Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/q1fv_8.c @ 10:37bf6b4a2645

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
9:c0fb53affa76 10:37bf6b4a2645
1 /*
2 * Copyright (c) 2003, 2007-11 Matteo Frigo
3 * Copyright (c) 2003, 2007-11 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 Nov 25 07:39:31 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twidsq_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 8 -dif -name q1fv_8 -include q1f.h */
29
30 /*
31 * This function contains 264 FP additions, 192 FP multiplications,
32 * (or, 184 additions, 112 multiplications, 80 fused multiply/add),
33 * 117 stack variables, 1 constants, and 128 memory accesses
34 */
35 #include "q1f.h"
36
37 static void q1fv_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 = ri;
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 T42, T43, T1U, T1V, T2Y, T2Z, TT, TS;
46 {
47 V T3, Te, T1E, T1P, Tu, Tp, T25, T20, T2b, T2m, T3M, T2x, T2C, T3X, TA;
48 V TL, T48, T4d, T17, T11, TW, T1i, T2I, T1y, T1t, T2T, T3f, T3q, T34, T39;
49 V T3G, T3B, Ts, Tv, Tf, Ta, T23, T26, T1Q, T1L, T2A, T2D, T2n, T2i, T4b;
50 V T4e, T3Y, T3T, TZ, T12, TM, TH, T35, T2L, T3j, T1w, T1z, T1j, T1e, T36;
51 V T2O, T3C, T3i, T3k;
52 {
53 V T3d, T32, T3e, T3o, T3p, T33;
54 {
55 V T2v, T2w, T3V, T46, T3W;
56 {
57 V T1, T2, Tc, Td, T1C, T1D, T1N, T1O;
58 T1 = LD(&(x[0]), ms, &(x[0]));
59 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
60 Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
61 Td = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
62 T1C = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
63 T1D = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
64 T1N = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
65 T1O = LD(&(x[WS(vs, 3) + WS(rs, 6)]), ms, &(x[WS(vs, 3)]));
66 {
67 V T29, T1Y, T1Z, T2a, T2k, T2l, Tn, To, T3K, T3L;
68 T29 = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
69 T3 = VSUB(T1, T2);
70 Tn = VADD(T1, T2);
71 Te = VSUB(Tc, Td);
72 To = VADD(Tc, Td);
73 T1E = VSUB(T1C, T1D);
74 T1Y = VADD(T1C, T1D);
75 T1P = VSUB(T1N, T1O);
76 T1Z = VADD(T1N, T1O);
77 T2a = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
78 T2k = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
79 T2l = LD(&(x[WS(vs, 4) + WS(rs, 6)]), ms, &(x[WS(vs, 4)]));
80 Tu = VSUB(Tn, To);
81 Tp = VADD(Tn, To);
82 T3K = LD(&(x[WS(vs, 7)]), ms, &(x[WS(vs, 7)]));
83 T3L = LD(&(x[WS(vs, 7) + WS(rs, 4)]), ms, &(x[WS(vs, 7)]));
84 T25 = VSUB(T1Y, T1Z);
85 T20 = VADD(T1Y, T1Z);
86 T2v = VADD(T29, T2a);
87 T2b = VSUB(T29, T2a);
88 T2w = VADD(T2k, T2l);
89 T2m = VSUB(T2k, T2l);
90 T3V = LD(&(x[WS(vs, 7) + WS(rs, 2)]), ms, &(x[WS(vs, 7)]));
91 T46 = VADD(T3K, T3L);
92 T3M = VSUB(T3K, T3L);
93 T3W = LD(&(x[WS(vs, 7) + WS(rs, 6)]), ms, &(x[WS(vs, 7)]));
94 }
95 }
96 {
97 V T15, TU, T16, T1g, TV, T1h;
98 {
99 V Ty, Tz, TJ, TK, T47;
100 Ty = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
101 Tz = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
102 TJ = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
103 T2x = VADD(T2v, T2w);
104 T2C = VSUB(T2v, T2w);
105 TK = LD(&(x[WS(vs, 1) + WS(rs, 6)]), ms, &(x[WS(vs, 1)]));
106 T47 = VADD(T3V, T3W);
107 T3X = VSUB(T3V, T3W);
108 T15 = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
109 TA = VSUB(Ty, Tz);
110 TU = VADD(Ty, Tz);
111 T16 = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
112 T1g = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
113 TL = VSUB(TJ, TK);
114 TV = VADD(TJ, TK);
115 T48 = VADD(T46, T47);
116 T4d = VSUB(T46, T47);
117 T1h = LD(&(x[WS(vs, 2) + WS(rs, 6)]), ms, &(x[WS(vs, 2)]));
118 }
119 {
120 V T2G, T1r, T2H, T2R, T1s, T2S;
121 T2G = LD(&(x[WS(vs, 5)]), ms, &(x[WS(vs, 5)]));
122 T17 = VSUB(T15, T16);
123 T1r = VADD(T15, T16);
124 T2H = LD(&(x[WS(vs, 5) + WS(rs, 4)]), ms, &(x[WS(vs, 5)]));
125 T11 = VSUB(TU, TV);
126 TW = VADD(TU, TV);
127 T2R = LD(&(x[WS(vs, 5) + WS(rs, 2)]), ms, &(x[WS(vs, 5)]));
128 T1i = VSUB(T1g, T1h);
129 T1s = VADD(T1g, T1h);
130 T2S = LD(&(x[WS(vs, 5) + WS(rs, 6)]), ms, &(x[WS(vs, 5)]));
131 T3d = LD(&(x[WS(vs, 6)]), ms, &(x[WS(vs, 6)]));
132 T2I = VSUB(T2G, T2H);
133 T32 = VADD(T2G, T2H);
134 T3e = LD(&(x[WS(vs, 6) + WS(rs, 4)]), ms, &(x[WS(vs, 6)]));
135 T3o = LD(&(x[WS(vs, 6) + WS(rs, 2)]), ms, &(x[WS(vs, 6)]));
136 T3p = LD(&(x[WS(vs, 6) + WS(rs, 6)]), ms, &(x[WS(vs, 6)]));
137 T1y = VSUB(T1r, T1s);
138 T1t = VADD(T1r, T1s);
139 T33 = VADD(T2R, T2S);
140 T2T = VSUB(T2R, T2S);
141 }
142 }
143 }
144 {
145 V T2y, T2e, T3Q, T2z, T2h, T49, T3P, T3R;
146 {
147 V T6, Tq, T1I, Tr, T9, T21, T1H, T1J;
148 {
149 V T4, T3z, T3A, T5, T7, T8, T1F, T1G;
150 T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
151 T3f = VSUB(T3d, T3e);
152 T3z = VADD(T3d, T3e);
153 T3q = VSUB(T3o, T3p);
154 T3A = VADD(T3o, T3p);
155 T5 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
