Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/q1bv_8.c @ 95:89f5e221ed7b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam <cannam@all-day-breakfast.com>
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
94:d278df1123f9 95:89f5e221ed7b
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 07:39:33 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 q1bv_8 -include 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 * 121 stack variables, 1 constants, and 128 memory accesses
34 */
35 #include "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 T42, T43, T1U, T1V, T2Y, T2Z, TT, TS, T45, T44;
46 {
47 V T3, Te, T1E, T1P, Tv, Tp, T26, T20, T2b, T2m, T3M, T2x, T2D, T3X, TA;
48 V TL, T48, T4e, T17, T12, TW, T1i, T2I, T1z, T1t, T2T, T3f, T3q, T34, T3a;
49 V T3H, T3B, Ts, Tw, Tf, Ta, T23, T27, T1Q, T1L, T2A, T2E, T2n, T2i, T4b;
50 V T4f, T3Y, T3T, TZ, T13, TM, TH, T35, T2L, T3j, T1w, T1A, 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 Tv = VADD(Tn, To);
81 Tp = VSUB(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 T26 = VADD(T1Y, T1Z);
85 T20 = VSUB(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 = VSUB(T2v, T2w);
104 T2D = VADD(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 = VSUB(T46, T47);
116 T4e = VADD(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 T12 = VADD(TU, TV);
126 TW = VSUB(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 T1z = VADD(T1r, T1s);
138 T1t = VSUB(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 = VSUB(T32, T33);
158 T3a = VADD(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 T3H = VADD(T3z, T3A);
163 T3B = VSUB(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 = VSUB(Tq, Tr);
179 Tw = VADD(Tq, Tr);
180 Tf = VSUB(T6, T9);
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 = VSUB(T21, T22);
191 T27 = VADD(T21, T22);
192 T1Q = VSUB(T1H, T1K);
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 = VSUB(T2y, T2z);
209 T2E = VADD(T2y, T2z);
210 T2n = VSUB(T2e, T2h);
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 = VSUB(T49, T4a);
221 T4f = VADD(T49, T4a);
222 T3Y = VSUB(T3P, T3S);
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 = VSUB(TX, TY);
236 T13 = VADD(TX, TY);
237 TM = VSUB(TD, TG);
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 = VSUB(T1u, T1v);
248 T1A = VADD(T1u, T1v);
249 T1j = VSUB(T1a, T1d);
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 T3b, T2U, T2P, T3I, T3r, T3m, T11, T25, T39, T4d;
262 {
263 V T37, T3E, T2B, T24;
264 {
265 V T3D, T3l, Tt, T4c;
266 ST(&(x[0]), VADD(Tv, Tw), ms, &(x[0]));
267 ST(&(x[WS(rs, 2)]), VADD(T1z, T1A), ms, &(x[0]));
268 ST(&(x[WS(rs, 7)]), VADD(T4e, T4f), ms, &(x[WS(rs, 1)]));
269 T37 = VSUB(T35, T36);
270 T3b = VADD(T35, T36);
271 T2U = VSUB(T2L, T2O);
272 T2P = VADD(T2L, T2O);
273 T3D = VADD(T3j, T3k);
274 T3l = VSUB(T3j, T3k);
275 ST(&(x[WS(rs, 4)]), VADD(T2D, T2E), ms, &(x[0]));
276 ST(&(x[WS(rs, 3)]), VADD(T26, T27), ms, &(x[WS(rs, 1)]));
277 ST(&(x[WS(rs, 1)]), VADD(T12, T13), ms, &(x[WS(rs, 1)]));
278 ST(&(x[WS(rs, 5)]), VADD(T3a, T3b), ms, &(x[WS(rs, 1)]));
279 Tt = BYTW(&(W[TWVL * 10]), VFNMSI(Ts, Tp));
280 T4c = BYTW(&(W[TWVL * 10]), VFNMSI(T4b, T48));
281 T3E = VSUB(T3C, T3D);
282 T3I = VADD(T3C, T3D);
283 T3r = VSUB(T3i, T3l);
284 T3m = VADD(T3i, T3l);
285 T2B = BYTW(&(W[TWVL * 10]), VFNMSI(T2A, T2x));
286 T24 = BYTW(&(W[TWVL * 10]), VFNMSI(T23, T20));
287 ST(&(x[WS(vs, 6)]), Tt, ms, &(x[WS(vs, 6)]));
288 ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
289 }
290 {
291 V T38, T1y, Tu, T10, T1x, T3F, T2C, T3G;
292 T10 = BYTW(&(W[TWVL * 10]), VFNMSI(TZ, TW));
293 ST(&(x[WS(rs, 6)]), VADD(T3H, T3I), ms, &(x[0]));
294 T1x = BYTW(&(W[TWVL * 10]), VFNMSI(T1w, T1t));
295 T3F = BYTW(&(W[TWVL * 10]), VFNMSI(T3E, T3B));
296 ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 6)]));
297 ST(&(x[WS(vs, 6) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
298 T38 = BYTW(&(W[TWVL * 10]), VFNMSI(T37, T34));
299 T1y = BYTW(&(W[TWVL * 2]), VFMAI(T1w, T1t));
