Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/q1fv_5.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 5 -dif -name q1fv_5 -include q1f.h */
29
30 /*
31 * This function contains 100 FP additions, 95 FP multiplications,
32 * (or, 55 additions, 50 multiplications, 45 fused multiply/add),
33 * 69 stack variables, 4 constants, and 50 memory accesses
34 */
35 #include "q1f.h"
36
37 static void q1fv_5(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
38 {
39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
43 {
44 INT m;
45 R *x;
46 x = ri;
47 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(10, rs), MAKE_VOLATILE_STRIDE(10, vs)) {
48 V Te, T1w, Ty, TS, TW, Tb, T1t, Tv, T1g, T1c, TP, TV, T1f, T19, TY;
49 V TX;
50 {
51 V T1, T1j, Tl, Ti, Ta, T8, T1A, T1q, T1s, T9, TF, T1r, TZ, TR, TL;
52 V TC, Ts, Tu, TQ, TI, T15, T1b, T10, T11, Tt;
53 {
54 V T1n, T1o, T1k, T1l, T7, Td, T4, Tc;
55 {
56 V T5, T6, T2, T3;
57 T1 = LD(&(x[0]), ms, &(x[0]));
58 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
59 T6 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
60 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
61 T3 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
62 T1j = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
63 T1n = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
64 T1o = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
65 T1k = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
66 T1l = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
67 T7 = VADD(T5, T6);
68 Td = VSUB(T5, T6);
69 T4 = VADD(T2, T3);
70 Tc = VSUB(T2, T3);
71 }
72 {
73 V Tm, Tn, Tr, Tx, T1v, T1p;
74 Tl = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
75 T1v = VSUB(T1n, T1o);
76 T1p = VADD(T1n, T1o);
77 {
78 V T1u, T1m, Tp, Tq;
79 T1u = VSUB(T1k, T1l);
80 T1m = VADD(T1k, T1l);
81 Tp = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
82 Ti = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tc, Td));
83 Te = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Td, Tc));
84 Ta = VSUB(T4, T7);
85 T8 = VADD(T4, T7);
86 Tq = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
87 T1w = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1v, T1u));
88 T1A = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1u, T1v));
89 T1q = VADD(T1m, T1p);
90 T1s = VSUB(T1m, T1p);
91 Tm = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
92 T9 = VFNMS(LDK(KP250000000), T8, T1);
93 Tn = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
94 Tr = VADD(Tp, Tq);
95 Tx = VSUB(Tp, Tq);
96 }
97 {
98 V TJ, TK, TG, Tw, To, TH, T13, T14;
99 TF = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
100 T1r = VFNMS(LDK(KP250000000), T1q, T1j);
101 TJ = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
102 TK = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
103 TG = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
104 Tw = VSUB(Tm, Tn);
105 To = VADD(Tm, Tn);
106 TH = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
107 TZ = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
108 T13 = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
109 T14 = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
110 TR = VSUB(TJ, TK);
111 TL = VADD(TJ, TK);
112 Ty = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tx, Tw));