156 T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
157 T34 = VADD(T32, T33);
158 T39 = VSUB(T32, T33);
159 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
160 T1F = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
161 T1G = LD(&(x[WS(vs, 3) + WS(rs, 5)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
162 T3G = VSUB(T3z, T3A);
163 T3B = VADD(T3z, T3A);
164 T6 = VSUB(T4, T5);
165 Tq = VADD(T4, T5);
166 T1I = LD(&(x[WS(vs, 3) + WS(rs, 7)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
167 Tr = VADD(T7, T8);
168 T9 = VSUB(T7, T8);
169 T21 = VADD(T1F, T1G);
170 T1H = VSUB(T1F, T1G);
171 T1J = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
172 }
173 {
174 V T2f, T22, T1K, T2g, T2c, T2d, T3N, T3O;
175 T2c = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
176 T2d = LD(&(x[WS(vs, 4) + WS(rs, 5)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
177 T2f = LD(&(x[WS(vs, 4) + WS(rs, 7)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
178 Ts = VADD(Tq, Tr);
179 Tv = VSUB(Tr, Tq);
180 Tf = VSUB(T9, T6);
181 Ta = VADD(T6, T9);
182 T22 = VADD(T1I, T1J);
183 T1K = VSUB(T1I, T1J);
184 T2y = VADD(T2c, T2d);
185 T2e = VSUB(T2c, T2d);
186 T2g = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
187 T3N = LD(&(x[WS(vs, 7) + WS(rs, 1)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
188 T3O = LD(&(x[WS(vs, 7) + WS(rs, 5)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
189 T3Q = LD(&(x[WS(vs, 7) + WS(rs, 7)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
190 T23 = VADD(T21, T22);
191 T26 = VSUB(T22, T21);
192 T1Q = VSUB(T1K, T1H);
193 T1L = VADD(T1H, T1K);
194 T2z = VADD(T2f, T2g);
195 T2h = VSUB(T2f, T2g);
196 T49 = VADD(T3N, T3O);
197 T3P = VSUB(T3N, T3O);
198 T3R = LD(&(x[WS(vs, 7) + WS(rs, 3)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
199 }
200 }
201 {
202 V TX, TD, T1b, TY, TG, T1u, T1a, T1c;
203 {
204 V TE, T4a, T3S, TF, TB, TC, T18, T19;
205 TB = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
206 TC = LD(&(x[WS(vs, 1) + WS(rs, 5)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
207 TE = LD(&(x[WS(vs, 1) + WS(rs, 7)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
208 T2A = VADD(T2y, T2z);
209 T2D = VSUB(T2z, T2y);
210 T2n = VSUB(T2h, T2e);
211 T2i = VADD(T2e, T2h);
212 T4a = VADD(T3Q, T3R);
213 T3S = VSUB(T3Q, T3R);
214 TX = VADD(TB, TC);
215 TD = VSUB(TB, TC);
216 TF = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
217 T18 = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
218 T19 = LD(&(x[WS(vs, 2) + WS(rs, 5)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
219 T1b = LD(&(x[WS(vs, 2) + WS(rs, 7)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
220 T4b = VADD(T49, T4a);
221 T4e = VSUB(T4a, T49);
222 T3Y = VSUB(T3S, T3P);
223 T3T = VADD(T3P, T3S);
224 TY = VADD(TE, TF);
225 TG = VSUB(TE, TF);
226 T1u = VADD(T18, T19);
227 T1a = VSUB(T18, T19);
228 T1c = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
229 }
230 {
231 V T2M, T1v, T1d, T2N, T2J, T2K, T3g, T3h;
232 T2J = LD(&(x[WS(vs, 5) + WS(rs, 1)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
233 T2K = LD(&(x[WS(vs, 5) + WS(rs, 5)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
234 T2M = LD(&(x[WS(vs, 5) + WS(rs, 7)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
235 TZ = VADD(TX, TY);
236 T12 = VSUB(TY, TX);
237 TM = VSUB(TG, TD);
238 TH = VADD(TD, TG);
239 T1v = VADD(T1b, T1c);
240 T1d = VSUB(T1b, T1c);
241 T35 = VADD(T2J, T2K);
242 T2L = VSUB(T2J, T2K);
243 T2N = LD(&(x[WS(vs, 5) + WS(rs, 3)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
244 T3g = LD(&(x[WS(vs, 6) + WS(rs, 1)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
245 T3h = LD(&(x[WS(vs, 6) + WS(rs, 5)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
246 T3j = LD(&(x[WS(vs, 6) + WS(rs, 7)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
247 T1w = VADD(T1u, T1v);
248 T1z = VSUB(T1v, T1u);
249 T1j = VSUB(T1d, T1a);
250 T1e = VADD(T1a, T1d);
251 T36 = VADD(T2M, T2N);
252 T2O = VSUB(T2M, T2N);
253 T3C = VADD(T3g, T3h);
254 T3i = VSUB(T3g, T3h);
255 T3k = LD(&(x[WS(vs, 6) + WS(rs, 3)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
256 }
257 }
258 }
259 }
260 {
261 V T3a, T2U, T2P, T3H, T3r, T3m, T13, T27, T3b, T4f;
262 {
263 V T37, T3E, T2B, T24;
264 {
265 V T3D, T3l, Tt, T4c;
266 ST(&(x[0]), VADD(Tp, Ts), ms, &(x[0]));
267 ST(&(x[WS(rs, 2)]), VADD(T1t, T1w), ms, &(x[0]));
268 ST(&(x[WS(rs, 7)]), VADD(T48, T4b), ms, &(x[WS(rs, 1)]));
269 T37 = VADD(T35, T36);
270 T3a = VSUB(T36, T35);
271 T2U = VSUB(T2O, T2L);
272 T2P = VADD(T2L, T2O);
273 T3D = VADD(T3j, T3k);
274 T3l = VSUB(T3j, T3k);
275 ST(&(x[WS(rs, 4)]), VADD(T2x, T2A), ms, &(x[0]));
276 ST(&(x[WS(rs, 3)]), VADD(T20, T23), ms, &(x[WS(rs, 1)]));
277 ST(&(x[WS(rs, 5)]), VADD(T34, T37), ms, &(x[WS(rs, 1)]));
278 ST(&(x[WS(rs, 1)]), VADD(TW, TZ), ms, &(x[WS(rs, 1)]));