300 ST(&(x[WS(vs, 6) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
301 Tu = BYTW(&(W[TWVL * 2]), VFMAI(Ts, Tp));
302 ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 6)]));
303 ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 6)]));
304 T2C = BYTW(&(W[TWVL * 2]), VFMAI(T2A, T2x));
305 T3G = BYTW(&(W[TWVL * 2]), VFMAI(T3E, T3B));
306 ST(&(x[WS(vs, 6) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
307 ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1y, ms, &(x[WS(vs, 2)]));
308 T11 = BYTW(&(W[TWVL * 2]), VFMAI(TZ, TW));
309 ST(&(x[WS(vs, 2)]), Tu, ms, &(x[WS(vs, 2)]));
310 T25 = BYTW(&(W[TWVL * 2]), VFMAI(T23, T20));
311 T39 = BYTW(&(W[TWVL * 2]), VFMAI(T37, T34));
312 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T2C, ms, &(x[WS(vs, 2)]));
313 ST(&(x[WS(vs, 2) + WS(rs, 6)]), T3G, ms, &(x[WS(vs, 2)]));
314 T4d = BYTW(&(W[TWVL * 2]), VFMAI(T4b, T48));
315 }
316 }
317 {
318 V Tj, Tk, T2r, T2j, T2o, T2s, Ti, Th, T1M, T1R, T41, T40;
319 {
320 V T3c, T4g, T3J, T2F, Tx, T1B;
321 Tx = BYTW(&(W[TWVL * 6]), VSUB(Tv, Tw));
322 ST(&(x[WS(vs, 2) + WS(rs, 1)]), T11, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
323 T1B = BYTW(&(W[TWVL * 6]), VSUB(T1z, T1A));
324 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T25, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
325 ST(&(x[WS(vs, 2) + WS(rs, 5)]), T39, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
326 T3c = BYTW(&(W[TWVL * 6]), VSUB(T3a, T3b));
327 T4g = BYTW(&(W[TWVL * 6]), VSUB(T4e, T4f));
328 ST(&(x[WS(vs, 2) + WS(rs, 7)]), T4d, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
329 ST(&(x[WS(vs, 4)]), Tx, ms, &(x[WS(vs, 4)]));
330 T3J = BYTW(&(W[TWVL * 6]), VSUB(T3H, T3I));
331 ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 4)]));
332 T2F = BYTW(&(W[TWVL * 6]), VSUB(T2D, T2E));
333 {
334 V T14, Tb, Tg, T28, T3U, T3Z;
335 T28 = BYTW(&(W[TWVL * 6]), VSUB(T26, T27));
336 ST(&(x[WS(vs, 4) + WS(rs, 5)]), T3c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
337 ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4g, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
338 T14 = BYTW(&(W[TWVL * 6]), VSUB(T12, T13));
339 Tj = VFMA(LDK(KP707106781), Ta, T3);
340 Tb = VFNMS(LDK(KP707106781), Ta, T3);
341 ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3J, ms, &(x[WS(vs, 4)]));
342 Tk = VFMA(LDK(KP707106781), Tf, Te);
343 Tg = VFNMS(LDK(KP707106781), Tf, Te);
344 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2F, ms, &(x[WS(vs, 4)]));
345 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T28, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
346 T3U = VFNMS(LDK(KP707106781), T3T, T3M);
347 T42 = VFMA(LDK(KP707106781), T3T, T3M);
348 T43 = VFMA(LDK(KP707106781), T3Y, T3X);
349 T3Z = VFNMS(LDK(KP707106781), T3Y, T3X);
350 ST(&(x[WS(vs, 4) + WS(rs, 1)]), T14, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
351 T2r = VFMA(LDK(KP707106781), T2i, T2b);
352 T2j = VFNMS(LDK(KP707106781), T2i, T2b);
353 T2o = VFNMS(LDK(KP707106781), T2n, T2m);
354 T2s = VFMA(LDK(KP707106781), T2n, T2m);
355 Ti = BYTW(&(W[TWVL * 8]), VFMAI(Tg, Tb));
356 Th = BYTW(&(W[TWVL * 4]), VFNMSI(Tg, Tb));
357 T1U = VFMA(LDK(KP707106781), T1L, T1E);
358 T1M = VFNMS(LDK(KP707106781), T1L, T1E);
359 T1R = VFNMS(LDK(KP707106781), T1Q, T1P);
360 T1V = VFMA(LDK(KP707106781), T1Q, T1P);
361 T41 = BYTW(&(W[TWVL * 8]), VFMAI(T3Z, T3U));
362 T40 = BYTW(&(W[TWVL * 4]), VFNMSI(T3Z, T3U));
363 }
364 }
365 {
366 V TQ, TR, T1n, T1o, T3v, T3w;
367 {
368 V TI, TN, T1f, T1k, T3n, T3s;
369 {
370 V T1T, T1S, T2q, T2p;
371 TQ = VFMA(LDK(KP707106781), TH, TA);
372 TI = VFNMS(LDK(KP707106781), TH, TA);
373 T2q = BYTW(&(W[TWVL * 8]), VFMAI(T2o, T2j));
374 T2p = BYTW(&(W[TWVL * 4]), VFNMSI(T2o, T2j));
375 ST(&(x[WS(vs, 5)]), Ti, ms, &(x[WS(vs, 5)]));
376 ST(&(x[WS(vs, 3)]), Th, ms, &(x[WS(vs, 3)]));
377 T1T = BYTW(&(W[TWVL * 8]), VFMAI(T1R, T1M));
378 T1S = BYTW(&(W[TWVL * 4]), VFNMSI(T1R, T1M));
379 ST(&(x[WS(vs, 5) + WS(rs, 7)]), T41, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
