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 15 -name n1fv_15 -include dft/simd/n1f.h */
29
30 /*
31 * This function contains 78 FP additions, 49 FP multiplications,
32 * (or, 36 additions, 7 multiplications, 42 fused multiply/add),
33 * 53 stack variables, 8 constants, and 30 memory accesses
34 */
35 #include "dft/simd/n1f.h"
36
n1fv_15(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1fv_15(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP910592997, +0.910592997310029334643087372129977886038870291);
40 DVK(KP823639103, +0.823639103546331925877420039278190003029660514);
41 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
42 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
44 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
45 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
46 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
47 {
48 INT i;
49 const R *xi;
50 R *xo;
51 xi = ri;
52 xo = ro;
53 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(30, is), MAKE_VOLATILE_STRIDE(30, os)) {
54 V T5, TX, TB, TO, TU, TV, TR, Ta, Tf, Tg, Tl, Tq, Tr, TE, TH;
55 V TI, T10, T12, T1f, T1g;
56 {
57 V T1, T2, T3, T4;
58 T1 = LD(&(xi[0]), ivs, &(xi[0]));
59 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
60 T3 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
61 T4 = VADD(T2, T3);
62 T5 = VADD(T1, T4);
63 TX = VSUB(T3, T2);
64 TB = VFNMS(LDK(KP500000000), T4, T1);
65 }
66 {
67 V T6, T9, TC, TM, Tm, Tp, TG, TQ, Tb, Te, TD, TN, Th, Tk, TF;
68 V TP, TY, TZ;
69 {
70 V T7, T8, Tn, To;
71 T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
72 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
73 T8 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
74 T9 = VADD(T7, T8);
75 TC = VFNMS(LDK(KP500000000), T9, T6);
76 TM = VSUB(T8, T7);
77 Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
78 Tn = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
79 To = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
80 Tp = VADD(Tn, To);
81 TG = VFNMS(LDK(KP500000000), Tp, Tm);
82 TQ = VSUB(To, Tn);
83 }
84 {
85 V Tc, Td, Ti, Tj;
86 Tb = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
87 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
88 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
89 Te = VADD(Tc, Td);
90 TD = VFNMS(LDK(KP500000000), Te, Tb);
91 TN = VSUB(Td, Tc);
92 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
93 Ti = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
94 Tj = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
95 Tk = VADD(Ti, Tj);
96 TF = VFNMS(LDK(KP500000000), Tk, Th);
97 TP = VSUB(Tj, Ti);
98 }
99 TO = VSUB(TM, TN);
100 TU = VSUB(TC, TD);
101 TV = VSUB(TF, TG);
102 TR = VSUB(TP, TQ);
103 Ta = VADD(T6, T9);
104 Tf = VADD(Tb, Te);
105 Tg = VADD(Ta, Tf);
106 Tl = VADD(Th, Tk);
107 Tq = VADD(Tm, Tp);
108 Tr = VADD(Tl, Tq);
109 TE = VADD(TC, TD);
110 TH = VADD(TF, TG);
111 TI = VADD(TE, TH);