279 Tt = BYTWJ(&(W[TWVL * 6]), VSUB(Tp, Ts));
280 T4c = BYTWJ(&(W[TWVL * 6]), VSUB(T48, T4b));
281 T3E = VADD(T3C, T3D);
282 T3H = VSUB(T3D, T3C);
283 T3r = VSUB(T3l, T3i);
284 T3m = VADD(T3i, T3l);
285 T2B = BYTWJ(&(W[TWVL * 6]), VSUB(T2x, T2A));
286 T24 = BYTWJ(&(W[TWVL * 6]), VSUB(T20, T23));
287 ST(&(x[WS(vs, 4)]), Tt, ms, &(x[WS(vs, 4)]));
288 ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
289 ST(&(x[WS(rs, 6)]), VADD(T3B, T3E), ms, &(x[0]));
290 }
291 {
292 V T38, T1A, Tw, T10, T1x, T3F, T2E, T3I;
293 T10 = BYTWJ(&(W[TWVL * 6]), VSUB(TW, TZ));
294 T1x = BYTWJ(&(W[TWVL * 6]), VSUB(T1t, T1w));
295 T3F = BYTWJ(&(W[TWVL * 6]), VSUB(T3B, T3E));
296 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 4)]));
297 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
298 T38 = BYTWJ(&(W[TWVL * 6]), VSUB(T34, T37));
299 T1A = BYTWJ(&(W[TWVL * 10]), VFNMSI(T1z, T1y));
300 Tw = BYTWJ(&(W[TWVL * 10]), VFNMSI(Tv, Tu));
301 ST(&(x[WS(vs, 4) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
302 ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 4)]));
303 ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 4)]));
304 T2E = BYTWJ(&(W[TWVL * 10]), VFNMSI(T2D, T2C));
305 T3I = BYTWJ(&(W[TWVL * 10]), VFNMSI(T3H, T3G));
306 ST(&(x[WS(vs, 4) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
307 ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1A, ms, &(x[WS(vs, 6)]));
308 ST(&(x[WS(vs, 6)]), Tw, ms, &(x[WS(vs, 6)]));
309 T13 = BYTWJ(&(W[TWVL * 10]), VFNMSI(T12, T11));
310 T27 = BYTWJ(&(W[TWVL * 10]), VFNMSI(T26, T25));
311 T3b = BYTWJ(&(W[TWVL * 10]), VFNMSI(T3a, T39));
312 ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2E, ms, &(x[WS(vs, 6)]));
313 ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3I, ms, &(x[WS(vs, 6)]));
314 T4f = BYTWJ(&(W[TWVL * 10]), VFNMSI(T4e, T4d));
315 }
316 }
317 {
318 V Tj, Tk, T2r, T2j, Ti, Th, T2o, T2s, T1M, T1R, T41, T40;
319 {
320 V T3c, T4g, T3J, T2F, Tx, T1B;
321 Tx = BYTWJ(&(W[TWVL * 2]), VFMAI(Tv, Tu));
322 T1B = BYTWJ(&(W[TWVL * 2]), VFMAI(T1z, T1y));
323 ST(&(x[WS(vs, 6) + WS(rs, 1)]), T13, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
324 ST(&(x[WS(vs, 6) + WS(rs, 3)]), T27, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
325 ST(&(x[WS(vs, 6) + WS(rs, 5)]), T3b, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
326 T3c = BYTWJ(&(W[TWVL * 2]), VFMAI(T3a, T39));
327 T4g = BYTWJ(&(W[TWVL * 2]), VFMAI(T4e, T4d));
328 ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4f, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
329 ST(&(x[WS(vs, 2)]), Tx, ms, &(x[WS(vs, 2)]));
330 ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 2)]));
331 T3J = BYTWJ(&(W[TWVL * 2]), VFMAI(T3H, T3G));
332 T2F = BYTWJ(&(W[TWVL * 2]), VFMAI(T2D, T2C));
333 {
334 V T14, Tb, Tg, T28, T3U, T3Z;
335 T28 = BYTWJ(&(W[TWVL * 2]), VFMAI(T26, T25));
336 ST(&(x[WS(vs, 2) + WS(rs, 5)]), T3c, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
337 ST(&(x[WS(vs, 2) + WS(rs, 7)]), T4g, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
338 T14 = BYTWJ(&(W[TWVL * 2]), VFMAI(T12, T11));
339 Tj = VFNMS(LDK(KP707106781), Ta, T3);
340 Tb = VFMA(LDK(KP707106781), Ta, T3);
341 Tg = VFNMS(LDK(KP707106781), Tf, Te);
342 Tk = VFMA(LDK(KP707106781), Tf, Te);
343 ST(&(x[WS(vs, 2) + WS(rs, 6)]), T3J, ms, &(x[WS(vs, 2)]));
344 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T2F, ms, &(x[WS(vs, 2)]));
345 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T28, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
346 T3U = VFMA(LDK(KP707106781), T3T, T3M);
347 T42 = VFNMS(LDK(KP707106781), T3T, T3M);
348 T43 = VFMA(LDK(KP707106781), T3Y, T3X);
349 T3Z = VFNMS(LDK(KP707106781), T3Y, T3X);
350 ST(&(x[WS(vs, 2) + WS(rs, 1)]), T14, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
351 T2r = VFNMS(LDK(KP707106781), T2i, T2b);
352 T2j = VFMA(LDK(KP707106781), T2i, T2b);
353 Ti = BYTWJ(&(W[TWVL * 12]), VFMAI(Tg, Tb));
354 Th = BYTWJ(&(W[0]), VFNMSI(Tg, Tb));
355 T2o = VFNMS(LDK(KP707106781), T2n, T2m);
356 T2s = VFMA(LDK(KP707106781), T2n, T2m);
357 T1U = VFNMS(LDK(KP707106781), T1L, T1E);
358 T1M = VFMA(LDK(KP707106781), T1L, T1E);
359 T1R = VFNMS(LDK(KP707106781), T1Q, T1P);
360 T1V = VFMA(LDK(KP707106781), T1Q, T1P);
361 T41 = BYTWJ(&(W[TWVL * 12]), VFMAI(T3Z, T3U));
362 T40 = BYTWJ(&(W[0]), VFNMSI(T3Z, T3U));
363 }
364 }
365 {
366 V TQ, TR, T1n, T1o, T3v, T3w;
367 {
368 V T1f, T1k, T3n, TP, TO, T3s, T2Q, T2V;
369 {
370 V TI, T2q, T2p, T1T, T1S, TN;
371 TQ = VFNMS(LDK(KP707106781), TH, TA);
372 TI = VFMA(LDK(KP707106781), TH, TA);
373 ST(&(x[WS(vs, 7)]), Ti, ms, &(x[WS(vs, 7)]));
374 ST(&(x[WS(vs, 1)]), Th, ms, &(x[WS(vs, 1)]));
375 T2q = BYTWJ(&(W[TWVL * 12]), VFMAI(T2o, T2j));
376 T2p = BYTWJ(&(W[0]), VFNMSI(T2o, T2j));
377 T1T = BYTWJ(&(W[TWVL * 12]), VFMAI(T1R, T1M));
378 T1S = BYTWJ(&(W[0]), VFNMSI(T1R, T1M));