380 ST(&(x[WS(vs, 3) + WS(rs, 7)]), T40, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
381 ST(&(x[WS(vs, 5) + WS(rs, 4)]), T2q, ms, &(x[WS(vs, 5)]));
382 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T2p, ms, &(x[WS(vs, 3)]));
383 TN = VFNMS(LDK(KP707106781), TM, TL);
384 TR = VFMA(LDK(KP707106781), TM, TL);
385 T1n = VFMA(LDK(KP707106781), T1e, T17);
386 T1f = VFNMS(LDK(KP707106781), T1e, T17);
387 ST(&(x[WS(vs, 5) + WS(rs, 3)]), T1T, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
388 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1S, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
389 T1k = VFNMS(LDK(KP707106781), T1j, T1i);
390 T1o = VFMA(LDK(KP707106781), T1j, T1i);
391 T3v = VFMA(LDK(KP707106781), T3m, T3f);
392 T3n = VFNMS(LDK(KP707106781), T3m, T3f);
393 T3s = VFNMS(LDK(KP707106781), T3r, T3q);
394 T3w = VFMA(LDK(KP707106781), T3r, T3q);
395 }
396 {
397 V T2Q, TP, TO, T2V, T2X, T2W;
398 T2Y = VFMA(LDK(KP707106781), T2P, T2I);
399 T2Q = VFNMS(LDK(KP707106781), T2P, T2I);
400 TP = BYTW(&(W[TWVL * 8]), VFMAI(TN, TI));
401 TO = BYTW(&(W[TWVL * 4]), VFNMSI(TN, TI));
402 T2V = VFNMS(LDK(KP707106781), T2U, T2T);
403 T2Z = VFMA(LDK(KP707106781), T2U, T2T);
404 {
405 V T1m, T1l, T3u, T3t;
406 T1m = BYTW(&(W[TWVL * 8]), VFMAI(T1k, T1f));
407 T1l = BYTW(&(W[TWVL * 4]), VFNMSI(T1k, T1f));
408 T3u = BYTW(&(W[TWVL * 8]), VFMAI(T3s, T3n));
409 T3t = BYTW(&(W[TWVL * 4]), VFNMSI(T3s, T3n));
410 ST(&(x[WS(vs, 5) + WS(rs, 1)]), TP, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
411 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TO, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
412 T2X = BYTW(&(W[TWVL * 8]), VFMAI(T2V, T2Q));
413 T2W = BYTW(&(W[TWVL * 4]), VFNMSI(T2V, T2Q));
414 ST(&(x[WS(vs, 5) + WS(rs, 2)]), T1m, ms, &(x[WS(vs, 5)]));
415 ST(&(x[WS(vs, 3) + WS(rs, 2)]), T1l, ms, &(x[WS(vs, 3)]));
416 ST(&(x[WS(vs, 5) + WS(rs, 6)]), T3u, ms, &(x[WS(vs, 5)]));
417 ST(&(x[WS(vs, 3) + WS(rs, 6)]), T3t, ms, &(x[WS(vs, 3)]));
418 }
419 ST(&(x[WS(vs, 5) + WS(rs, 5)]), T2X, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
420 ST(&(x[WS(vs, 3) + WS(rs, 5)]), T2W, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
421 }
422 }
423 {
424 V T3y, T3x, T1q, T1p;
425 T1q = BYTW(&(W[TWVL * 12]), VFNMSI(T1o, T1n));
426 T1p = BYTW(&(W[0]), VFMAI(T1o, T1n));
427 {
428 V Tm, Tl, T2u, T2t;
429 Tm = BYTW(&(W[TWVL * 12]), VFNMSI(Tk, Tj));
430 Tl = BYTW(&(W[0]), VFMAI(Tk, Tj));
431 T2u = BYTW(&(W[TWVL * 12]), VFNMSI(T2s, T2r));
432 T2t = BYTW(&(W[0]), VFMAI(T2s, T2r));
433 ST(&(x[WS(vs, 7) + WS(rs, 2)]), T1q, ms, &(x[WS(vs, 7)]));
434 ST(&(x[WS(vs, 1) + WS(rs, 2)]), T1p, ms, &(x[WS(vs, 1)]));
435 T3y = BYTW(&(W[TWVL * 12]), VFNMSI(T3w, T3v));
436 T3x = BYTW(&(W[0]), VFMAI(T3w, T3v));
437 ST(&(x[WS(vs, 7)]), Tm, ms, &(x[WS(vs, 7)]));
438 ST(&(x[WS(vs, 1)]), Tl, ms, &(x[WS(vs, 1)]));
439 ST(&(x[WS(vs, 7) + WS(rs, 4)]), T2u, ms, &(x[WS(vs, 7)]));
440 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T2t, ms, &(x[WS(vs, 1)]));
441 }
442 ST(&(x[WS(vs, 7) + WS(rs, 6)]), T3y, ms, &(x[WS(vs, 7)]));
443 ST(&(x[WS(vs, 1) + WS(rs, 6)]), T3x, ms, &(x[WS(vs, 1)]));
444 TT = BYTW(&(W[TWVL * 12]), VFNMSI(TR, TQ));
445 TS = BYTW(&(W[0]), VFMAI(TR, TQ));
446 }
447 }
448 }
449 }
450 }
451 {
452 V T1X, T1W, T31, T30;
453 T1X = BYTW(&(W[TWVL * 12]), VFNMSI(T1V, T1U));
454 T1W = BYTW(&(W[0]), VFMAI(T1V, T1U));
455 T31 = BYTW(&(W[TWVL * 12]), VFNMSI(T2Z, T2Y));
456 T30 = BYTW(&(W[0]), VFMAI(T2Z, T2Y));
457 ST(&(x[WS(vs, 7) + WS(rs, 1)]), TT, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
458 ST(&(x[WS(vs, 1) + WS(rs, 1)]), TS, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
459 T45 = BYTW(&(W[TWVL * 12]), VFNMSI(T43, T42));
460 T44 = BYTW(&(W[0]), VFMAI(T43, T42));
461 ST(&(x[WS(vs, 7) + WS(rs, 3)]), T1X, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
462 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1W, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
463 ST(&(x[WS(vs, 7) + WS(rs, 5)]), T31, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