112 TY = VADD(TM, TN);
113 TZ = VADD(TP, TQ);
114 T10 = VADD(TY, TZ);
115 T12 = VSUB(TY, TZ);
116 }
117 T1f = VADD(TB, TI);
118 T1g = VMUL(LDK(KP866025403), VADD(TX, T10));
119 ST(&(xo[WS(os, 5)]), VFNMSI(T1g, T1f), ovs, &(xo[WS(os, 1)]));
120 ST(&(xo[WS(os, 10)]), VFMAI(T1g, T1f), ovs, &(xo[0]));
121 {
122 V Tu, Ts, Tt, Ty, TA, Tw, Tx, Tz, Tv;
123 Tu = VSUB(Tg, Tr);
124 Ts = VADD(Tg, Tr);
125 Tt = VFNMS(LDK(KP250000000), Ts, T5);
126 Tw = VSUB(Tl, Tq);
127 Tx = VSUB(Ta, Tf);
128 Ty = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tx, Tw));
129 TA = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tw, Tx));
130 ST(&(xo[0]), VADD(T5, Ts), ovs, &(xo[0]));
131 Tz = VFMA(LDK(KP559016994), Tu, Tt);
132 ST(&(xo[WS(os, 6)]), VFNMSI(TA, Tz), ovs, &(xo[0]));
133 ST(&(xo[WS(os, 9)]), VFMAI(TA, Tz), ovs, &(xo[WS(os, 1)]));
134 Tv = VFNMS(LDK(KP559016994), Tu, Tt);
135 ST(&(xo[WS(os, 3)]), VFNMSI(Ty, Tv), ovs, &(xo[WS(os, 1)]));
136 ST(&(xo[WS(os, 12)]), VFMAI(Ty, Tv), ovs, &(xo[0]));
137 }
138 {
139 V TS, TW, T1a, T18, T13, T1b, TL, T17, T11, TJ, TK;
140 TS = VFMA(LDK(KP618033988), TR, TO);
141 TW = VFMA(LDK(KP618033988), TV, TU);
142 T1a = VFNMS(LDK(KP618033988), TU, TV);
143 T18 = VFNMS(LDK(KP618033988), TO, TR);
144 T11 = VFNMS(LDK(KP250000000), T10, TX);
145 T13 = VFMA(LDK(KP559016994), T12, T11);
146 T1b = VFNMS(LDK(KP559016994), T12, T11);
147 TJ = VFNMS(LDK(KP250000000), TI, TB);
148 TK = VSUB(TE, TH);
149 TL = VFMA(LDK(KP559016994), TK, TJ);
150 T17 = VFNMS(LDK(KP559016994), TK, TJ);
151 {
152 V TT, T14, T1d, T1e;
153 TT = VFMA(LDK(KP823639103), TS, TL);
154 T14 = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T13, TW));
155 ST(&(xo[WS(os, 1)]), VFNMSI(T14, TT), ovs, &(xo[WS(os, 1)]));
156 ST(&(xo[WS(os, 14)]), VFMAI(T14, TT), ovs, &(xo[0]));
157 T1d = VFNMS(LDK(KP823639103), T18, T17);
158 T1e = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1b, T1a));
159 ST(&(xo[WS(os, 8)]), VFNMSI(T1e, T1d), ovs, &(xo[0]));
160 ST(&(xo[WS(os, 7)]), VFMAI(T1e, T1d), ovs, &(xo[WS(os, 1)]));
161 }
162 {
163 V T15, T16, T19, T1c;
164 T15 = VFNMS(LDK(KP823639103), TS, TL);
165 T16 = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T13, TW));
166 ST(&(xo[WS(os, 11)]), VFNMSI(T16, T15), ovs, &(xo[WS(os, 1)]));
167 ST(&(xo[WS(os, 4)]), VFMAI(T16, T15), ovs, &(xo[0]));
168 T19 = VFMA(LDK(KP823639103), T18, T17);
169 T1c = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1b, T1a));
170 ST(&(xo[WS(os, 13)]), VFNMSI(T1c, T19), ovs, &(xo[WS(os, 1)]));
171 ST(&(xo[WS(os, 2)]), VFMAI(T1c, T19), ovs, &(xo[0]));
172 }
173 }
174 }
175 }
176 VLEAVE();
177 }
178
179 static const kdft_desc desc = { 15, XSIMD_STRING("n1fv_15"), { 36, 7, 42, 0 }, &GENUS, 0, 0, 0, 0 };
180
XSIMD(codelet_n1fv_15)181 void XSIMD(codelet_n1fv_15) (planner *p) { X(kdft_register) (p, n1fv_15, &desc);
182 }
183
184 #else
185
186 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 15 -name n1fv_15 -include dft/simd/n1f.h */