379 ST(&(x[WS(vs, 7) + WS(rs, 7)]), T41, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
380 ST(&(x[WS(vs, 1) + WS(rs, 7)]), T40, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
381 TN = VFNMS(LDK(KP707106781), TM, TL);
382 TR = VFMA(LDK(KP707106781), TM, TL);
383 T1n = VFNMS(LDK(KP707106781), T1e, T17);
384 T1f = VFMA(LDK(KP707106781), T1e, T17);
385 ST(&(x[WS(vs, 7) + WS(rs, 4)]), T2q, ms, &(x[WS(vs, 7)]));
386 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T2p, ms, &(x[WS(vs, 1)]));
387 ST(&(x[WS(vs, 7) + WS(rs, 3)]), T1T, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
388 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1S, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
389 T1k = VFNMS(LDK(KP707106781), T1j, T1i);
390 T1o = VFMA(LDK(KP707106781), T1j, T1i);
391 T3v = VFNMS(LDK(KP707106781), T3m, T3f);
392 T3n = VFMA(LDK(KP707106781), T3m, T3f);
393 TP = BYTWJ(&(W[TWVL * 12]), VFMAI(TN, TI));
394 TO = BYTWJ(&(W[0]), VFNMSI(TN, TI));
395 T3s = VFNMS(LDK(KP707106781), T3r, T3q);
396 T3w = VFMA(LDK(KP707106781), T3r, T3q);
397 }
398 T2Y = VFNMS(LDK(KP707106781), T2P, T2I);
399 T2Q = VFMA(LDK(KP707106781), T2P, T2I);
400 T2V = VFNMS(LDK(KP707106781), T2U, T2T);
401 T2Z = VFMA(LDK(KP707106781), T2U, T2T);
402 {
403 V T3u, T3t, T2X, T2W, T1m, T1l;
404 T1m = BYTWJ(&(W[TWVL * 12]), VFMAI(T1k, T1f));
405 T1l = BYTWJ(&(W[0]), VFNMSI(T1k, T1f));
406 ST(&(x[WS(vs, 7) + WS(rs, 1)]), TP, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
407 ST(&(x[WS(vs, 1) + WS(rs, 1)]), TO, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
408 T3u = BYTWJ(&(W[TWVL * 12]), VFMAI(T3s, T3n));
409 T3t = BYTWJ(&(W[0]), VFNMSI(T3s, T3n));
410 T2X = BYTWJ(&(W[TWVL * 12]), VFMAI(T2V, T2Q));
411 T2W = BYTWJ(&(W[0]), VFNMSI(T2V, T2Q));
412 ST(&(x[WS(vs, 7) + WS(rs, 2)]), T1m, ms, &(x[WS(vs, 7)]));
413 ST(&(x[WS(vs, 1) + WS(rs, 2)]), T1l, ms, &(x[WS(vs, 1)]));
414 ST(&(x[WS(vs, 7) + WS(rs, 6)]), T3u, ms, &(x[WS(vs, 7)]));
415 ST(&(x[WS(vs, 1) + WS(rs, 6)]), T3t, ms, &(x[WS(vs, 1)]));
416 ST(&(x[WS(vs, 7) + WS(rs, 5)]), T2X, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
417 ST(&(x[WS(vs, 1) + WS(rs, 5)]), T2W, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
418 }
419 }
420 {
421 V T2u, T2t, T3y, T3x;
422 {
423 V T1q, T1p, Tm, Tl;
424 T1q = BYTWJ(&(W[TWVL * 4]), VFMAI(T1o, T1n));
425 T1p = BYTWJ(&(W[TWVL * 8]), VFNMSI(T1o, T1n));
426 Tm = BYTWJ(&(W[TWVL * 4]), VFMAI(Tk, Tj));
427 Tl = BYTWJ(&(W[TWVL * 8]), VFNMSI(Tk, Tj));
428 ST(&(x[WS(vs, 3) + WS(rs, 2)]), T1q, ms, &(x[WS(vs, 3)]));
429 ST(&(x[WS(vs, 5) + WS(rs, 2)]), T1p, ms, &(x[WS(vs, 5)]));
430 T2u = BYTWJ(&(W[TWVL * 4]), VFMAI(T2s, T2r));
431 T2t = BYTWJ(&(W[TWVL * 8]), VFNMSI(T2s, T2r));
432 T3y = BYTWJ(&(W[TWVL * 4]), VFMAI(T3w, T3v));
433 T3x = BYTWJ(&(W[TWVL * 8]), VFNMSI(T3w, T3v));
434 ST(&(x[WS(vs, 3)]), Tm, ms, &(x[WS(vs, 3)]));
435 ST(&(x[WS(vs, 5)]), Tl, ms, &(x[WS(vs, 5)]));
436 }
437 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T2u, ms, &(x[WS(vs, 3)]));
438 ST(&(x[WS(vs, 5) + WS(rs, 4)]), T2t, ms, &(x[WS(vs, 5)]));
439 ST(&(x[WS(vs, 3) + WS(rs, 6)]), T3y, ms, &(x[WS(vs, 3)]));
440 ST(&(x[WS(vs, 5) + WS(rs, 6)]), T3x, ms, &(x[WS(vs, 5)]));
441 TT = BYTWJ(&(W[TWVL * 4]), VFMAI(TR, TQ));
442 TS = BYTWJ(&(W[TWVL * 8]), VFNMSI(TR, TQ));
443 }
444 }
445 }
446 }
447 }
448 {
449 V T31, T30, T45, T44, T1X, T1W;
450 T1X = BYTWJ(&(W[TWVL * 4]), VFMAI(T1V, T1U));
451 T1W = BYTWJ(&(W[TWVL * 8]), VFNMSI(T1V, T1U));
452 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TT, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
453 ST(&(x[WS(vs, 5) + WS(rs, 1)]), TS, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
454 T31 = BYTWJ(&(W[TWVL * 4]), VFMAI(T2Z, T2Y));
455 T30 = BYTWJ(&(W[TWVL * 8]), VFNMSI(T2Z, T2Y));
456 T45 = BYTWJ(&(W[TWVL * 4]), VFMAI(T43, T42));
457 T44 = BYTWJ(&(W[TWVL * 8]), VFNMSI(T43, T42));
458 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1X, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
459 ST(&(x[WS(vs, 5) + WS(rs, 3)]), T1W, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
460 ST(&(x[WS(vs, 3) + WS(rs, 5)]), T31, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
461 ST(&(x[WS(vs, 5) + WS(rs, 5)]), T30, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
462 ST(&(x[WS(vs, 3) + WS(rs, 7)]), T45, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
463 ST(&(x[WS(vs, 5) + WS(rs, 7)]), T44, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
464 }
465 }
466 }
467 VLEAVE();
468 }
469
470 static const tw_instr twinstr[] = {
471 VTW(0, 1),
472 VTW(0, 2),
473 VTW(0, 3),
474 VTW(0, 4),
475 VTW(0, 5),
476 VTW(0, 6),
477 VTW(0, 7),
478 {TW_NEXT, VL, 0}
479 };
480
481 static const ct_desc desc = { 8, XSIMD_STRING("q1fv_8"), twinstr, &GENUS, {184, 112, 80, 0}, 0, 0, 0 };
482
483 void XSIMD(codelet_q1fv_8) (planner *p) {
484 X(kdft_difsq_register) (p, q1fv_8, &desc);
