1 /*
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 *
19 */
20
21 /* This file was automatically generated --- DO NOT EDIT */
22 /* Generated on Thu Dec 10 07:05:41 EST 2020 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_twidsq_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 5 -dif -name q1bv_5 -include dft/simd/q1b.h -sign 1 */
29
30 /*
31 * This function contains 100 FP additions, 95 FP multiplications,
32 * (or, 55 additions, 50 multiplications, 45 fused multiply/add),
33 * 44 stack variables, 4 constants, and 50 memory accesses
34 */
35 #include "dft/simd/q1b.h"
36
q1bv_5(R * ri,R * ii,const R * W,stride rs,stride vs,INT mb,INT me,INT ms)37 static void q1bv_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 = ii;
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 T1, Ta, Ti, Te, T8, T9, T1j, T1s, T1A, T1w, T1q, T1r, Tl, Tu, TC;
49 V Ty, Ts, Tt, TF, TO, TW, TS, TM, TN, TZ, T18, T1g, T1c, T16, T17;
50 {
51 V T7, Td, T4, Tc;
52 T1 = LD(&(x[0]), ms, &(x[0]));
53 {
54 V T5, T6, T2, T3;
55 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
56 T6 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
57 T7 = VADD(T5, T6);
58 Td = VSUB(T5, T6);
59 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
60 T3 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
61 T4 = VADD(T2, T3);
62 Tc = VSUB(T2, T3);
63 }
64 Ta = VSUB(T4, T7);
65 Ti = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tc, Td));
66 Te = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Td, Tc));
67 T8 = VADD(T4, T7);
68 T9 = VFNMS(LDK(KP250000000), T8, T1);
69 }
70 {
71 V T1p, T1v, T1m, T1u;
72 T1j = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
73 {
74 V T1n, T1o, T1k, T1l;
75 T1n = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
76 T1o = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
77 T1p = VADD(T1n, T1o);
78 T1v = VSUB(T1n, T1o);
79 T1k = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
80 T1l = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
81 T1m = VADD(T1k, T1l);
82 T1u = VSUB(T1k, T1l);
83 }
84 T1s = VSUB(T1m, T1p);
85 T1A = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1u, T1v));
86 T1w = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1v, T1u));
87 T1q = VADD(T1m, T1p);
88 T1r = VFNMS(LDK(KP250000000), T1q, T1j);
89 }
90 {
91 V Tr, Tx, To, Tw;
92 Tl = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
93 {
94 V Tp, Tq, Tm, Tn;
95 Tp = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
96 Tq = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
97 Tr = VADD(Tp, Tq);
98 Tx = VSUB(Tp, Tq);
99 Tm = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
100 Tn = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
101 To = VADD(Tm, Tn);
102 Tw = VSUB(Tm, Tn);
103 }
104 Tu = VSUB(To, Tr);
105 TC = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tw, Tx));
106 Ty = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tx, Tw));
107 Ts = VADD(To, Tr);
108 Tt = VFNMS(LDK(KP250000000), Ts, Tl);
109 }
110 {
111 V TL, TR, TI, TQ;
112 TF = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
113 {
114 V TJ, TK, TG, TH;
115 TJ = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
116 TK = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
117 TL = VADD(TJ, TK);
118 TR = VSUB(TJ, TK);
