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 07:04:41 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_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n1fv_16 -include dft/simd/n1f.h */
29
30 /*
31 * This function contains 72 FP additions, 34 FP multiplications,
32 * (or, 38 additions, 0 multiplications, 34 fused multiply/add),
33 * 30 stack variables, 3 constants, and 32 memory accesses
34 */
35 #include "dft/simd/n1f.h"
36
n1fv_16(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1fv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
41 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
42 {
43 INT i;
44 const R *xi;
45 R *xo;
46 xi = ri;
47 xo = ro;
48 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
49 V T7, TU, Tz, TH, Tu, TV, TA, TK, Te, TX, TC, TO, Tl, TY, TD;
50 V TR;
51 {
52 V T1, T2, T3, T4, T5, T6;
53 T1 = LD(&(xi[0]), ivs, &(xi[0]));
54 T2 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
55 T3 = VADD(T1, T2);
56 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
57 T5 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
58 T6 = VADD(T4, T5);
59 T7 = VSUB(T3, T6);
60 TU = VSUB(T4, T5);
61 Tz = VADD(T3, T6);
62 TH = VSUB(T1, T2);
63 }
64 {
65 V Tq, TJ, Tt, TI;
66 {
67 V To, Tp, Tr, Ts;
68 To = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
69 Tp = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
70 Tq = VADD(To, Tp);
71 TJ = VSUB(To, Tp);
72 Tr = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
73 Ts = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
74 Tt = VADD(Tr, Ts);
75 TI = VSUB(Tr, Ts);
76 }
77 Tu = VSUB(Tq, Tt);
78 TV = VSUB(TJ, TI);
79 TA = VADD(Tt, Tq);
80 TK = VADD(TI, TJ);
81 }
82 {
83 V Ta, TM, Td, TN;
84 {
85 V T8, T9, Tb, Tc;
86 T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
87 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
88 Ta = VADD(T8, T9);
89 TM = VSUB(T8, T9);
90 Tb = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
91 Tc = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
92 Td = VADD(Tb, Tc);
93 TN = VSUB(Tb, Tc);
94 }
95 Te = VSUB(Ta, Td);
96 TX = VFMA(LDK(KP414213562), TM, TN);
97 TC = VADD(Ta, Td);
98 TO = VFNMS(LDK(KP414213562), TN, TM);
99 }
100 {
101 V Th, TP, Tk, TQ;
102 {
103 V Tf, Tg, Ti, Tj;
104 Tf = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
105 Tg = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
106 Th = VADD(Tf, Tg);
107 TP = VSUB(Tf, Tg);
108 Ti = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
109 Tj = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
110 Tk = VADD(Ti, Tj);
111 TQ = VSUB(Tj, Ti);
112 }
113 Tl = VSUB(Th, Tk);
114 TY = VFMA(LDK(KP414213562), TP, TQ);
115 TD = VADD(Th, Tk);
116 TR = VFNMS(LDK(KP414213562), TQ, TP);
117 }
118 {
119 V TB, TE, TF, TG;
120 TB = VADD(Tz, TA);
121 TE = VADD(TC, TD);
122 ST(&(xo[WS(os, 8)]), VSUB(TB, TE), ovs, &(xo[0]));
123 ST(&(xo[0]), VADD(TB, TE), ovs, &(xo[0]));
124 TF = VSUB(Tz, TA);
125 TG = VSUB(TD, TC);
126 ST(&(xo[WS(os, 12)]), VFNMSI(TG, TF), ovs, &(xo[0]));
127 ST(&(xo[WS(os, 4)]), VFMAI(TG, TF), ovs, &(xo[0]));
128 }
129 {
130 V Tn, Tx, Tw, Ty, Tm, Tv;
131 Tm = VADD(Te, Tl);
132 Tn = VFNMS(LDK(KP707106781), Tm, T7);
133 Tx = VFMA(LDK(KP707106781), Tm, T7);
134 Tv = VSUB(Tl, Te);
135 Tw = VFNMS(LDK(KP707106781), Tv, Tu);
136 Ty = VFMA(LDK(KP707106781), Tv, Tu);
137 ST(&(xo[WS(os, 6)]), VFNMSI(Tw, Tn), ovs, &(xo[0]));
138 ST(&(xo[WS(os, 2)]), VFMAI(Ty, Tx), ovs, &(xo[0]));