485 }
486 #else /* HAVE_FMA */
487
488 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -dif -name q1fv_8 -include q1f.h */
489
490 /*
491 * This function contains 264 FP additions, 128 FP multiplications,
492 * (or, 264 additions, 128 multiplications, 0 fused multiply/add),
493 * 77 stack variables, 1 constants, and 128 memory accesses
494 */
495 #include "q1f.h"
496
497 static void q1fv_8(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
498 {
499 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
500 {
501 INT m;
502 R *x;
503 x = ri;
504 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)) {
505 V T3, Tu, Tf, Tp, T1E, T25, T1Q, T20, T2b, T2C, T2n, T2x, T3M, T4d, T3Y;
506 V T48, TA, T11, TM, TW, T17, T1y, T1j, T1t, T2I, T39, T2U, T34, T3f, T3G;
507 V T3r, T3B, Ta, Tv, Tc, Ts, T1L, T26, T1N, T23, T2i, T2D, T2k, T2A, T3T;
508 V T4e, T3V, T4b, TH, T12, TJ, TZ, T1e, T1z, T1g, T1w, T2P, T3a, T2R, T37;
509 V T3m, T3H, T3o, T3E, T28, T14;
510 {
511 V T1, T2, Tn, Td, Te, To;
512 T1 = LD(&(x[0]), ms, &(x[0]));
513 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
514 Tn = VADD(T1, T2);
515 Td = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
516 Te = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
517 To = VADD(Td, Te);
518 T3 = VSUB(T1, T2);
519 Tu = VSUB(Tn, To);
520 Tf = VSUB(Td, Te);
521 Tp = VADD(Tn, To);
522 }
523 {
524 V T1C, T1D, T1Y, T1O, T1P, T1Z;
525 T1C = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
526 T1D = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
527 T1Y = VADD(T1C, T1D);
528 T1O = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
529 T1P = LD(&(x[WS(vs, 3) + WS(rs, 6)]), ms, &(x[WS(vs, 3)]));
530 T1Z = VADD(T1O, T1P);
531 T1E = VSUB(T1C, T1D);
532 T25 = VSUB(T1Y, T1Z);
533 T1Q = VSUB(T1O, T1P);
534 T20 = VADD(T1Y, T1Z);
535 }
536 {
537 V T29, T2a, T2v, T2l, T2m, T2w;
538 T29 = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
539 T2a = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
540 T2v = VADD(T29, T2a);
541 T2l = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
542 T2m = LD(&(x[WS(vs, 4) + WS(rs, 6)]), ms, &(x[WS(vs, 4)]));
543 T2w = VADD(T2l, T2m);
544 T2b = VSUB(T29, T2a);
545 T2C = VSUB(T2v, T2w);
546 T2n = VSUB(T2l, T2m);
547 T2x = VADD(T2v, T2w);
548 }
549 {
550 V T3K, T3L, T46, T3W, T3X, T47;
551 T3K = LD(&(x[WS(vs, 7)]), ms, &(x[WS(vs, 7)]));
552 T3L = LD(&(x[WS(vs, 7) + WS(rs, 4)]), ms, &(x[WS(vs, 7)]));
553 T46 = VADD(T3K, T3L);
554 T3W = LD(&(x[WS(vs, 7) + WS(rs, 2)]), ms, &(x[WS(vs, 7)]));
555 T3X = LD(&(x[WS(vs, 7) + WS(rs, 6)]), ms, &(x[WS(vs, 7)]));
556 T47 = VADD(T3W, T3X);
557 T3M = VSUB(T3K, T3L);
558 T4d = VSUB(T46, T47);
559 T3Y = VSUB(T3W, T3X);
560 T48 = VADD(T46, T47);
561 }
562 {
563 V Ty, Tz, TU, TK, TL, TV;
564 Ty = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
565 Tz = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
566 TU = VADD(Ty, Tz);
567 TK = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
568 TL = LD(&(x[WS(vs, 1) + WS(rs, 6)]), ms, &(x[WS(vs, 1)]));
569 TV = VADD(TK, TL);
570 TA = VSUB(Ty, Tz);
571 T11 = VSUB(TU, TV);
572 TM = VSUB(TK, TL);
573 TW = VADD(TU, TV);
574 }
575 {
576 V T15, T16, T1r, T1h, T1i, T1s;
577 T15 = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
578 T16 = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
579 T1r = VADD(T15, T16);
580 T1h = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
581 T1i = LD(&(x[WS(vs, 2) + WS(rs, 6)]), ms, &(x[WS(vs, 2)]));
582 T1s = VADD(T1h, T1i);
583 T17 = VSUB(T15, T16);
584 T1y = VSUB(T1r, T1s);
585 T1j = VSUB(T1h, T1i);
586 T1t = VADD(T1r, T1s);
587 }
588 {
589 V T2G, T2H, T32, T2S, T2T, T33;
590 T2G = LD(&(x[WS(vs, 5)]), ms, &(x[WS(vs, 5)]));
591 T2H = LD(&(x[WS(vs, 5) + WS(rs, 4)]), ms, &(x[WS(vs, 5)]));
592 T32 = VADD(T2G, T2H);
593 T2S = LD(&(x[WS(vs, 5) + WS(rs, 2)]), ms, &(x[WS(vs, 5)]));
594 T2T = LD(&(x[WS(vs, 5) + WS(rs, 6)]), ms, &(x[WS(vs, 5)]));
595 T33 = VADD(T2S, T2T);
596 T2I = VSUB(T2G, T2H);
597 T39 = VSUB(T32, T33);
598 T2U = VSUB(T2S, T2T);
599 T34 = VADD(T32, T33);
600 }
601 {
602 V T3d, T3e, T3z, T3p, T3q, T3A;
603 T3d = LD(&(x[WS(vs, 6)]), ms, &(x[WS(vs, 6)]));
604 T3e = LD(&(x[WS(vs, 6) + WS(rs, 4)]), ms, &(x[WS(vs, 6)]));
605 T3z = VADD(T3d, T3e);
606 T3p = LD(&(x[WS(vs, 6) + WS(rs, 2)]), ms, &(x[WS(vs, 6)]));
607 T3q = LD(&(x[WS(vs, 6) + WS(rs, 6)]), ms, &(x[WS(vs, 6)]));
608 T3A = VADD(T3p, T3q);
609 T3f = VSUB(T3d, T3e);
610 T3G = VSUB(T3z, T3A);
611 T3r = VSUB(T3p, T3q);
612 T3B = VADD(T3z, T3A);
613 }
614 {
615 V T6, Tq, T9, Tr;
616 {
617 V T4, T5, T7, T8;
618 T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
619 T5 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
620 T6 = VSUB(T4, T5);
621 Tq = VADD(T4, T5);
622 T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
623 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