119 TG = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
120 TH = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
121 TI = VADD(TG, TH);
122 TQ = VSUB(TG, TH);
123 }
124 TO = VSUB(TI, TL);
125 TW = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TQ, TR));
126 TS = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TR, TQ));
127 TM = VADD(TI, TL);
128 TN = VFNMS(LDK(KP250000000), TM, TF);
129 }
130 {
131 V T15, T1b, T12, T1a;
132 TZ = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
133 {
134 V T13, T14, T10, T11;
135 T13 = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
136 T14 = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
137 T15 = VADD(T13, T14);
138 T1b = VSUB(T13, T14);
139 T10 = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
140 T11 = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
141 T12 = VADD(T10, T11);
142 T1a = VSUB(T10, T11);
143 }
144 T18 = VSUB(T12, T15);
145 T1g = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1a, T1b));
146 T1c = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1b, T1a));
147 T16 = VADD(T12, T15);
148 T17 = VFNMS(LDK(KP250000000), T16, TZ);
149 }
150 ST(&(x[0]), VADD(T1, T8), ms, &(x[0]));
151 ST(&(x[WS(rs, 4)]), VADD(T1j, T1q), ms, &(x[0]));
152 ST(&(x[WS(rs, 2)]), VADD(TF, TM), ms, &(x[0]));
153 ST(&(x[WS(rs, 3)]), VADD(TZ, T16), ms, &(x[WS(rs, 1)]));
154 ST(&(x[WS(rs, 1)]), VADD(Tl, Ts), ms, &(x[WS(rs, 1)]));
155 {
156 V Tj, Tk, Th, T1B, T1C, T1z;
157 Th = VFNMS(LDK(KP559016994), Ta, T9);
158 Tj = BYTW(&(W[TWVL * 2]), VFNMSI(Ti, Th));
159 Tk = BYTW(&(W[TWVL * 4]), VFMAI(Ti, Th));
160 ST(&(x[WS(vs, 2)]), Tj, ms, &(x[WS(vs, 2)]));
161 ST(&(x[WS(vs, 3)]), Tk, ms, &(x[WS(vs, 3)]));
162 T1z = VFNMS(LDK(KP559016994), T1s, T1r);
163 T1B = BYTW(&(W[TWVL * 2]), VFNMSI(T1A, T1z));
164 T1C = BYTW(&(W[TWVL * 4]), VFMAI(T1A, T1z));
165 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T1B, ms, &(x[WS(vs, 2)]));
166 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T1C, ms, &(x[WS(vs, 3)]));
167 }
168 {
169 V T1h, T1i, T1f, TD, TE, TB;
170 T1f = VFNMS(LDK(KP559016994), T18, T17);
171 T1h = BYTW(&(W[TWVL * 2]), VFNMSI(T1g, T1f));
172 T1i = BYTW(&(W[TWVL * 4]), VFMAI(T1g, T1f));
173 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T1h, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
174 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1i, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
175 TB = VFNMS(LDK(KP559016994), Tu, Tt);
176 TD = BYTW(&(W[TWVL * 2]), VFNMSI(TC, TB));
177 TE = BYTW(&(W[TWVL * 4]), VFMAI(TC, TB));
178 ST(&(x[WS(vs, 2) + WS(rs, 1)]), TD, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
179 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
180 }
181 {
182 V TX, TY, TV, TT, TU, TP;
183 TV = VFNMS(LDK(KP559016994), TO, TN);
184 TX = BYTW(&(W[TWVL * 2]), VFNMSI(TW, TV));
185 TY = BYTW(&(W[TWVL * 4]), VFMAI(TW, TV));
186 ST(&(x[WS(vs, 2) + WS(rs, 2)]), TX, ms, &(x[WS(vs, 2)]));
187 ST(&(x[WS(vs, 3) + WS(rs, 2)]), TY, ms, &(x[WS(vs, 3)]));
188 TP = VFMA(LDK(KP559016994), TO, TN);
189 TT = BYTW(&(W[0]), VFMAI(TS, TP));
190 TU = BYTW(&(W[TWVL * 6]), VFNMSI(TS, TP));
191 ST(&(x[WS(vs, 1) + WS(rs, 2)]), TT, ms, &(x[WS(vs, 1)]));
192 ST(&(x[WS(vs, 4) + WS(rs, 2)]), TU, ms, &(x[WS(vs, 4)]));
193 }
194 {
195 V Tf, Tg, Tb, Tz, TA, Tv;