139 ST(&(xo[WS(os, 10)]), VFMAI(Tw, Tn), ovs, &(xo[0]));
140 ST(&(xo[WS(os, 14)]), VFNMSI(Ty, Tx), ovs, &(xo[0]));
141 }
142 {
143 V TT, T11, T10, T12;
144 {
145 V TL, TS, TW, TZ;
146 TL = VFMA(LDK(KP707106781), TK, TH);
147 TS = VADD(TO, TR);
148 TT = VFNMS(LDK(KP923879532), TS, TL);
149 T11 = VFMA(LDK(KP923879532), TS, TL);
150 TW = VFNMS(LDK(KP707106781), TV, TU);
151 TZ = VSUB(TX, TY);
152 T10 = VFNMS(LDK(KP923879532), TZ, TW);
153 T12 = VFMA(LDK(KP923879532), TZ, TW);
154 }
155 ST(&(xo[WS(os, 9)]), VFNMSI(T10, TT), ovs, &(xo[WS(os, 1)]));
156 ST(&(xo[WS(os, 15)]), VFMAI(T12, T11), ovs, &(xo[WS(os, 1)]));
157 ST(&(xo[WS(os, 7)]), VFMAI(T10, TT), ovs, &(xo[WS(os, 1)]));
158 ST(&(xo[WS(os, 1)]), VFNMSI(T12, T11), ovs, &(xo[WS(os, 1)]));
159 }
160 {
161 V T15, T19, T18, T1a;
162 {
163 V T13, T14, T16, T17;
164 T13 = VFNMS(LDK(KP707106781), TK, TH);
165 T14 = VADD(TX, TY);
166 T15 = VFNMS(LDK(KP923879532), T14, T13);
167 T19 = VFMA(LDK(KP923879532), T14, T13);
168 T16 = VFMA(LDK(KP707106781), TV, TU);
169 T17 = VSUB(TR, TO);
170 T18 = VFNMS(LDK(KP923879532), T17, T16);
171 T1a = VFMA(LDK(KP923879532), T17, T16);
172 }
173 ST(&(xo[WS(os, 5)]), VFNMSI(T18, T15), ovs, &(xo[WS(os, 1)]));
174 ST(&(xo[WS(os, 13)]), VFNMSI(T1a, T19), ovs, &(xo[WS(os, 1)]));
175 ST(&(xo[WS(os, 11)]), VFMAI(T18, T15), ovs, &(xo[WS(os, 1)]));
176 ST(&(xo[WS(os, 3)]), VFMAI(T1a, T19), ovs, &(xo[WS(os, 1)]));
177 }
178 }
179 }
180 VLEAVE();
181 }
182
183 static const kdft_desc desc = { 16, XSIMD_STRING("n1fv_16"), { 38, 0, 34, 0 }, &GENUS, 0, 0, 0, 0 };
184
XSIMD(codelet_n1fv_16)185 void XSIMD(codelet_n1fv_16) (planner *p) { X(kdft_register) (p, n1fv_16, &desc);
186 }
187
188 #else
189
190 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n1fv_16 -include dft/simd/n1f.h */
191
192 /*
193 * This function contains 72 FP additions, 12 FP multiplications,
194 * (or, 68 additions, 8 multiplications, 4 fused multiply/add),
195 * 30 stack variables, 3 constants, and 32 memory accesses
196 */
197 #include "dft/simd/n1f.h"
198
n1fv_16(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)199 static void n1fv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
200 {
201 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
202 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
203 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
204 {
205 INT i;
206 const R *xi;
207 R *xo;
208 xi = ri;
209 xo = ro;
210 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
211 V Tp, T13, Tu, TN, Tm, T14, Tv, TY, T7, T17, Ty, TT, Te, T16, Tx;
212 V TQ;
213 {
214 V Tn, To, TM, Ts, Tt, TL;
215 Tn = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
216 To = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
217 TM = VADD(Tn, To);
218 Ts = LD(&(xi[0]), ivs, &(xi[0]));
219 Tt = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
220 TL = VADD(Ts, Tt);
221 Tp = VSUB(Tn, To);
222 T13 = VADD(TL, TM);
223 Tu = VSUB(Ts, Tt);
224 TN = VSUB(TL, TM);
225 }
226 {
227 V Ti, TW, Tl, TX;
228 {
229 V Tg, Th, Tj, Tk;
230 Tg = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
231 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
232 Ti = VSUB(Tg, Th);
233 TW = VADD(Tg, Th);
234 Tj = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
235 Tk = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
236 Tl = VSUB(Tj, Tk);
237 TX = VADD(Tj, Tk);
238 }
239 Tm = VMUL(LDK(KP707106781), VSUB(Ti, Tl));