624 T9 = VSUB(T7, T8);
625 Tr = VADD(T7, T8);
626 }
627 Ta = VMUL(LDK(KP707106781), VADD(T6, T9));
628 Tv = VBYI(VSUB(Tr, Tq));
629 Tc = VMUL(LDK(KP707106781), VSUB(T9, T6));
630 Ts = VADD(Tq, Tr);
631 }
632 {
633 V T1H, T21, T1K, T22;
634 {
635 V T1F, T1G, T1I, T1J;
636 T1F = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
637 T1G = LD(&(x[WS(vs, 3) + WS(rs, 5)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
638 T1H = VSUB(T1F, T1G);
639 T21 = VADD(T1F, T1G);
640 T1I = LD(&(x[WS(vs, 3) + WS(rs, 7)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
641 T1J = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
642 T1K = VSUB(T1I, T1J);
643 T22 = VADD(T1I, T1J);
644 }
645 T1L = VMUL(LDK(KP707106781), VADD(T1H, T1K));
646 T26 = VBYI(VSUB(T22, T21));
647 T1N = VMUL(LDK(KP707106781), VSUB(T1K, T1H));
648 T23 = VADD(T21, T22);
649 }
650 {
651 V T2e, T2y, T2h, T2z;
652 {
653 V T2c, T2d, T2f, T2g;
654 T2c = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
655 T2d = LD(&(x[WS(vs, 4) + WS(rs, 5)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
656 T2e = VSUB(T2c, T2d);
657 T2y = VADD(T2c, T2d);
658 T2f = LD(&(x[WS(vs, 4) + WS(rs, 7)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
659 T2g = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
660 T2h = VSUB(T2f, T2g);
661 T2z = VADD(T2f, T2g);
662 }
663 T2i = VMUL(LDK(KP707106781), VADD(T2e, T2h));
664 T2D = VBYI(VSUB(T2z, T2y));
665 T2k = VMUL(LDK(KP707106781), VSUB(T2h, T2e));
666 T2A = VADD(T2y, T2z);
667 }
668 {
669 V T3P, T49, T3S, T4a;
670 {
671 V T3N, T3O, T3Q, T3R;
672 T3N = LD(&(x[WS(vs, 7) + WS(rs, 1)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
673 T3O = LD(&(x[WS(vs, 7) + WS(rs, 5)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
674 T3P = VSUB(T3N, T3O);
675 T49 = VADD(T3N, T3O);
676 T3Q = LD(&(x[WS(vs, 7) + WS(rs, 7)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
677 T3R = LD(&(x[WS(vs, 7) + WS(rs, 3)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
678 T3S = VSUB(T3Q, T3R);
679 T4a = VADD(T3Q, T3R);
680 }
681 T3T = VMUL(LDK(KP707106781), VADD(T3P, T3S));
682 T4e = VBYI(VSUB(T4a, T49));
683 T3V = VMUL(LDK(KP707106781), VSUB(T3S, T3P));
684 T4b = VADD(T49, T4a);
685 }
686 {
687 V TD, TX, TG, TY;
688 {
689 V TB, TC, TE, TF;
690 TB = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
691 TC = LD(&(x[WS(vs, 1) + WS(rs, 5)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
692 TD = VSUB(TB, TC);
693 TX = VADD(TB, TC);
694 TE = LD(&(x[WS(vs, 1) + WS(rs, 7)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
695 TF = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
696 TG = VSUB(TE, TF);
697 TY = VADD(TE, TF);
698 }
699 TH = VMUL(LDK(KP707106781), VADD(TD, TG));
700 T12 = VBYI(VSUB(TY, TX));
701 TJ = VMUL(LDK(KP707106781), VSUB(TG, TD));
702 TZ = VADD(TX, TY);
703 }
704 {
705 V T1a, T1u, T1d, T1v;
706 {
707 V T18, T19, T1b, T1c;
708 T18 = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
709 T19 = LD(&(x[WS(vs, 2) + WS(rs, 5)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
710 T1a = VSUB(T18, T19);
711 T1u = VADD(T18, T19);
712 T1b = LD(&(x[WS(vs, 2) + WS(rs, 7)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
713 T1c = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
714 T1d = VSUB(T1b, T1c);
715 T1v = VADD(T1b, T1c);
716 }
717 T1e = VMUL(LDK(KP707106781), VADD(T1a, T1d));
718 T1z = VBYI(VSUB(T1v, T1u));
719 T1g = VMUL(LDK(KP707106781), VSUB(T1d, T1a));
720 T1w = VADD(T1u, T1v);
721 }
722 {
723 V T2L, T35, T2O, T36;
724 {
725 V T2J, T2K, T2M, T2N;
726 T2J = LD(&(x[WS(vs, 5) + WS(rs, 1)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
727 T2K = LD(&(x[WS(vs, 5) + WS(rs, 5)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
728 T2L = VSUB(T2J, T2K);
729 T35 = VADD(T2J, T2K);
730 T2M = LD(&(x[WS(vs, 5) + WS(rs, 7)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
731 T2N = LD(&(x[WS(vs, 5) + WS(rs, 3)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
732 T2O = VSUB(T2M, T2N);
733 T36 = VADD(T2M, T2N);
734 }
735 T2P = VMUL(LDK(KP707106781), VADD(T2L, T2O));
736 T3a = VBYI(VSUB(T36, T35));
737 T2R = VMUL(LDK(KP707106781), VSUB(T2O, T2L));
738 T37 = VADD(T35, T36);
739 }
740 {
741 V T3i, T3C, T3l, T3D;
742 {
743 V T3g, T3h, T3j, T3k;
744 T3g = LD(&(x[WS(vs, 6) + WS(rs, 1)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
745 T3h = LD(&(x[WS(vs, 6) + WS(rs, 5)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
746 T3i = VSUB(T3g, T3h);
747 T3C = VADD(T3g, T3h);
748 T3j = LD(&(x[WS(vs, 6) + WS(rs, 7)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
749 T3k = LD(&(x[WS(vs, 6) + WS(rs, 3)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
750 T3l = VSUB(T3j, T3k);
751 T3D = VADD(T3j, T3k);
752 }
753 T3m = VMUL(LDK(KP707106781), VADD(T3i, T3l));
754 T3H = VBYI(VSUB(T3D, T3C));
755 T3o = VMUL(LDK(KP707106781), VSUB(T3l, T3i));