196 Tb = VFMA(LDK(KP559016994), Ta, T9);
197 Tf = BYTW(&(W[0]), VFMAI(Te, Tb));
198 Tg = BYTW(&(W[TWVL * 6]), VFNMSI(Te, Tb));
199 ST(&(x[WS(vs, 1)]), Tf, ms, &(x[WS(vs, 1)]));
200 ST(&(x[WS(vs, 4)]), Tg, ms, &(x[WS(vs, 4)]));
201 Tv = VFMA(LDK(KP559016994), Tu, Tt);
202 Tz = BYTW(&(W[0]), VFMAI(Ty, Tv));
203 TA = BYTW(&(W[TWVL * 6]), VFNMSI(Ty, Tv));
204 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tz, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
205 ST(&(x[WS(vs, 4) + WS(rs, 1)]), TA, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
206 }
207 {
208 V T1d, T1e, T19, T1x, T1y, T1t;
209 T19 = VFMA(LDK(KP559016994), T18, T17);
210 T1d = BYTW(&(W[0]), VFMAI(T1c, T19));
211 T1e = BYTW(&(W[TWVL * 6]), VFNMSI(T1c, T19));
212 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1d, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
213 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T1e, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
214 T1t = VFMA(LDK(KP559016994), T1s, T1r);
215 T1x = BYTW(&(W[0]), VFMAI(T1w, T1t));
216 T1y = BYTW(&(W[TWVL * 6]), VFNMSI(T1w, T1t));
217 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T1x, ms, &(x[WS(vs, 1)]));
218 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T1y, ms, &(x[WS(vs, 4)]));
219 }
220 }
221 }
222 VLEAVE();
223 }
224
225 static const tw_instr twinstr[] = {
226 VTW(0, 1),
227 VTW(0, 2),
228 VTW(0, 3),
229 VTW(0, 4),
230 { TW_NEXT, VL, 0 }
231 };
232
233 static const ct_desc desc = { 5, XSIMD_STRING("q1bv_5"), twinstr, &GENUS, { 55, 50, 45, 0 }, 0, 0, 0 };
234
XSIMD(codelet_q1bv_5)235 void XSIMD(codelet_q1bv_5) (planner *p) {
236 X(kdft_difsq_register) (p, q1bv_5, &desc);
237 }
238 #else
239
240 /* Generated by: ../../../genfft/gen_twidsq_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 5 -dif -name q1bv_5 -include dft/simd/q1b.h -sign 1 */
241
242 /*
243 * This function contains 100 FP additions, 70 FP multiplications,
244 * (or, 85 additions, 55 multiplications, 15 fused multiply/add),
245 * 44 stack variables, 4 constants, and 50 memory accesses
246 */
247 #include "dft/simd/q1b.h"
248
q1bv_5(R * ri,R * ii,const R * W,stride rs,stride vs,INT mb,INT me,INT ms)249 static void q1bv_5(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
250 {
251 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
252 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
253 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
254 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
255 {
256 INT m;
257 R *x;
258 x = ii;
259 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)) {
260 V Tb, T7, Th, Ta, Tc, Td, T1t, T1p, T1z, T1s, T1u, T1v, Tv, Tr, TB;
261 V Tu, Tw, Tx, TP, TL, TV, TO, TQ, TR, T19, T15, T1f, T18, T1a, T1b;
262 {
263 V T6, T9, T3, T8;
264 Tb = LD(&(x[0]), ms, &(x[0]));
265 {
266 V T4, T5, T1, T2;
267 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
268 T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
269 T6 = VSUB(T4, T5);
270 T9 = VADD(T4, T5);
271 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
272 T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
273 T3 = VSUB(T1, T2);
274 T8 = VADD(T1, T2);
275 }
276 T7 = VBYI(VFMA(LDK(KP951056516), T3, VMUL(LDK(KP587785252), T6)));
277 Th = VBYI(VFNMS(LDK(KP951056516), T6, VMUL(LDK(KP587785252), T3)));