464 ST(&(x[WS(vs, 1) + WS(rs, 5)]), T30, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
465 }
466 ST(&(x[WS(vs, 7) + WS(rs, 7)]), T45, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
467 ST(&(x[WS(vs, 1) + WS(rs, 7)]), T44, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
468 }
469 }
470 VLEAVE();
471 }
472
473 static const tw_instr twinstr[] = {
474 VTW(0, 1),
475 VTW(0, 2),
476 VTW(0, 3),
477 VTW(0, 4),
478 VTW(0, 5),
479 VTW(0, 6),
480 VTW(0, 7),
481 {TW_NEXT, VL, 0}
482 };
483
484 static const ct_desc desc = { 8, XSIMD_STRING("q1bv_8"), twinstr, &GENUS, {184, 112, 80, 0}, 0, 0, 0 };
485
486 void XSIMD(codelet_q1bv_8) (planner *p) {
487 X(kdft_difsq_register) (p, q1bv_8, &desc);
488 }
489 #else /* HAVE_FMA */
490
491 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -dif -name q1bv_8 -include q1b.h -sign 1 */
492
493 /*
494 * This function contains 264 FP additions, 128 FP multiplications,
495 * (or, 264 additions, 128 multiplications, 0 fused multiply/add),
496 * 77 stack variables, 1 constants, and 128 memory accesses
497 */
498 #include "q1b.h"
499
500 static void q1bv_8(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
501 {
502 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
503 {
504 INT m;
505 R *x;
506 x = ii;
507 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)) {
508 V Ta, Tv, Te, Tp, T1L, T26, T1P, T20, T2i, T2D, T2m, T2x, T3T, T4e, T3X;
509 V T48, TH, T12, TL, TW, T1e, T1z, T1i, T1t, T2P, T3a, T2T, T34, T3m, T3H;
510 V T3q, T3B, T7, Tw, Tf, Ts, T1I, T27, T1Q, T23, T2f, T2E, T2n, T2A, T3Q;
511 V T4f, T3Y, T4b, TE, T13, TM, TZ, T1b, T1A, T1j, T1w, T2M, T3b, T2U, T37;
512 V T3j, T3I, T3r, T3E, T28, T14;
513 {
514 V T8, T9, To, Tc, Td, Tn;
515 T8 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
516 T9 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
517 To = VADD(T8, T9);
518 Tc = LD(&(x[0]), ms, &(x[0]));
519 Td = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
520 Tn = VADD(Tc, Td);
521 Ta = VSUB(T8, T9);
522 Tv = VADD(Tn, To);
523 Te = VSUB(Tc, Td);
524 Tp = VSUB(Tn, To);
525 }
526 {
527 V T1J, T1K, T1Z, T1N, T1O, T1Y;
528 T1J = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
529 T1K = LD(&(x[WS(vs, 3) + WS(rs, 6)]), ms, &(x[WS(vs, 3)]));
530 T1Z = VADD(T1J, T1K);
531 T1N = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
532 T1O = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
533 T1Y = VADD(T1N, T1O);
534 T1L = VSUB(T1J, T1K);
535 T26 = VADD(T1Y, T1Z);
536 T1P = VSUB(T1N, T1O);
537 T20 = VSUB(T1Y, T1Z);
538 }
539 {
540 V T2g, T2h, T2w, T2k, T2l, T2v;
541 T2g = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
542 T2h = LD(&(x[WS(vs, 4) + WS(rs, 6)]), ms, &(x[WS(vs, 4)]));
543 T2w = VADD(T2g, T2h);
544 T2k = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
545 T2l = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
546 T2v = VADD(T2k, T2l);
547 T2i = VSUB(T2g, T2h);
548 T2D = VADD(T2v, T2w);
549 T2m = VSUB(T2k, T2l);
550 T2x = VSUB(T2v, T2w);
551 }
552 {
553 V T3R, T3S, T47, T3V, T3W, T46;
554 T3R = LD(&(x[WS(vs, 7) + WS(rs, 2)]), ms, &(x[WS(vs, 7)]));
555 T3S = LD(&(x[WS(vs, 7) + WS(rs, 6)]), ms, &(x[WS(vs, 7)]));
556 T47 = VADD(T3R, T3S);
557 T3V = LD(&(x[WS(vs, 7)]), ms, &(x[WS(vs, 7)]));
558 T3W = LD(&(x[WS(vs, 7) + WS(rs, 4)]), ms, &(x[WS(vs, 7)]));
559 T46 = VADD(T3V, T3W);
560 T3T = VSUB(T3R, T3S);
561 T4e = VADD(T46, T47);
562 T3X = VSUB(T3V, T3W);
563 T48 = VSUB(T46, T47);
564 }
565 {
566 V TF, TG, TV, TJ, TK, TU;
567 TF = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
568 TG = LD(&(x[WS(vs, 1) + WS(rs, 6)]), ms, &(x[WS(vs, 1)]));
569 TV = VADD(TF, TG);
570 TJ = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
571 TK = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
572 TU = VADD(TJ, TK);
573 TH = VSUB(TF, TG);
574 T12 = VADD(TU, TV);
575 TL = VSUB(TJ, TK);
576 TW = VSUB(TU, TV);
577 }
578 {
579 V T1c, T1d, T1s, T1g, T1h, T1r;
580 T1c = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