113 TC = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tw, Tx));
114 Ts = VADD(To, Tr);
115 Tu = VSUB(To, Tr);
116 TQ = VSUB(TG, TH);
117 TI = VADD(TG, TH);
118 T15 = VADD(T13, T14);
119 T1b = VSUB(T13, T14);
120 T10 = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
121 T11 = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
122 Tt = VFNMS(LDK(KP250000000), Ts, Tl);
123 }
124 }
125 }
126 {
127 V TO, T12, T1a, Th, T1z, TN, TM, T18, T17;
128 ST(&(x[0]), VADD(T1, T8), ms, &(x[0]));
129 TS = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TR, TQ));
130 TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TQ, TR));
131 TM = VADD(TI, TL);
132 TO = VSUB(TI, TL);
133 ST(&(x[WS(rs, 4)]), VADD(T1j, T1q), ms, &(x[0]));
134 T12 = VADD(T10, T11);
135 T1a = VSUB(T10, T11);
136 ST(&(x[WS(rs, 1)]), VADD(Tl, Ts), ms, &(x[WS(rs, 1)]));
137 Th = VFNMS(LDK(KP559016994), Ta, T9);
138 Tb = VFMA(LDK(KP559016994), Ta, T9);
139 T1t = VFMA(LDK(KP559016994), T1s, T1r);
140 T1z = VFNMS(LDK(KP559016994), T1s, T1r);
141 ST(&(x[WS(rs, 2)]), VADD(TF, TM), ms, &(x[0]));
142 TN = VFNMS(LDK(KP250000000), TM, TF);
143 {
144 V T16, Tk, Tj, T1C, T1B, TD, TE, TB;
145 TB = VFNMS(LDK(KP559016994), Tu, Tt);
146 Tv = VFMA(LDK(KP559016994), Tu, Tt);
147 T1g = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1a, T1b));
148 T1c = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1b, T1a));
149 T18 = VSUB(T12, T15);
150 T16 = VADD(T12, T15);
151 Tk = BYTWJ(&(W[TWVL * 4]), VFNMSI(Ti, Th));
152 Tj = BYTWJ(&(W[TWVL * 2]), VFMAI(Ti, Th));
153 T1C = BYTWJ(&(W[TWVL * 4]), VFNMSI(T1A, T1z));
154 T1B = BYTWJ(&(W[TWVL * 2]), VFMAI(T1A, T1z));
155 TD = BYTWJ(&(W[TWVL * 2]), VFMAI(TC, TB));
156 TE = BYTWJ(&(W[TWVL * 4]), VFNMSI(TC, TB));
157 ST(&(x[WS(rs, 3)]), VADD(TZ, T16), ms, &(x[WS(rs, 1)]));
158 T17 = VFNMS(LDK(KP250000000), T16, TZ);
159 ST(&(x[WS(vs, 3)]), Tk, ms, &(x[WS(vs, 3)]));
160 ST(&(x[WS(vs, 2)]), Tj, ms, &(x[WS(vs, 2)]));
161 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T1C, ms, &(x[WS(vs, 3)]));
162 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T1B, ms, &(x[WS(vs, 2)]));
163 ST(&(x[WS(vs, 2) + WS(rs, 1)]), TD, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
164 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
165 }
166 TP = VFMA(LDK(KP559016994), TO, TN);
167 TV = VFNMS(LDK(KP559016994), TO, TN);
168 T1f = VFNMS(LDK(KP559016994), T18, T17);
169 T19 = VFMA(LDK(KP559016994), T18, T17);
170 }
171 }
172 TY = BYTWJ(&(W[TWVL * 4]), VFNMSI(TW, TV));
173 TX = BYTWJ(&(W[TWVL * 2]), VFMAI(TW, TV));
174 {
175 V T1i, T1h, TU, TT;
176 T1i = BYTWJ(&(W[TWVL * 4]), VFNMSI(T1g, T1f));
177 T1h = BYTWJ(&(W[TWVL * 2]), VFMAI(T1g, T1f));
178 TU = BYTWJ(&(W[TWVL * 6]), VFMAI(TS, TP));
179 TT = BYTWJ(&(W[0]), VFNMSI(TS, TP));
180 {
181 V Tg, Tf, TA, Tz;
182 Tg = BYTWJ(&(W[TWVL * 6]), VFMAI(Te, Tb));
183 Tf = BYTWJ(&(W[0]), VFNMSI(Te, Tb));
184 TA = BYTWJ(&(W[TWVL * 6]), VFMAI(Ty, Tv));
185 Tz = BYTWJ(&(W[0]), VFNMSI(Ty, Tv));
186 {
187 V T1e, T1d, T1y, T1x;
188 T1e = BYTWJ(&(W[TWVL * 6]), VFMAI(T1c, T19));
189 T1d = BYTWJ(&(W[0]), VFNMSI(T1c, T19));
190 T1y = BYTWJ(&(W[TWVL * 6]), VFMAI(T1w, T1t));
191 T1x = BYTWJ(&(W[0]), VFNMSI(T1w, T1t));
192 ST(&(x[WS(vs, 3) + WS(rs, 2)]), TY, ms, &(x[WS(vs, 3)]));