240 T14 = VADD(TX, TW);
241 Tv = VMUL(LDK(KP707106781), VADD(Tl, Ti));
242 TY = VSUB(TW, TX);
243 }
244 {
245 V T3, TR, T6, TS;
246 {
247 V T1, T2, T4, T5;
248 T1 = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
249 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
250 T3 = VSUB(T1, T2);
251 TR = VADD(T1, T2);
252 T4 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
253 T5 = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
254 T6 = VSUB(T4, T5);
255 TS = VADD(T4, T5);
256 }
257 T7 = VFNMS(LDK(KP923879532), T6, VMUL(LDK(KP382683432), T3));
258 T17 = VADD(TR, TS);
259 Ty = VFMA(LDK(KP923879532), T3, VMUL(LDK(KP382683432), T6));
260 TT = VSUB(TR, TS);
261 }
262 {
263 V Ta, TO, Td, TP;
264 {
265 V T8, T9, Tb, Tc;
266 T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
267 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
268 Ta = VSUB(T8, T9);
269 TO = VADD(T8, T9);
270 Tb = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
271 Tc = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
272 Td = VSUB(Tb, Tc);
273 TP = VADD(Tb, Tc);
274 }
275 Te = VFMA(LDK(KP382683432), Ta, VMUL(LDK(KP923879532), Td));
276 T16 = VADD(TO, TP);
277 Tx = VFNMS(LDK(KP382683432), Td, VMUL(LDK(KP923879532), Ta));
278 TQ = VSUB(TO, TP);
279 }
280 {
281 V T15, T18, T19, T1a;
282 T15 = VADD(T13, T14);
283 T18 = VADD(T16, T17);
284 ST(&(xo[WS(os, 8)]), VSUB(T15, T18), ovs, &(xo[0]));
285 ST(&(xo[0]), VADD(T15, T18), ovs, &(xo[0]));
286 T19 = VSUB(T13, T14);
287 T1a = VBYI(VSUB(T17, T16));
288 ST(&(xo[WS(os, 12)]), VSUB(T19, T1a), ovs, &(xo[0]));
289 ST(&(xo[WS(os, 4)]), VADD(T19, T1a), ovs, &(xo[0]));
290 }
291 {
292 V TV, T11, T10, T12, TU, TZ;
293 TU = VMUL(LDK(KP707106781), VADD(TQ, TT));
294 TV = VADD(TN, TU);
295 T11 = VSUB(TN, TU);
296 TZ = VMUL(LDK(KP707106781), VSUB(TT, TQ));
297 T10 = VBYI(VADD(TY, TZ));
298 T12 = VBYI(VSUB(TZ, TY));
299 ST(&(xo[WS(os, 14)]), VSUB(TV, T10), ovs, &(xo[0]));
300 ST(&(xo[WS(os, 6)]), VADD(T11, T12), ovs, &(xo[0]));
301 ST(&(xo[WS(os, 2)]), VADD(TV, T10), ovs, &(xo[0]));
302 ST(&(xo[WS(os, 10)]), VSUB(T11, T12), ovs, &(xo[0]));
303 }
304 {
305 V Tr, TB, TA, TC;
306 {
307 V Tf, Tq, Tw, Tz;
308 Tf = VSUB(T7, Te);
309 Tq = VSUB(Tm, Tp);
310 Tr = VBYI(VSUB(Tf, Tq));
311 TB = VBYI(VADD(Tq, Tf));
312 Tw = VADD(Tu, Tv);
313 Tz = VADD(Tx, Ty);
314 TA = VSUB(Tw, Tz);
315 TC = VADD(Tw, Tz);
316 }
317 ST(&(xo[WS(os, 7)]), VADD(Tr, TA), ovs, &(xo[WS(os, 1)]));
318 ST(&(xo[WS(os, 15)]), VSUB(TC, TB), ovs, &(xo[WS(os, 1)]));
319 ST(&(xo[WS(os, 9)]), VSUB(TA, Tr), ovs, &(xo[WS(os, 1)]));
320 ST(&(xo[WS(os, 1)]), VADD(TB, TC), ovs, &(xo[WS(os, 1)]));
321 }
322 {
323 V TF, TJ, TI, TK;
324 {
325 V TD, TE, TG, TH;
326 TD = VSUB(Tu, Tv);
327 TE = VADD(Te, T7);
328 TF = VADD(TD, TE);
329 TJ = VSUB(TD, TE);
330 TG = VADD(Tp, Tm);
331 TH = VSUB(Ty, Tx);
332 TI = VBYI(VADD(TG, TH));
333 TK = VBYI(VSUB(TH, TG));
334 }
335 ST(&(xo[WS(os, 13)]), VSUB(TF, TI), ovs, &(xo[WS(os, 1)]));
336 ST(&(xo[WS(os, 5)]), VADD(TJ, TK), ovs, &(xo[WS(os, 1)]));
337 ST(&(xo[WS(os, 3)]), VADD(TF, TI), ovs, &(xo[WS(os, 1)]));
338 ST(&(xo[WS(os, 11)]), VSUB(TJ, TK), ovs, &(xo[WS(os, 1)]));
339 }
340 }
341 }
342 VLEAVE();
343 }
344
345 static const kdft_desc desc = { 16, XSIMD_STRING("n1fv_16"), { 68, 8, 4, 0 }, &GENUS, 0, 0, 0, 0 };
346
XSIMD(codelet_n1fv_16)347 void XSIMD(codelet_n1fv_16) (planner *p) { X(kdft_register) (p, n1fv_16, &desc);
348 }
349
350 #endif
351