278 Ta = VMUL(LDK(KP559016994), VSUB(T8, T9));
279 Tc = VADD(T8, T9);
280 Td = VFNMS(LDK(KP250000000), Tc, Tb);
281 }
282 {
283 V T1o, T1r, T1l, T1q;
284 T1t = LD(&(x[WS(vs, 4)]), ms, &(x[WS(vs, 4)]));
285 {
286 V T1m, T1n, T1j, T1k;
287 T1m = LD(&(x[WS(vs, 4) + WS(rs, 2)]), ms, &(x[WS(vs, 4)]));
288 T1n = LD(&(x[WS(vs, 4) + WS(rs, 3)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
289 T1o = VSUB(T1m, T1n);
290 T1r = VADD(T1m, T1n);
291 T1j = LD(&(x[WS(vs, 4) + WS(rs, 1)]), ms, &(x[WS(vs, 4) + WS(rs, 1)]));
292 T1k = LD(&(x[WS(vs, 4) + WS(rs, 4)]), ms, &(x[WS(vs, 4)]));
293 T1l = VSUB(T1j, T1k);
294 T1q = VADD(T1j, T1k);
295 }
296 T1p = VBYI(VFMA(LDK(KP951056516), T1l, VMUL(LDK(KP587785252), T1o)));
297 T1z = VBYI(VFNMS(LDK(KP951056516), T1o, VMUL(LDK(KP587785252), T1l)));
298 T1s = VMUL(LDK(KP559016994), VSUB(T1q, T1r));
299 T1u = VADD(T1q, T1r);
300 T1v = VFNMS(LDK(KP250000000), T1u, T1t);
301 }
302 {
303 V Tq, Tt, Tn, Ts;
304 Tv = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)]));
305 {
306 V To, Tp, Tl, Tm;
307 To = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)]));
308 Tp = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
309 Tq = VSUB(To, Tp);
310 Tt = VADD(To, Tp);
311 Tl = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)]));
312 Tm = LD(&(x[WS(vs, 1) + WS(rs, 4)]), ms, &(x[WS(vs, 1)]));
313 Tn = VSUB(Tl, Tm);
314 Ts = VADD(Tl, Tm);
315 }
316 Tr = VBYI(VFMA(LDK(KP951056516), Tn, VMUL(LDK(KP587785252), Tq)));
317 TB = VBYI(VFNMS(LDK(KP951056516), Tq, VMUL(LDK(KP587785252), Tn)));
318 Tu = VMUL(LDK(KP559016994), VSUB(Ts, Tt));
319 Tw = VADD(Ts, Tt);
320 Tx = VFNMS(LDK(KP250000000), Tw, Tv);
321 }
322 {
323 V TK, TN, TH, TM;
324 TP = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)]));
325 {
326 V TI, TJ, TF, TG;
327 TI = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)]));
328 TJ = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
329 TK = VSUB(TI, TJ);
330 TN = VADD(TI, TJ);
331 TF = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)]));
332 TG = LD(&(x[WS(vs, 2) + WS(rs, 4)]), ms, &(x[WS(vs, 2)]));
333 TH = VSUB(TF, TG);
334 TM = VADD(TF, TG);
335 }
336 TL = VBYI(VFMA(LDK(KP951056516), TH, VMUL(LDK(KP587785252), TK)));
337 TV = VBYI(VFNMS(LDK(KP951056516), TK, VMUL(LDK(KP587785252), TH)));
338 TO = VMUL(LDK(KP559016994), VSUB(TM, TN));
339 TQ = VADD(TM, TN);
340 TR = VFNMS(LDK(KP250000000), TQ, TP);
341 }
342 {
343 V T14, T17, T11, T16;
344 T19 = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)]));
345 {
346 V T12, T13, TZ, T10;
347 T12 = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)]));
348 T13 = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
349 T14 = VSUB(T12, T13);
350 T17 = VADD(T12, T13);
351 TZ = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)]));
352 T10 = LD(&(x[WS(vs, 3) + WS(rs, 4)]), ms, &(x[WS(vs, 3)]));
353 T11 = VSUB(TZ, T10);
354 T16 = VADD(TZ, T10);
355 }
356 T15 = VBYI(VFMA(LDK(KP951056516), T11, VMUL(LDK(KP587785252), T14)));
357 T1f = VBYI(VFNMS(LDK(KP951056516), T14, VMUL(LDK(KP587785252), T11)));
358 T18 = VMUL(LDK(KP559016994), VSUB(T16, T17));
359 T1a = VADD(T16, T17);
360 T1b = VFNMS(LDK(KP250000000), T1a, T19);
361 }
362 ST(&(x[0]), VADD(Tb, Tc), ms, &(x[0]));
363 ST(&(x[WS(rs, 4)]), VADD(T1t, T1u), ms, &(x[0]));