581 T1d = LD(&(x[WS(vs, 2) + WS(rs, 6)]), ms, &(x[WS(vs, 2)]));
582 T1s = VADD(T1c, T1d);
583 T1g = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
584 T1h = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
585 T1r = VADD(T1g, T1h);
586 T1e = VSUB(T1c, T1d);
587 T1z = VADD(T1r, T1s);
588 T1i = VSUB(T1g, T1h);
589 T1t = VSUB(T1r, T1s);
590 }
591 {
592 V T2N, T2O, T33, T2R, T2S, T32;
593 T2N = LD(&(x[WS(vs, 5) + WS(rs, 2)]), ms, &(x[WS(vs, 5)]));
594 T2O = LD(&(x[WS(vs, 5) + WS(rs, 6)]), ms, &(x[WS(vs, 5)]));
595 T33 = VADD(T2N, T2O);
596 T2R = LD(&(x[WS(vs, 5)]), ms, &(x[WS(vs, 5)]));
597 T2S = LD(&(x[WS(vs, 5) + WS(rs, 4)]), ms, &(x[WS(vs, 5)]));
598 T32 = VADD(T2R, T2S);
599 T2P = VSUB(T2N, T2O);
600 T3a = VADD(T32, T33);
601 T2T = VSUB(T2R, T2S);
602 T34 = VSUB(T32, T33);
603 }
604 {
605 V T3k, T3l, T3A, T3o, T3p, T3z;
606 T3k = LD(&(x[WS(vs, 6) + WS(rs, 2)]), ms, &(x[WS(vs, 6)]));
607 T3l = LD(&(x[WS(vs, 6) + WS(rs, 6)]), ms, &(x[WS(vs, 6)]));
608 T3A = VADD(T3k, T3l);
609 T3o = LD(&(x[WS(vs, 6)]), ms, &(x[WS(vs, 6)]));
610 T3p = LD(&(x[WS(vs, 6) + WS(rs, 4)]), ms, &(x[WS(vs, 6)]));
611 T3z = VADD(T3o, T3p);
612 T3m = VSUB(T3k, T3l);
613 T3H = VADD(T3z, T3A);
614 T3q = VSUB(T3o, T3p);
615 T3B = VSUB(T3z, T3A);
616 }
617 {
618 V T3, Tq, T6, Tr;
619 {
620 V T1, T2, T4, T5;
621 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
622 T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
623 T3 = VSUB(T1, T2);
624 Tq = VADD(T1, T2);
625 T4 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
626 T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
627 T6 = VSUB(T4, T5);
628 Tr = VADD(T4, T5);
629 }
630 T7 = VMUL(LDK(KP707106781), VSUB(T3, T6));
631 Tw = VADD(Tq, Tr);
632 Tf = VMUL(LDK(KP707106781), VADD(T3, T6));
633 Ts = VBYI(VSUB(Tq, Tr));
634 }
635 {
636 V T1E, T21, T1H, T22;
637 {
638 V T1C, T1D, T1F, T1G;
639 T1C = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
640 T1D = LD(&(x[WS(vs, 3) + WS(rs, 5)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
641 T1E = VSUB(T1C, T1D);
642 T21 = VADD(T1C, T1D);
643 T1F = LD(&(x[WS(vs, 3) + WS(rs, 7)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
644 T1G = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
645 T1H = VSUB(T1F, T1G);
646 T22 = VADD(T1F, T1G);
647 }
648 T1I = VMUL(LDK(KP707106781), VSUB(T1E, T1H));
649 T27 = VADD(T21, T22);
650 T1Q = VMUL(LDK(KP707106781), VADD(T1E, T1H));
651 T23 = VBYI(VSUB(T21, T22));
652 }
653 {
654 V T2b, T2y, T2e, T2z;
655 {
656 V T29, T2a, T2c, T2d;
657 T29 = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
658 T2a = LD(&(x[WS(vs, 4) + WS(rs, 5)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
659 T2b = VSUB(T29, T2a);
660 T2y = VADD(T29, T2a);
661 T2c = LD(&(x[WS(vs, 4) + WS(rs, 7)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
662 T2d = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
663 T2e = VSUB(T2c, T2d);
664 T2z = VADD(T2c, T2d);
665 }
666 T2f = VMUL(LDK(KP707106781), VSUB(T2b, T2e));
667 T2E = VADD(T2y, T2z);
668 T2n = VMUL(LDK(KP707106781), VADD(T2b, T2e));
669 T2A = VBYI(VSUB(T2y, T2z));
670 }
671 {
672 V T3M, T49, T3P, T4a;
673 {
674 V T3K, T3L, T3N, T3O;
675 T3K = LD(&(x[WS(vs, 7) + WS(rs, 1)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
676 T3L = LD(&(x[WS(vs, 7) + WS(rs, 5)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
677 T3M = VSUB(T3K, T3L);
678 T49 = VADD(T3K, T3L);
679 T3N = LD(&(x[WS(vs, 7) + WS(rs, 7)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
680 T3O = LD(&(x[WS(vs, 7) + WS(rs, 3)]), ms, &(x[WS(vs, 7) + WS(rs, 1)]));
681 T3P = VSUB(T3N, T3O);
682 T4a = VADD(T3N, T3O);
683 }
684 T3Q = VMUL(LDK(KP707106781), VSUB(T3M, T3P));
685 T4f = VADD(T49, T4a);
686 T3Y = VMUL(LDK(KP707106781), VADD(T3M, T3P));
687 T4b = VBYI(VSUB(T49, T4a));
688 }
689 {
690 V TA, TX, TD, TY;
691 {
692 V Ty, Tz, TB, TC;
693 Ty = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
694 Tz = LD(&(x[WS(vs, 1) + WS(rs, 5)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
695 TA = VSUB(Ty, Tz);
