
1003 ST(&(x[WS(vs, 1) + WS(rs, 5)]), T30, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
1004 ST(&(x[WS(vs, 7) + WS(rs, 5)]), T31, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
1005 }
1006 {
1007 V T44, T45, T42, T43;
1008 T42 = VBYI(VADD(T3T, T3Q));
1009 T43 = VADD(T3X, T3Y);
1010 T44 = BYTW(&(W[0]), VADD(T42, T43));
1011 T45 = BYTW(&(W[TWVL * 12]), VSUB(T43, T42));
1012 ST(&(x[WS(vs, 1) + WS(rs, 7)]), T44, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
1013 ST(&(x[WS(vs, 7) + WS(rs, 7)]), T45, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
1014 }
1015 }
1016 }
1017 VLEAVE();
1018 }
1019
1020 static const tw_instr twinstr[] = {
1021 VTW(0, 1),
1022 VTW(0, 2),
1023 VTW(0, 3),
1024 VTW(0, 4),
1025 VTW(0, 5),
1026 VTW(0, 6),
1027 VTW(0, 7),
1028 {TW_NEXT, VL, 0}
1029 };
1030
1031 static const ct_desc desc = { 8, XSIMD_STRING("q1bv_8"), twinstr, &GENUS, {264, 128, 0, 0}, 0, 0, 0 };
1032
1033 void XSIMD(codelet_q1bv_8) (planner *p) {
1034 X(kdft_difsq_register) (p, q1bv_8, &desc);
1035 }
1036 #endif