193 ST(&(x[WS(vs, 2) + WS(rs, 2)]), TX, ms, &(x[WS(vs, 2)]));
194 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1i, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
195 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T1h, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
196 ST(&(x[WS(vs, 4) + WS(rs, 2)]), TU, ms, &(x[WS(vs, 4)]));
197 ST(&(x[WS(vs, 1) + WS(rs, 2)]), TT, ms, &(x[WS(vs, 1)]));
198 ST(&(x[WS(vs, 4)]), Tg, ms, &(x[WS(vs, 4)]));
199 ST(&(x[WS(vs, 1)]), Tf, ms, &(x[WS(vs, 1)]));
200 ST(&(x[WS(vs, 4) + WS(rs, 1)]), TA, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
201 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tz, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
202 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T1e, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
203 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1d, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
204 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T1y, ms, &(x[WS(vs, 4)]));
205 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T1x, ms, &(x[WS(vs, 1)]));
206 }
207 }
208 }
209 }
210 }
211 VLEAVE();
212 }
213
214 static const tw_instr twinstr[] = {
215 VTW(0, 1),
216 VTW(0, 2),
217 VTW(0, 3),
218 VTW(0, 4),
219 {TW_NEXT, VL, 0}
220 };
221
222 static const ct_desc desc = { 5, XSIMD_STRING("q1fv_5"), twinstr, &GENUS, {55, 50, 45, 0}, 0, 0, 0 };
223
224 void XSIMD(codelet_q1fv_5) (planner *p) {
225 X(kdft_difsq_register) (p, q1fv_5, &desc);
226 }
227 #else /* HAVE_FMA */
228
229 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 5 -dif -name q1fv_5 -include q1f.h */
230
231 /*
232 * This function contains 100 FP additions, 70 FP multiplications,
233 * (or, 85 additions, 55 multiplications, 15 fused multiply/add),
234 * 44 stack variables, 4 constants, and 50 memory accesses
235 */
236 #include "q1f.h"
237
238 static void q1fv_5(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
239 {
240 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
241 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
242 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
243 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
244 {
245 INT m;
246 R *x;
247 x = ri;
248 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(10, rs), MAKE_VOLATILE_STRIDE(10, vs)) {
249 V T8, T7, Th, Te, T9, Ta, T1q, T1p, T1z, T1w, T1r, T1s, Ts, Tr, TB;
250 V Ty, Tt, Tu, TM, TL, TV, TS, TN, TO, T16, T15, T1f, T1c, T17, T18;
251 {
252 V T6, Td, T3, Tc;
253 T8 = LD(&(x[0]), ms, &(x[0]));
254 {
255 V T4, T5, T1, T2;
256 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
257 T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
258 T6 = VADD(T4, T5);
259 Td = VSUB(T4, T5);
260 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
261 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
262 T3 = VADD(T1, T2);
263 Tc = VSUB(T1, T2);
264 }
265 T7 = VMUL(LDK(KP559016994), VSUB(T3, T6));
266 Th = VBYI(VFNMS(LDK(KP587785252), Tc, VMUL(LDK(KP951056516), Td)));
267 Te = VBYI(VFMA(LDK(KP951056516), Tc, VMUL(LDK(KP587785252), Td)));
268 T9 = VADD(T3, T6);
269 Ta = VFNMS(LDK(KP250000000), T9, T8);
270 }
271 {
272 V T1o, T1v, T1l, T1u;
273 T1q = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