364 ST(&(x[WS(rs, 2)]), VADD(TP, TQ), ms, &(x[0]));
365 ST(&(x[WS(rs, 3)]), VADD(T19, T1a), ms, &(x[WS(rs, 1)]));
366 ST(&(x[WS(rs, 1)]), VADD(Tv, Tw), ms, &(x[WS(rs, 1)]));
367 {
368 V Tj, Tk, Ti, T1B, T1C, T1A;
369 Ti = VSUB(Td, Ta);
370 Tj = BYTW(&(W[TWVL * 2]), VADD(Th, Ti));
371 Tk = BYTW(&(W[TWVL * 4]), VSUB(Ti, Th));
372 ST(&(x[WS(vs, 2)]), Tj, ms, &(x[WS(vs, 2)]));
373 ST(&(x[WS(vs, 3)]), Tk, ms, &(x[WS(vs, 3)]));
374 T1A = VSUB(T1v, T1s);
375 T1B = BYTW(&(W[TWVL * 2]), VADD(T1z, T1A));
376 T1C = BYTW(&(W[TWVL * 4]), VSUB(T1A, T1z));
377 ST(&(x[WS(vs, 2) + WS(rs, 4)]), T1B, ms, &(x[WS(vs, 2)]));
378 ST(&(x[WS(vs, 3) + WS(rs, 4)]), T1C, ms, &(x[WS(vs, 3)]));
379 }
380 {
381 V T1h, T1i, T1g, TD, TE, TC;
382 T1g = VSUB(T1b, T18);
383 T1h = BYTW(&(W[TWVL * 2]), VADD(T1f, T1g));
384 T1i = BYTW(&(W[TWVL * 4]), VSUB(T1g, T1f));
385 ST(&(x[WS(vs, 2) + WS(rs, 3)]), T1h, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
386 ST(&(x[WS(vs, 3) + WS(rs, 3)]), T1i, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
387 TC = VSUB(Tx, Tu);
388 TD = BYTW(&(W[TWVL * 2]), VADD(TB, TC));
389 TE = BYTW(&(W[TWVL * 4]), VSUB(TC, TB));
390 ST(&(x[WS(vs, 2) + WS(rs, 1)]), TD, ms, &(x[WS(vs, 2) + WS(rs, 1)]));
391 ST(&(x[WS(vs, 3) + WS(rs, 1)]), TE, ms, &(x[WS(vs, 3) + WS(rs, 1)]));
392 }
393 {
394 V TX, TY, TW, TT, TU, TS;
395 TW = VSUB(TR, TO);
396 TX = BYTW(&(W[TWVL * 2]), VADD(TV, TW));
397 TY = BYTW(&(W[TWVL * 4]), VSUB(TW, TV));
398 ST(&(x[WS(vs, 2) + WS(rs, 2)]), TX, ms, &(x[WS(vs, 2)]));
399 ST(&(x[WS(vs, 3) + WS(rs, 2)]), TY, ms, &(x[WS(vs, 3)]));
400 TS = VADD(TO, TR);
401 TT = BYTW(&(W[0]), VADD(TL, TS));
402 TU = BYTW(&(W[TWVL * 6]), VSUB(TS, TL));
403 ST(&(x[WS(vs, 1) + WS(rs, 2)]), TT, ms, &(x[WS(vs, 1)]));
404 ST(&(x[WS(vs, 4) + WS(rs, 2)]), TU, ms, &(x[WS(vs, 4)]));
405 }
406 {
407 V Tf, Tg, Te, Tz, TA, Ty;
408 Te = VADD(Ta, Td);
409 Tf = BYTW(&(W[0]), VADD(T7, Te));
410 Tg = BYTW(&(W[TWVL * 6]), VSUB(Te, T7));
411 ST(&(x[WS(vs, 1)]), Tf, ms, &(x[WS(vs, 1)]));
412 ST(&(x[WS(vs, 4)]), Tg, ms, &(x[WS(vs, 4)]));
413 Ty = VADD(Tu, Tx);
414 Tz = BYTW(&(W[0]), VADD(Tr, Ty));
415 TA = BYTW(&(W[TWVL * 6]), VSUB(Ty, Tr));
416 ST(&(x[WS(vs, 1) + WS(rs, 1)]), Tz, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
417 ST(&(x[WS(vs, 4) + WS(rs, 1)]), TA, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
418 }
419 {
420 V T1d, T1e, T1c, T1x, T1y, T1w;
421 T1c = VADD(T18, T1b);
422 T1d = BYTW(&(W[0]), VADD(T15, T1c));
423 T1e = BYTW(&(W[TWVL * 6]), VSUB(T1c, T15));
424 ST(&(x[WS(vs, 1) + WS(rs, 3)]), T1d, ms, &(x[WS(vs, 1) + WS(rs, 1)]));
425 ST(&(x[WS(vs, 4) + WS(rs, 3)]), T1e, ms, &(x[WS(vs, 4) + WS(rs, 1)]));
426 T1w = VADD(T1s, T1v);
427 T1x = BYTW(&(W[0]), VADD(T1p, T1w));
428 T1y = BYTW(&(W[TWVL * 6]), VSUB(T1w, T1p));
429 ST(&(x[WS(vs, 1) + WS(rs, 4)]), T1x, ms, &(x[WS(vs, 1)]));
430 ST(&(x[WS(vs, 4) + WS(rs, 4)]), T1y, ms, &(x[WS(vs, 4)]));
431 }
432 }
433 }
434 VLEAVE();
435 }
436
437 static const tw_instr twinstr[] = {
438 VTW(0, 1),
439 VTW(0, 2),
440 VTW(0, 3),
441 VTW(0, 4),
442 { TW_NEXT, VL, 0 }
443 };
444
445 static const ct_desc desc = { 5, XSIMD_STRING("q1bv_5"), twinstr, &GENUS, { 85, 55, 15, 0 }, 0, 0, 0 };
446
XSIMD(codelet_q1bv_5)447 void XSIMD(codelet_q1bv_5) (planner *p) {
448 X(kdft_difsq_register) (p, q1bv_5, &desc);
449 }
450 #endif
451