696 TX = VADD(Ty, Tz);
697 TB = LD(&(x[WS(vs, 1) + WS(rs, 7)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
698 TC = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
699 TD = VSUB(TB, TC);
700 TY = VADD(TB, TC);
701 }
702 TE = VMUL(LDK(KP707106781), VSUB(TA, TD));
703 T13 = VADD(TX, TY);
704 TM = VMUL(LDK(KP707106781), VADD(TA, TD));
705 TZ = VBYI(VSUB(TX, TY));
706 }
707 {
708 V T17, T1u, T1a, T1v;
709 {
710 V T15, T16, T18, T19;
711 T15 = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
712 T16 = LD(&(x[WS(vs, 2) + WS(rs, 5)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
713 T17 = VSUB(T15, T16);
714 T1u = VADD(T15, T16);
715 T18 = LD(&(x[WS(vs, 2) + WS(rs, 7)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
716 T19 = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
717 T1a = VSUB(T18, T19);
718 T1v = VADD(T18, T19);
719 }
720 T1b = VMUL(LDK(KP707106781), VSUB(T17, T1a));
721 T1A = VADD(T1u, T1v);
722 T1j = VMUL(LDK(KP707106781), VADD(T17, T1a));
723 T1w = VBYI(VSUB(T1u, T1v));
724 }
725 {
726 V T2I, T35, T2L, T36;
727 {
728 V T2G, T2H, T2J, T2K;
729 T2G = LD(&(x[WS(vs, 5) + WS(rs, 1)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
730 T2H = LD(&(x[WS(vs, 5) + WS(rs, 5)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
731 T2I = VSUB(T2G, T2H);
732 T35 = VADD(T2G, T2H);
733 T2J = LD(&(x[WS(vs, 5) + WS(rs, 7)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
734 T2K = LD(&(x[WS(vs, 5) + WS(rs, 3)]), ms, &(x[WS(vs, 5) + WS(rs, 1)]));
735 T2L = VSUB(T2J, T2K);
736 T36 = VADD(T2J, T2K);
737 }
738 T2M = VMUL(LDK(KP707106781), VSUB(T2I, T2L));
739 T3b = VADD(T35, T36);
740 T2U = VMUL(LDK(KP707106781), VADD(T2I, T2L));
741 T37 = VBYI(VSUB(T35, T36));
742 }
743 {
744 V T3f, T3C, T3i, T3D;
745 {
746 V T3d, T3e, T3g, T3h;
747 T3d = LD(&(x[WS(vs, 6) + WS(rs, 1)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
748 T3e = LD(&(x[WS(vs, 6) + WS(rs, 5)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
749 T3f = VSUB(T3d, T3e);
750 T3C = VADD(T3d, T3e);
751 T3g = LD(&(x[WS(vs, 6) + WS(rs, 7)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
752 T3h = LD(&(x[WS(vs, 6) + WS(rs, 3)]), ms, &(x[WS(vs, 6) + WS(rs, 1)]));
753 T3i = VSUB(T3g, T3h);
754 T3D = VADD(T3g, T3h);
755 }
756 T3j = VMUL(LDK(KP707106781), VSUB(T3f, T3i));
757 T3I = VADD(T3C, T3D);
758 T3r = VMUL(LDK(KP707106781), VADD(T3f, T3i));
759 T3E = VBYI(VSUB(T3C, T3D));
760 }
761 ST(&(x[0]), VADD(Tv, Tw), ms, &(x[0]));
762 ST(&(x[WS(rs, 2)]), VADD(T1z, T1A), ms, &(x[0]));
763 ST(&(x[WS(rs, 5)]), VADD(T3a, T3b), ms, &(x[WS(rs, 1)]));
764 ST(&(x[WS(rs, 7)]), VADD(T4e, T4f), ms, &(x[WS(rs, 1)]));
765 ST(&(x[WS(rs, 6)]), VADD(T3H, T3I), ms, &(x[0]));
766 ST(&(x[WS(rs, 4)]), VADD(T2D, T2E), ms, &(x[0]));
767 {
768 V Tt, T4c, T2B, T24;
769 ST(&(x[WS(rs, 3)]), VADD(T26, T27), ms, &(x[WS(rs, 1)]));
770 ST(&(x[WS(rs, 1)]), VADD(T12, T13), ms, &(x[WS(rs, 1)]));
771 Tt = BYTW(&(W[TWVL * 10]), VSUB(Tp, Ts));
772 ST(&(x[WS(vs, 6)]), Tt, ms, &(x[WS(vs, 6)]));
773 T4c = BYTW(&(W[TWVL * 10]), VSUB(T48, T4b));
774 ST(&(x[WS(vs, 6) + WS(rs, 7)]), T4c, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
775 T2B = BYTW(&(W[TWVL * 10]), VSUB(T2x, T2A));
776 ST(&(x[WS(vs, 6) + WS(rs, 4)]), T2B, ms, &(x[WS(vs, 6)]));
777 T24 = BYTW(&(W[TWVL * 10]), VSUB(T20, T23));
778 ST(&(x[WS(vs, 6) + WS(rs, 3)]), T24, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
779 }
780 {
781 V T10, T1x, T3F, T38, T1y, Tu;
782 T10 = BYTW(&(W[TWVL * 10]), VSUB(TW, TZ));
783 ST(&(x[WS(vs, 6) + WS(rs, 1)]), T10, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
784 T1x = BYTW(&(W[TWVL * 10]), VSUB(T1t, T1w));
785 ST(&(x[WS(vs, 6) + WS(rs, 2)]), T1x, ms, &(x[WS(vs, 6)]));
786 T3F = BYTW(&(W[TWVL * 10]), VSUB(T3B, T3E));
787 ST(&(x[WS(vs, 6) + WS(rs, 6)]), T3F, ms, &(x[WS(vs, 6)]));
788 T38 = BYTW(&(W[TWVL * 10]), VSUB(T34, T37));
789 ST(&(x[WS(vs, 6) + WS(rs, 5)]), T38, ms, &(x[WS(vs, 6) + WS(rs, 1)]));
790 T1y = BYTW(&(W[TWVL * 2]), VADD(T1t, T1w));
791 ST(&(x[WS(vs, 2) + WS(rs, 2)]), T1y, ms, &(x[WS(vs, 2)]));