274 {
275 V T1m, T1n, T1j, T1k;
276 T1m = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
277 T1n = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
278 T1o = VADD(T1m, T1n);
279 T1v = VSUB(T1m, T1n);
280 T1j = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
281 T1k = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
282 T1l = VADD(T1j, T1k);
283 T1u = VSUB(T1j, T1k);
284 }
285 T1p = VMUL(LDK(KP559016994), VSUB(T1l, T1o));
286 T1z = VBYI(VFNMS(LDK(KP587785252), T1u, VMUL(LDK(KP951056516), T1v)));
287 T1w = VBYI(VFMA(LDK(KP951056516), T1u, VMUL(LDK(KP587785252), T1v)));
288 T1r = VADD(T1l, T1o);
289 T1s = VFNMS(LDK(KP250000000), T1r, T1q);
290 }
291 {
292 V Tq, Tx, Tn, Tw;
293 Ts = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
294 {
295 V To, Tp, Tl, Tm;
296 To = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
297 Tp = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
298 Tq = VADD(To, Tp);
299 Tx = VSUB(To, Tp);
300 Tl = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
301 Tm = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
302 Tn = VADD(Tl, Tm);
303 Tw = VSUB(Tl, Tm);
304 }
305 Tr = VMUL(LDK(KP559016994), VSUB(Tn, Tq));
306 TB = VBYI(VFNMS(LDK(KP587785252), Tw, VMUL(LDK(KP951056516), Tx)));
307 Ty = VBYI(VFMA(LDK(KP951056516), Tw, VMUL(LDK(KP587785252), Tx)));
308 Tt = VADD(Tn, Tq);
309 Tu = VFNMS(LDK(KP250000000), Tt, Ts);
310 }
311 {
312 V TK, TR, TH, TQ;
313 TM = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
314 {
315 V TI, TJ, TF, TG;
316 TI = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
317 TJ = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
318 TK = VADD(TI, TJ);
319 TR = VSUB(TI, TJ);
320 TF = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
321 TG = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
322 TH = VADD(TF, TG);
323 TQ = VSUB(TF, TG);
324 }
325 TL = VMUL(LDK(KP559016994), VSUB(TH, TK));
326 TV = VBYI(VFNMS(LDK(KP587785252), TQ, VMUL(LDK(KP951056516), TR)));
327 TS = VBYI(VFMA(LDK(KP951056516), TQ, VMUL(LDK(KP587785252), TR)));
328 TN = VADD(TH, TK);
329 TO = VFNMS(LDK(KP250000000), TN, TM);
330 }
331 {
332 V T14, T1b, T11, T1a;
333 T16 = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
334 {
335 V T12, T13, TZ, T10;
336 T12 = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
337 T13 = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
338 T14 = VADD(T12, T13);
339 T1b = VSUB(T12, T13);
340 TZ = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
341 T10 = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
342 T11 = VADD(TZ, T10);
343 T1a = VSUB(TZ, T10);
344 }
345 T15 = VMUL(LDK(KP559016994), VSUB(T11, T14));
346 T1f = VBYI(VFNMS(LDK(KP587785252), T1a, VMUL(LDK(KP951056516), T1b)));
347 T1c = VBYI(VFMA(LDK(KP951056516), T1a, VMUL(LDK(KP587785252), T1b)));
348 T17 = VADD(T11, T14);
349 T18 = VFNMS(LDK(KP250000000), T17, T16);
350 }
351 ST(&(x[0]), VADD(T8, T9), ms, &(x[0]));
352 ST(&(x[WS(rs, 4)]), VADD(T1q, T1r), ms, &(x[0]));
353 ST(&(x[WS(rs, 2)]), VADD(TM, TN), ms, &(x[0]));
354 ST(&(x[WS(rs, 3)]), VADD(T16, T17), ms, &(x[WS(rs, 1)]));
355 ST(&(x[WS(rs, 1)]), VADD(Ts, Tt), ms, &(x[WS(rs, 1)]));
356 {
357 V Tj, Tk, Ti, T1B, T1C, T1A;
358 Ti = VSUB(Ta, T7);
359 Tj = BYTWJ(&(W[TWVL * 2]), VADD(Th, Ti));