756 T3E = VADD(T3C, T3D);
757 }
758 ST(&(x[0]), VADD(Tp, Ts), ms, &(x[0]));
759 ST(&(x[WS(rs, 2)]), VADD(T1t, T1w), ms, &(x[0]));
760 ST(&(x[WS(rs, 5)]), VADD(T34, T37), ms, &(x[WS(rs, 1)]));
761 ST(&(x[WS(rs, 7)]), VADD(T48, T4b), ms, &(x[WS(rs, 1)]));
762 ST(&(x[WS(rs, 6)]), VADD(T3B, T3E), ms, &(x[0]));
763 ST(&(x[WS(rs, 4)]), VADD(T2x, T2A), ms, &(x[0]));
764 {
765 V Tt, T4c, T2B, T24;
766 ST(&(x[WS(rs, 3)]), VADD(T20, T23), ms, &(x[WS(rs, 1)]));
767 ST(&(x[WS(rs, 1)]), VADD(TW, TZ), ms, &(x[WS(rs, 1)]));
768 Tt = BYTWJ(&(W[TWVL * 6]), VSUB(Tp, Ts));
769 ST(&(x[WS(vs, 4)]), Tt, ms, &(x[WS(vs, 4)]));
770 T4c = BYTWJ(&(W[TWVL * 6]), VSUB(T48, T4b));
771 ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
772 T2B = BYTWJ(&(W[TWVL * 6]), VSUB(T2x, T2A));
773 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 4)]));
774 T24 = BYTWJ(&(W[TWVL * 6]), VSUB(T20, T23));
775 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
776 }
777 {
778 V T10, T1x, T3F, T38, T1A, Tw;
779 T10 = BYTWJ(&(W[TWVL * 6]), VSUB(TW, TZ));
780 ST(&(x[WS(vs, 4) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
781 T1x = BYTWJ(&(W[TWVL * 6]), VSUB(T1t, T1w));
782 ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 4)]));
783 T3F = BYTWJ(&(W[TWVL * 6]), VSUB(T3B, T3E));
784 ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 4)]));
785 T38 = BYTWJ(&(W[TWVL * 6]), VSUB(T34, T37));
786 ST(&(x[WS(vs, 4) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
787 T1A = BYTWJ(&(W[TWVL * 10]), VSUB(T1y, T1z));
788 ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1A, ms, &(x[WS(vs, 6)]));
789 Tw = BYTWJ(&(W[TWVL * 10]), VSUB(Tu, Tv));
790 ST(&(x[WS(vs, 6)]), Tw, ms, &(x[WS(vs, 6)]));
791 }
792 {
793 V T2E, T3I, T13, T27, T3b, T4f;
794 T2E = BYTWJ(&(W[TWVL * 10]), VSUB(T2C, T2D));
795 ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2E, ms, &(x[WS(vs, 6)]));
796 T3I = BYTWJ(&(W[TWVL * 10]), VSUB(T3G, T3H));
797 ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3I, ms, &(x[WS(vs, 6)]));
798 T13 = BYTWJ(&(W[TWVL * 10]), VSUB(T11, T12));
799 ST(&(x[WS(vs, 6) + WS(rs, 1)]), T13, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
800 T27 = BYTWJ(&(W[TWVL * 10]), VSUB(T25, T26));
801 ST(&(x[WS(vs, 6) + WS(rs, 3)]), T27, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
802 T3b = BYTWJ(&(W[TWVL * 10]), VSUB(T39, T3a));
803 ST(&(x[WS(vs, 6) + WS(rs, 5)]), T3b, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
804 T4f = BYTWJ(&(W[TWVL * 10]), VSUB(T4d, T4e));
805 ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4f, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
806 }
807 {
808 V Tx, T1B, T3c, T4g, T3J, T2F;
809 Tx = BYTWJ(&(W[TWVL * 2]), VADD(Tu, Tv));
810 ST(&(x[WS(vs, 2)]), Tx, ms, &(x[WS(vs, 2)]));
811 T1B = BYTWJ(&(W[TWVL * 2]), VADD(T1y, T1z));
812 ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 2)]));
813 T3c = BYTWJ(&(W[TWVL * 2]), VADD(T39, T3a));
814 ST(&(x[WS(vs, 2) + WS(rs, 5)]), T3c, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
815 T4g = BYTWJ(&(W[TWVL * 2]), VADD(T4d, T4e));
816 ST(&(x[WS(vs, 2) + WS(rs, 7)]), T4g, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
817 T3J = BYTWJ(&(W[TWVL * 2]), VADD(T3G, T3H));
818 ST(&(x[WS(vs, 2) + WS(rs, 6)]), T3J, ms, &(x[WS(vs, 2)]));
819 T2F = BYTWJ(&(W[TWVL * 2]), VADD(T2C, T2D));
820 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T2F, ms, &(x[WS(vs, 2)]));
821 }
822 T28 = BYTWJ(&(W[TWVL * 2]), VADD(T25, T26));
823 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T28, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
824 T14 = BYTWJ(&(W[TWVL * 2]), VADD(T11, T12));
825 ST(&(x[WS(vs, 2) + WS(rs, 1)]), T14, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
826 {
827 V Th, Ti, Tb, Tg;
828 Tb = VADD(T3, Ta);
829 Tg = VBYI(VSUB(Tc, Tf));
830 Th = BYTWJ(&(W[TWVL * 12]), VSUB(Tb, Tg));
831 Ti = BYTWJ(&(W[0]), VADD(Tb, Tg));
832 ST(&(x[WS(vs, 7)]), Th, ms, &(x[WS(vs, 7)]));
833 ST(&(x[WS(vs, 1)]), Ti, ms, &(x[WS(vs, 1)]));
834 }
835 {
836 V T40, T41, T3U, T3Z;
837 T3U = VADD(T3M, T3T);
838 T3Z = VBYI(VSUB(T3V, T3Y));
839 T40 = BYTWJ(&(W[TWVL * 12]), VSUB(T3U, T3Z));
840 T41 = BYTWJ(&(W[0]), VADD(T3U, T3Z));
841 ST(&(x[WS(vs, 7) + WS(rs, 7)]), T40, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
842 ST(&(x[WS(vs, 1) + WS(rs, 7)]), T41, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
843 }
844 {
845 V T2p, T2q, T2j, T2o;
846 T2j = VADD(T2b, T2i);
847 T2o = VBYI(VSUB(T2k, T2n));
848 T2p = BYTWJ(&(W[TWVL * 12]), VSUB(T2j, T2o));
849 T2q = BYTWJ(&(W[0]), VADD(T2j, T2o));
850 ST(&(x[WS(vs, 7) + WS(rs, 4)]), T2p, ms, &(x[WS(vs, 7)]));
851 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T2q, ms, &(x[WS(vs, 1)]));
852 }
853 {
854 V T1S, T1T, T1M, T1R;
855 T1M = VADD(T1E, T1L);
856 T1R = VBYI(VSUB(T1N, T1Q));
857 T1S = BYTWJ(&(W[TWVL * 12]), VSUB(T1M, T1R));
858 T1T = BYTWJ(&(W[0]), VADD(T1M, T1R));