187
188 /*
189 * This function contains 78 FP additions, 25 FP multiplications,
190 * (or, 64 additions, 11 multiplications, 14 fused multiply/add),
191 * 55 stack variables, 10 constants, and 30 memory accesses
192 */
193 #include "dft/simd/n1f.h"
194
n1fv_15(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)195 static void n1fv_15(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
196 {
197 DVK(KP216506350, +0.216506350946109661690930792688234045867850657);
198 DVK(KP509036960, +0.509036960455127183450980863393907648510733164);
199 DVK(KP823639103, +0.823639103546331925877420039278190003029660514);
200 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
201 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
202 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
203 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
204 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
205 DVK(KP484122918, +0.484122918275927110647408174972799951354115213);
206 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
207 {
208 INT i;
209 const R *xi;
210 R *xo;
211 xi = ri;
212 xo = ro;
213 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(30, is), MAKE_VOLATILE_STRIDE(30, os)) {
214 V T5, T10, TB, TO, TU, TV, TR, Ta, Tf, Tg, Tl, Tq, Tr, TE, TH;
215 V TI, TZ, T11, T1f, T1g;
216 {
217 V T1, T2, T3, T4;
218 T1 = LD(&(xi[0]), ivs, &(xi[0]));
219 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
220 T3 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
221 T4 = VADD(T2, T3);
222 T5 = VADD(T1, T4);
223 T10 = VSUB(T3, T2);
224 TB = VFNMS(LDK(KP500000000), T4, T1);
225 }
226 {
227 V T6, T9, TC, TP, Tm, Tp, TG, TN, Tb, Te, TD, TQ, Th, Tk, TF;
228 V TM, TX, TY;
229 {
230 V T7, T8, Tn, To;
231 T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
232 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
233 T8 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
234 T9 = VADD(T7, T8);
235 TC = VFNMS(LDK(KP500000000), T9, T6);
236 TP = VSUB(T8, T7);
237 Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
238 Tn = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
239 To = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
240 Tp = VADD(Tn, To);
241 TG = VFNMS(LDK(KP500000000), Tp, Tm);
242 TN = VSUB(To, Tn);
243 }
244 {
245 V Tc, Td, Ti, Tj;
246 Tb = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
247 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
248 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
249 Te = VADD(Tc, Td);
250 TD = VFNMS(LDK(KP500000000), Te, Tb);
251 TQ = VSUB(Td, Tc);
252 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
253 Ti = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
254 Tj = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
255 Tk = VADD(Ti, Tj);
256 TF = VFNMS(LDK(KP500000000), Tk, Th);
257 TM = VSUB(Tj, Ti);
258 }
259 TO = VSUB(TM, TN);
260 TU = VSUB(TF, TG);