792 Tu = BYTW(&(W[TWVL * 2]), VADD(Tp, Ts));
793 ST(&(x[WS(vs, 2)]), Tu, ms, &(x[WS(vs, 2)]));
794 }
795 {
796 V T2C, T3G, T11, T25, T39, T4d;
797 T2C = BYTW(&(W[TWVL * 2]), VADD(T2x, T2A));
798 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T2C, ms, &(x[WS(vs, 2)]));
799 T3G = BYTW(&(W[TWVL * 2]), VADD(T3B, T3E));
800 ST(&(x[WS(vs, 2) + WS(rs, 6)]), T3G, ms, &(x[WS(vs, 2)]));
801 T11 = BYTW(&(W[TWVL * 2]), VADD(TW, TZ));
802 ST(&(x[WS(vs, 2) + WS(rs, 1)]), T11, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
803 T25 = BYTW(&(W[TWVL * 2]), VADD(T20, T23));
804 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T25, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
805 T39 = BYTW(&(W[TWVL * 2]), VADD(T34, T37));
806 ST(&(x[WS(vs, 2) + WS(rs, 5)]), T39, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
807 T4d = BYTW(&(W[TWVL * 2]), VADD(T48, T4b));
808 ST(&(x[WS(vs, 2) + WS(rs, 7)]), T4d, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
809 }
810 {
811 V Tx, T1B, T3c, T4g, T3J, T2F;
812 Tx = BYTW(&(W[TWVL * 6]), VSUB(Tv, Tw));
813 ST(&(x[WS(vs, 4)]), Tx, ms, &(x[WS(vs, 4)]));
814 T1B = BYTW(&(W[TWVL * 6]), VSUB(T1z, T1A));
815 ST(&(x[WS(vs, 4) + WS(rs, 2)]), T1B, ms, &(x[WS(vs, 4)]));
816 T3c = BYTW(&(W[TWVL * 6]), VSUB(T3a, T3b));
817 ST(&(x[WS(vs, 4) + WS(rs, 5)]), T3c, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
818 T4g = BYTW(&(W[TWVL * 6]), VSUB(T4e, T4f));
819 ST(&(x[WS(vs, 4) + WS(rs, 7)]), T4g, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
820 T3J = BYTW(&(W[TWVL * 6]), VSUB(T3H, T3I));
821 ST(&(x[WS(vs, 4) + WS(rs, 6)]), T3J, ms, &(x[WS(vs, 4)]));
822 T2F = BYTW(&(W[TWVL * 6]), VSUB(T2D, T2E));
823 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T2F, ms, &(x[WS(vs, 4)]));
824 }
825 T28 = BYTW(&(W[TWVL * 6]), VSUB(T26, T27));
826 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T28, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
827 T14 = BYTW(&(W[TWVL * 6]), VSUB(T12, T13));
828 ST(&(x[WS(vs, 4) + WS(rs, 1)]), T14, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
829 {
830 V Th, Ti, Tb, Tg;
831 Tb = VBYI(VSUB(T7, Ta));
832 Tg = VSUB(Te, Tf);
833 Th = BYTW(&(W[TWVL * 4]), VADD(Tb, Tg));
834 Ti = BYTW(&(W[TWVL * 8]), VSUB(Tg, Tb));
835 ST(&(x[WS(vs, 3)]), Th, ms, &(x[WS(vs, 3)]));
836 ST(&(x[WS(vs, 5)]), Ti, ms, &(x[WS(vs, 5)]));
837 }
838 {
839 V T40, T41, T3U, T3Z;
840 T3U = VBYI(VSUB(T3Q, T3T));
841 T3Z = VSUB(T3X, T3Y);
842 T40 = BYTW(&(W[TWVL * 4]), VADD(T3U, T3Z));
843 T41 = BYTW(&(W[TWVL * 8]), VSUB(T3Z, T3U));
844 ST(&(x[WS(vs, 3) + WS(rs, 7)]), T40, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
845 ST(&(x[WS(vs, 5) + WS(rs, 7)]), T41, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
846 }
847 {
848 V T2p, T2q, T2j, T2o;
849 T2j = VBYI(VSUB(T2f, T2i));
850 T2o = VSUB(T2m, T2n);
851 T2p = BYTW(&(W[TWVL * 4]), VADD(T2j, T2o));
852 T2q = BYTW(&(W[TWVL * 8]), VSUB(T2o, T2j));
853 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T2p, ms, &(x[WS(vs, 3)]));
854 ST(&(x[WS(vs, 5) + WS(rs, 4)]), T2q, ms, &(x[WS(vs, 5)]));
855 }
856 {
857 V T1S, T1T, T1M, T1R;
858 T1M = VBYI(VSUB(T1I, T1L));
859 T1R = VSUB(T1P, T1Q);
860 T1S = BYTW(&(W[TWVL * 4]), VADD(T1M, T1R));
861 T1T = BYTW(&(W[TWVL * 8]), VSUB(T1R, T1M));
862 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1S, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
863 ST(&(x[WS(vs, 5) + WS(rs, 3)]), T1T, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
864 }
865 {
866 V TO, TP, TI, TN;
867 TI = VBYI(VSUB(TE, TH));
868 TN = VSUB(TL, TM);
869 TO = BYTW(&(W[TWVL * 4]), VADD(TI, TN));
870 TP = BYTW(&(W[TWVL * 8]), VSUB(TN, TI));
871 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TO, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
872 ST(&(x[WS(vs, 5) + WS(rs, 1)]), TP, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
873 }
874 {
875 V T1l, T1m, T1f, T1k;
876 T1f = VBYI(VSUB(T1b, T1e));
877 T1k = VSUB(T1i, T1j);
878 T1l = BYTW(&(W[TWVL * 4]), VADD(T1f, T1k));