360 Tk = BYTWJ(&(W[TWVL * 4]), VSUB(Ti, Th));
361 ST(&(x[WS(vs, 2)]), Tj, ms, &(x[WS(vs, 2)]));
362 ST(&(x[WS(vs, 3)]), Tk, ms, &(x[WS(vs, 3)]));
363 T1A = VSUB(T1s, T1p);
364 T1B = BYTWJ(&(W[TWVL * 2]), VADD(T1z, T1A));
365 T1C = BYTWJ(&(W[TWVL * 4]), VSUB(T1A, T1z));
366 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T1B, ms, &(x[WS(vs, 2)]));
367 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T1C, ms, &(x[WS(vs, 3)]));
368 }
369 {
370 V T1h, T1i, T1g, TD, TE, TC;
371 T1g = VSUB(T18, T15);
372 T1h = BYTWJ(&(W[TWVL * 2]), VADD(T1f, T1g));
373 T1i = BYTWJ(&(W[TWVL * 4]), VSUB(T1g, T1f));
374 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T1h, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
375 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1i, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
376 TC = VSUB(Tu, Tr);
377 TD = BYTWJ(&(W[TWVL * 2]), VADD(TB, TC));
378 TE = BYTWJ(&(W[TWVL * 4]), VSUB(TC, TB));
379 ST(&(x[WS(vs, 2) + WS(rs, 1)]), TD, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
380 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
381 }
382 {
383 V TX, TY, TW, TT, TU, TP;
384 TW = VSUB(TO, TL);
385 TX = BYTWJ(&(W[TWVL * 2]), VADD(TV, TW));
386 TY = BYTWJ(&(W[TWVL * 4]), VSUB(TW, TV));
387 ST(&(x[WS(vs, 2) + WS(rs, 2)]), TX, ms, &(x[WS(vs, 2)]));
388 ST(&(x[WS(vs, 3) + WS(rs, 2)]), TY, ms, &(x[WS(vs, 3)]));
389 TP = VADD(TL, TO);
390 TT = BYTWJ(&(W[0]), VSUB(TP, TS));
391 TU = BYTWJ(&(W[TWVL * 6]), VADD(TS, TP));
392 ST(&(x[WS(vs, 1) + WS(rs, 2)]), TT, ms, &(x[WS(vs, 1)]));
393 ST(&(x[WS(vs, 4) + WS(rs, 2)]), TU, ms, &(x[WS(vs, 4)]));
394 }
395 {
396 V Tf, Tg, Tb, Tz, TA, Tv;
397 Tb = VADD(T7, Ta);
398 Tf = BYTWJ(&(W[0]), VSUB(Tb, Te));
399 Tg = BYTWJ(&(W[TWVL * 6]), VADD(Te, Tb));
400 ST(&(x[WS(vs, 1)]), Tf, ms, &(x[WS(vs, 1)]));
401 ST(&(x[WS(vs, 4)]), Tg, ms, &(x[WS(vs, 4)]));
402 Tv = VADD(Tr, Tu);
403 Tz = BYTWJ(&(W[0]), VSUB(Tv, Ty));
404 TA = BYTWJ(&(W[TWVL * 6]), VADD(Ty, Tv));
405 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tz, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
406 ST(&(x[WS(vs, 4) + WS(rs, 1)]), TA, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
407 }
408 {
409 V T1d, T1e, T19, T1x, T1y, T1t;
410 T19 = VADD(T15, T18);
411 T1d = BYTWJ(&(W[0]), VSUB(T19, T1c));
412 T1e = BYTWJ(&(W[TWVL * 6]), VADD(T1c, T19));
413 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1d, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
414 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T1e, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
415 T1t = VADD(T1p, T1s);
416 T1x = BYTWJ(&(W[0]), VSUB(T1t, T1w));
417 T1y = BYTWJ(&(W[TWVL * 6]), VADD(T1w, T1t));
418 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T1x, ms, &(x[WS(vs, 1)]));
419 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T1y, ms, &(x[WS(vs, 4)]));
420 }
421 }
422 }
423 VLEAVE();
424 }
425
426 static const tw_instr twinstr[] = {
427 VTW(0, 1),
428 VTW(0, 2),
429 VTW(0, 3),
430 VTW(0, 4),
431 {TW_NEXT, VL, 0}
432 };
433
434 static const ct_desc desc = { 5, XSIMD_STRING("q1fv_5"), twinstr, &GENUS, {85, 55, 15, 0}, 0, 0, 0 };
435
436 void XSIMD(codelet_q1fv_5) (planner *p) {
437 X(kdft_difsq_register) (p, q1fv_5, &desc);
438 }
439 #endif /* HAVE_FMA */