261 TV = VSUB(TC, TD);
262 TR = VSUB(TP, TQ);
263 Ta = VADD(T6, T9);
264 Tf = VADD(Tb, Te);
265 Tg = VADD(Ta, Tf);
266 Tl = VADD(Th, Tk);
267 Tq = VADD(Tm, Tp);
268 Tr = VADD(Tl, Tq);
269 TE = VADD(TC, TD);
270 TH = VADD(TF, TG);
271 TI = VADD(TE, TH);
272 TX = VADD(TP, TQ);
273 TY = VADD(TM, TN);
274 TZ = VMUL(LDK(KP484122918), VSUB(TX, TY));
275 T11 = VADD(TX, TY);
276 }
277 T1f = VADD(TB, TI);
278 T1g = VBYI(VMUL(LDK(KP866025403), VADD(T10, T11)));
279 ST(&(xo[WS(os, 5)]), VSUB(T1f, T1g), ovs, &(xo[WS(os, 1)]));
280 ST(&(xo[WS(os, 10)]), VADD(T1f, T1g), ovs, &(xo[0]));
281 {
282 V Tu, Ts, Tt, Ty, TA, Tw, Tx, Tz, Tv;
283 Tu = VMUL(LDK(KP559016994), VSUB(Tg, Tr));
284 Ts = VADD(Tg, Tr);
285 Tt = VFNMS(LDK(KP250000000), Ts, T5);
286 Tw = VSUB(Tl, Tq);
287 Tx = VSUB(Ta, Tf);
288 Ty = VBYI(VFNMS(LDK(KP587785252), Tx, VMUL(LDK(KP951056516), Tw)));
289 TA = VBYI(VFMA(LDK(KP951056516), Tx, VMUL(LDK(KP587785252), Tw)));
290 ST(&(xo[0]), VADD(T5, Ts), ovs, &(xo[0]));
291 Tz = VADD(Tu, Tt);
292 ST(&(xo[WS(os, 6)]), VSUB(Tz, TA), ovs, &(xo[0]));
293 ST(&(xo[WS(os, 9)]), VADD(TA, Tz), ovs, &(xo[WS(os, 1)]));
294 Tv = VSUB(Tt, Tu);
295 ST(&(xo[WS(os, 3)]), VSUB(Tv, Ty), ovs, &(xo[WS(os, 1)]));
296 ST(&(xo[WS(os, 12)]), VADD(Ty, Tv), ovs, &(xo[0]));
297 }
298 {
299 V TS, TW, T1b, T18, T13, T1a, TL, T17, T12, TJ, TK;
300 TS = VFNMS(LDK(KP509036960), TR, VMUL(LDK(KP823639103), TO));
301 TW = VFNMS(LDK(KP587785252), TV, VMUL(LDK(KP951056516), TU));
302 T1b = VFMA(LDK(KP951056516), TV, VMUL(LDK(KP587785252), TU));
303 T18 = VFMA(LDK(KP823639103), TR, VMUL(LDK(KP509036960), TO));
304 T12 = VFNMS(LDK(KP216506350), T11, VMUL(LDK(KP866025403), T10));
305 T13 = VSUB(TZ, T12);
306 T1a = VADD(TZ, T12);
307 TJ = VFNMS(LDK(KP250000000), TI, TB);
308 TK = VMUL(LDK(KP559016994), VSUB(TE, TH));
309 TL = VSUB(TJ, TK);
310 T17 = VADD(TK, TJ);
311 {
312 V TT, T14, T1d, T1e;
313 TT = VSUB(TL, TS);
314 T14 = VBYI(VSUB(TW, T13));
315 ST(&(xo[WS(os, 8)]), VSUB(TT, T14), ovs, &(xo[0]));
316 ST(&(xo[WS(os, 7)]), VADD(TT, T14), ovs, &(xo[WS(os, 1)]));
317 T1d = VSUB(T17, T18);
318 T1e = VBYI(VADD(T1b, T1a));
319 ST(&(xo[WS(os, 11)]), VSUB(T1d, T1e), ovs, &(xo[WS(os, 1)]));
320 ST(&(xo[WS(os, 4)]), VADD(T1d, T1e), ovs, &(xo[0]));
321 }
322 {
323 V T15, T16, T19, T1c;
324 T15 = VADD(TL, TS);
325 T16 = VBYI(VADD(TW, T13));
326 ST(&(xo[WS(os, 13)]), VSUB(T15, T16), ovs, &(xo[WS(os, 1)]));
327 ST(&(xo[WS(os, 2)]), VADD(T15, T16), ovs, &(xo[0]));
328 T19 = VADD(T17, T18);
329 T1c = VBYI(VSUB(T1a, T1b));
330 ST(&(xo[WS(os, 14)]), VSUB(T19, T1c), ovs, &(xo[0]));
331 ST(&(xo[WS(os, 1)]), VADD(T19, T1c), ovs, &(xo[WS(os, 1)]));
332 }
333 }
334 }
335 }
336 VLEAVE();
337 }
338
339 static const kdft_desc desc = { 15, XSIMD_STRING("n1fv_15"), { 64, 11, 14, 0 }, &GENUS, 0, 0, 0, 0 };
340
XSIMD(codelet_n1fv_15)341 void XSIMD(codelet_n1fv_15) (planner *p) { X(kdft_register) (p, n1fv_15, &desc);
342 }
343
344 #endif
345