879 T1m = BYTW(&(W[TWVL * 8]), VSUB(T1k, T1f));
880 ST(&(x[WS(vs, 3) + WS(rs, 2)]), T1l, ms, &(x[WS(vs, 3)]));
881 ST(&(x[WS(vs, 5) + WS(rs, 2)]), T1m, ms, &(x[WS(vs, 5)]));
882 }
883 {
884 V T3t, T3u, T3n, T3s;
885 T3n = VBYI(VSUB(T3j, T3m));
886 T3s = VSUB(T3q, T3r);
887 T3t = BYTW(&(W[TWVL * 4]), VADD(T3n, T3s));
888 T3u = BYTW(&(W[TWVL * 8]), VSUB(T3s, T3n));
889 ST(&(x[WS(vs, 3) + WS(rs, 6)]), T3t, ms, &(x[WS(vs, 3)]));
890 ST(&(x[WS(vs, 5) + WS(rs, 6)]), T3u, ms, &(x[WS(vs, 5)]));
891 }
892 {
893 V T2W, T2X, T2Q, T2V;
894 T2Q = VBYI(VSUB(T2M, T2P));
895 T2V = VSUB(T2T, T2U);
896 T2W = BYTW(&(W[TWVL * 4]), VADD(T2Q, T2V));
897 T2X = BYTW(&(W[TWVL * 8]), VSUB(T2V, T2Q));
898 ST(&(x[WS(vs, 3) + WS(rs, 5)]), T2W, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
899 ST(&(x[WS(vs, 5) + WS(rs, 5)]), T2X, ms, &(x[WS(vs, 5) + WS(rs, 1)]));
900 }
901 {
902 V T1p, T1q, T1n, T1o;
903 T1n = VBYI(VADD(T1e, T1b));
904 T1o = VADD(T1i, T1j);
905 T1p = BYTW(&(W[0]), VADD(T1n, T1o));
906 T1q = BYTW(&(W[TWVL * 12]), VSUB(T1o, T1n));
907 ST(&(x[WS(vs, 1) + WS(rs, 2)]), T1p, ms, &(x[WS(vs, 1)]));
908 ST(&(x[WS(vs, 7) + WS(rs, 2)]), T1q, ms, &(x[WS(vs, 7)]));
909 }
910 {
911 V Tl, Tm, Tj, Tk;
912 Tj = VBYI(VADD(Ta, T7));
913 Tk = VADD(Te, Tf);
914 Tl = BYTW(&(W[0]), VADD(Tj, Tk));
915 Tm = BYTW(&(W[TWVL * 12]), VSUB(Tk, Tj));
916 ST(&(x[WS(vs, 1)]), Tl, ms, &(x[WS(vs, 1)]));
917 ST(&(x[WS(vs, 7)]), Tm, ms, &(x[WS(vs, 7)]));
918 }
919 {
920 V T2t, T2u, T2r, T2s;
921 T2r = VBYI(VADD(T2i, T2f));
922 T2s = VADD(T2m, T2n);
923 T2t = BYTW(&(W[0]), VADD(T2r, T2s));
924 T2u = BYTW(&(W[TWVL * 12]), VSUB(T2s, T2r));
925 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T2t, ms, &(x[WS(vs, 1)]));
926 ST(&(x[WS(vs, 7) + WS(rs, 4)]), T2u, ms, &(x[WS(vs, 7)]));
927 }
928 {
929 V T3x, T3y, T3v, T3w;
930 T3v = VBYI(VADD(T3m, T3j));
931 T3w = VADD(T3q, T3r);
932 T3x = BYTW(&(W[0]), VADD(T3v, T3w));
933 T3y = BYTW(&(W[TWVL * 12]), VSUB(T3w, T3v));
934 ST(&(x[WS(vs, 1) + WS(rs, 6)]), T3x, ms, &(x[WS(vs, 1)]));
935 ST(&(x[WS(vs, 7) + WS(rs, 6)]), T3y, ms, &(x[WS(vs, 7)]));
936 }
937 {
938 V TS, TT, TQ, TR;
939 TQ = VBYI(VADD(TH, TE));
940 TR = VADD(TL, TM);
941 TS = BYTW(&(W[0]), VADD(TQ, TR));
942 TT = BYTW(&(W[TWVL * 12]), VSUB(TR, TQ));
943 ST(&(x[WS(vs, 1) + WS(rs, 1)]), TS, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
944 ST(&(x[WS(vs, 7) + WS(rs, 1)]), TT, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
945 }
946 {
947 V T1W, T1X, T1U, T1V;
948 T1U = VBYI(VADD(T1L, T1I));
949 T1V = VADD(T1P, T1Q);
950 T1W = BYTW(&(W[0]), VADD(T1U, T1V));
951 T1X = BYTW(&(W[TWVL * 12]), VSUB(T1V, T1U));
952 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1W, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
953 ST(&(x[WS(vs, 7) + WS(rs, 3)]), T1X, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
954 }
955 {
956 V T30, T31, T2Y, T2Z;
957 T2Y = VBYI(VADD(T2P, T2M));
958 T2Z = VADD(T2T, T2U);
959 T30 = BYTW(&(W[0]), VADD(T2Y, T2Z));
960 T31 = BYTW(&(W[TWVL * 12]), VSUB(T2Z, T2Y));
961 ST(&(x[WS(vs, 1) + WS(rs, 5)]), T30, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
962 ST(&(x[WS(vs, 7) + WS(rs, 5)]), T31, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
963 }
964 {
965 V T44, T45, T42, T43;
966 T42 = VBYI(VADD(T3T, T3Q));
967 T43 = VADD(T3X, T3Y);
968 T44 = BYTW(&(W[0]), VADD(T42, T43));
969 T45 = BYTW(&(W[TWVL * 12]), VSUB(T43, T42));
970 ST(&(x[WS(vs, 1) + WS(rs, 7)]), T44, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
971 ST(&(x[WS(vs, 7) + WS(rs, 7)]), T45, ms, &(x[WS(vs, 7) + WS(rs, 1)]));
972 }
973 }
974 }
975 VLEAVE();
976 }
977
978 static const tw_instr twinstr[] = {
979 VTW(0, 1),
980 VTW(0, 2),
981 VTW(0, 3),
982 VTW(0, 4),
983 VTW(0, 5),
984 VTW(0, 6),
985 VTW(0, 7),
986 {TW_NEXT, VL, 0}
987 };
988
989 static const ct_desc desc = { 8, XSIMD_STRING("q1bv_8"), twinstr, &GENUS, {264, 128, 0, 0}, 0, 0, 0 };
990
991 void XSIMD(codelet_q1bv_8) (planner *p) {
992 X(kdft_difsq_register) (p, q1bv_8, &desc);
993 }
994 #endif /* HAVE_FMA */