859 ST(&(x[WS(vs, 7) + WS(rs, 3)]), T1S, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
860 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1T, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
861 }
862 {
863 V TO, TP, TI, TN;
864 TI = VADD(TA, TH);
865 TN = VBYI(VSUB(TJ, TM));
866 TO = BYTWJ(&(W[TWVL * 12]), VSUB(TI, TN));
867 TP = BYTWJ(&(W[0]), VADD(TI, TN));
868 ST(&(x[WS(vs, 7) + WS(rs, 1)]), TO, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
869 ST(&(x[WS(vs, 1) + WS(rs, 1)]), TP, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
870 }
871 {
872 V T1l, T1m, T1f, T1k;
873 T1f = VADD(T17, T1e);
874 T1k = VBYI(VSUB(T1g, T1j));
875 T1l = BYTWJ(&(W[TWVL * 12]), VSUB(T1f, T1k));
876 T1m = BYTWJ(&(W[0]), VADD(T1f, T1k));
877 ST(&(x[WS(vs, 7) + WS(rs, 2)]), T1l, ms, &(x[WS(vs, 7)]));
878 ST(&(x[WS(vs, 1) + WS(rs, 2)]), T1m, ms, &(x[WS(vs, 1)]));
879 }
880 {
881 V T3t, T3u, T3n, T3s;
882 T3n = VADD(T3f, T3m);
883 T3s = VBYI(VSUB(T3o, T3r));
884 T3t = BYTWJ(&(W[TWVL * 12]), VSUB(T3n, T3s));
885 T3u = BYTWJ(&(W[0]), VADD(T3n, T3s));
886 ST(&(x[WS(vs, 7) + WS(rs, 6)]), T3t, ms, &(x[WS(vs, 7)]));
887 ST(&(x[WS(vs, 1) + WS(rs, 6)]), T3u, ms, &(x[WS(vs, 1)]));
888 }
889 {
890 V T2W, T2X, T2Q, T2V;
891 T2Q = VADD(T2I, T2P);
892 T2V = VBYI(VSUB(T2R, T2U));
893 T2W = BYTWJ(&(W[TWVL * 12]), VSUB(T2Q, T2V));
894 T2X = BYTWJ(&(W[0]), VADD(T2Q, T2V));
895 ST(&(x[WS(vs, 7) + WS(rs, 5)]), T2W, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
896 ST(&(x[WS(vs, 1) + WS(rs, 5)]), T2X, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
897 }
898 {
899 V T1p, T1q, T1n, T1o;
900 T1n = VSUB(T17, T1e);
901 T1o = VBYI(VADD(T1j, T1g));
902 T1p = BYTWJ(&(W[TWVL * 8]), VSUB(T1n, T1o));
903 T1q = BYTWJ(&(W[TWVL * 4]), VADD(T1n, T1o));
904 ST(&(x[WS(vs, 5) + WS(rs, 2)]), T1p, ms, &(x[WS(vs, 5)]));
905 ST(&(x[WS(vs, 3) + WS(rs, 2)]), T1q, ms, &(x[WS(vs, 3)]));
906 }
907 {
908 V Tl, Tm, Tj, Tk;
909 Tj = VSUB(T3, Ta);
910 Tk = VBYI(VADD(Tf, Tc));
911 Tl = BYTWJ(&(W[TWVL * 8]), VSUB(Tj, Tk));
912 Tm = BYTWJ(&(W[TWVL * 4]), VADD(Tj, Tk));
913 ST(&(x[WS(vs, 5)]), Tl, ms, &(x[WS(vs, 5)]));
914 ST(&(x[WS(vs, 3)]), Tm, ms, &(x[WS(vs, 3)]));
915 }
916 {
917 V T2t, T2u, T2r, T2s;
918 T2r = VSUB(T2b, T2i);
919 T2s = VBYI(VADD(T2n, T2k));
920 T2t = BYTWJ(&(W[TWVL * 8]), VSUB(T2r, T2s));
921 T2u = BYTWJ(&(W[TWVL * 4]), VADD(T2r, T2s));
922 ST(&(x[WS(vs, 5) + WS(rs, 4)]), T2t, ms, &(x[WS(vs, 5)]));
923 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T2u, ms, &(x[WS(vs, 3)]));
924 }
925 {
926 V T3x, T3y, T3v, T3w;
927 T3v = VSUB(T3f, T3m);
928 T3w = VBYI(VADD(T3r, T3o));
929 T3x = BYTWJ(&(W[TWVL * 8]), VSUB(T3v, T3w));
930 T3y = BYTWJ(&(W[TWVL * 4]), VADD(T3v, T3w));
931 ST(&(x[WS(vs, 5) + WS(rs, 6)]), T3x, ms, &(x[WS(vs, 5)]));
932 ST(&(x[WS(vs, 3) + WS(rs, 6)]), T3y, ms, &(x[WS(vs, 3)]));
933 }
934 {
935 V TS, TT, TQ, TR;
936 TQ = VSUB(TA, TH);
937 TR = VBYI(VADD(TM, TJ));
938 TS = BYTWJ(&(W[TWVL * 8]), VSUB(TQ, TR));
939 TT = BYTWJ(&(W[TWVL * 4]), VADD(TQ, TR));
940 ST(&(x[WS(vs, 5) + WS(rs, 1)]), TS, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
941 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TT, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
942 }
943 {
944 V T1W, T1X, T1U, T1V;
945 T1U = VSUB(T1E, T1L);
946 T1V = VBYI(VADD(T1Q, T1N));
947 T1W = BYTWJ(&(W[TWVL * 8]), VSUB(T1U, T1V));
948 T1X = BYTWJ(&(W[TWVL * 4]), VADD(T1U, T1V));
949 ST(&(x[WS(vs, 5) + WS(rs, 3)]), T1W, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
950 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1X, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
951 }
952 {
953 V T30, T31, T2Y, T2Z;
954 T2Y = VSUB(T2I, T2P);
955 T2Z = VBYI(VADD(T2U, T2R));
956 T30 = BYTWJ(&(W[TWVL * 8]), VSUB(T2Y, T2Z));
957 T31 = BYTWJ(&(W[TWVL * 4]), VADD(T2Y, T2Z));
958 ST(&(x[WS(vs, 5) + WS(rs, 5)]), T30, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
959 ST(&(x[WS(vs, 3) + WS(rs, 5)]), T31, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
960 }
961 {
962 V T44, T45, T42, T43;
963 T42 = VSUB(T3M, T3T);
964 T43 = VBYI(VADD(T3Y, T3V));
965 T44 = BYTWJ(&(W[TWVL * 8]), VSUB(T42, T43));
966 T45 = BYTWJ(&(W[TWVL * 4]), VADD(T42, T43));
967 ST(&(x[WS(vs, 5) + WS(rs, 7)]), T44, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
968 ST(&(x[WS(vs, 3) + WS(rs, 7)]), T45, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
969 }
970 }
971 }
972 VLEAVE();
973 }
974
975 static const tw_instr twinstr[] = {
976 VTW(0, 1),
977 VTW(0, 2),
978 VTW(0, 3),
979 VTW(0, 4),
980 VTW(0, 5),
981 VTW(0, 6),
982 VTW(0, 7),
983 {TW_NEXT, VL, 0}
984 };
985
986 static const ct_desc desc = { 8, XSIMD_STRING("q1fv_8"), twinstr, &GENUS, {264, 128, 0, 0}, 0, 0, 0 };
987
988 void XSIMD(codelet_q1fv_8) (planner *p) {
989 X(kdft_difsq_register) (p, q1fv_8, &desc);
990 }
991 #endif /* HAVE_FMA */