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 07:04:08 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.native -fma -compact -variables 4 -pipeline-latency 4 -n 9 -name n1_9 -include dft/scalar/n.h */
29
30 /*
31 * This function contains 80 FP additions, 56 FP multiplications,
32 * (or, 24 additions, 0 multiplications, 56 fused multiply/add),
33 * 41 stack variables, 10 constants, and 36 memory accesses
34 */
35 #include "dft/scalar/n.h"
36
n1_9(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP954188894, +0.954188894138671133499268364187245676532219158);
40 DK(KP363970234, +0.363970234266202361351047882776834043890471784);
41 DK(KP852868531, +0.852868531952443209628250963940074071936020296);
42 DK(KP492403876, +0.492403876506104029683371512294761506835321626);
43 DK(KP984807753, +0.984807753012208059366743024589523013670643252);
44 DK(KP777861913, +0.777861913430206160028177977318626690410586096);
45 DK(KP839099631, +0.839099631177280011763127298123181364687434283);
46 DK(KP176326980, +0.176326980708464973471090386868618986121633062);
47 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
48 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
49 {
50 INT i;
51 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(36, is), MAKE_VOLATILE_STRIDE(36, os)) {
52 E T5, TL, Tm, Tl, T1f, TM, Ta, T1c, TF, TW, TI, TX, Tf, T1d, Ts;
53 E TZ, Tx, T10;
54 {
55 E T1, T2, T3, T4;
56 T1 = ri[0];
57 T2 = ri[WS(is, 3)];
58 T3 = ri[WS(is, 6)];
59 T4 = T2 + T3;
60 T5 = T1 + T4;
61 TL = FNMS(KP500000000, T4, T1);
62 Tm = T3 - T2;
63 }
64 {
65 E Th, Ti, Tj, Tk;
66 Th = ii[0];
67 Ti = ii[WS(is, 3)];
68 Tj = ii[WS(is, 6)];
69 Tk = Ti + Tj;
70 Tl = FNMS(KP500000000, Tk, Th);
71 T1f = Th + Tk;
72 TM = Ti - Tj;
73 }
74 {
75 E T6, Tz, T9, TE, TC, TH, TD, TG;
76 T6 = ri[WS(is, 1)];
77 Tz = ii[WS(is, 1)];
78 {
79 E T7, T8, TA, TB;
80 T7 = ri[WS(is, 4)];
81 T8 = ri[WS(is, 7)];
82 T9 = T7 + T8;
83 TE = T7 - T8;
84 TA = ii[WS(is, 4)];
85 TB = ii[WS(is, 7)];
86 TC = TA + TB;
87 TH = TB - TA;
88 }
89 Ta = T6 + T9;
90 T1c = Tz + TC;
91 TD = FNMS(KP500000000, TC, Tz);
92 TF = FNMS(KP866025403, TE, TD);
93 TW = FMA(KP866025403, TE, TD);
94 TG = FNMS(KP500000000, T9, T6);
95 TI = FNMS(KP866025403, TH, TG);
96 TX = FMA(KP866025403, TH, TG);
97 }
98 {
99 E Tb, Tt, Te, Tw, Tr, Tu, To, Tv;
100 Tb = ri[WS(is, 2)];
101 Tt = ii[WS(is, 2)];
102 {
103 E Tc, Td, Tp, Tq;
104 Tc = ri[WS(is, 5)];
105 Td = ri[WS(is, 8)];
106 Te = Tc + Td;
107 Tw = Td - Tc;
108 Tp = ii[WS(is, 5)];
109 Tq = ii[WS(is, 8)];
110 Tr = Tp - Tq;
111 Tu = Tp + Tq;
112 }
113 Tf = Tb + Te;
114 T1d = Tt + Tu;
115 To = FNMS(KP500000000, Te, Tb);
116 Ts = FMA(KP866025403, Tr, To);
117 TZ = FNMS(KP866025403, Tr, To);
118 Tv = FNMS(KP500000000, Tu, Tt);
119 Tx = FMA(KP866025403, Tw, Tv);
120 T10 = FNMS(KP866025403, Tw, Tv);
121 }
122 {
123 E T1e, Tg, T1b, T1i, T1g, T1h;
124 T1e = T1c - T1d;
125 Tg = Ta + Tf;
126 T1b = FNMS(KP500000000, Tg, T5);
127 ro[0] = T5 + Tg;
128 ro[WS(os, 3)] = FMA(KP866025403, T1e, T1b);
129 ro[WS(os, 6)] = FNMS(KP866025403, T1e, T1b);
130 T1i = Tf - Ta;
131 T1g = T1c + T1d;
132 T1h = FNMS(KP500000000, T1g, T1f);
133 io[WS(os, 3)] = FMA(KP866025403, T1i, T1h);
134 io[0] = T1f + T1g;
135 io[WS(os, 6)] = FNMS(KP866025403, T1i, T1h);
136 }
137 {
138 E Tn, TN, TK, TS, TQ, TU, TR, TT;
139 Tn = FMA(KP866025403, Tm, Tl);
140 TN = FMA(KP866025403, TM, TL);
141 {
142 E Ty, TJ, TO, TP;
143 Ty = FNMS(KP176326980, Tx, Ts);
144 TJ = FNMS(KP839099631, TI, TF);
145 TK = FNMS(KP777861913, TJ, Ty);
146 TS = FMA(KP777861913, TJ, Ty);
147 TO = FMA(KP176326980, Ts, Tx);
148 TP = FMA(KP839099631, TF, TI);
149 TQ = FMA(KP777861913, TP, TO);
150 TU = FNMS(KP777861913, TP, TO);
151 }
152 io[WS(os, 1)] = FNMS(KP984807753, TK, Tn);
153 ro[WS(os, 1)] = FMA(KP984807753, TQ, TN);
154 TR = FNMS(KP492403876, TQ, TN);
155 ro[WS(os, 4)] = FMA(KP852868531, TS, TR);
156 ro[WS(os, 7)] = FNMS(KP852868531, TS, TR);
157 TT = FMA(KP492403876, TK, Tn);
158 io[WS(os, 7)] = FNMS(KP852868531, TU, TT);
159 io[WS(os, 4)] = FMA(KP852868531, TU, TT);
160 }
161 {
162 E TV, T17, T12, T1a, T16, T18, T13, T19;
163 TV = FNMS(KP866025403, TM, TL);
164 T17 = FNMS(KP866025403, Tm, Tl);
165 {
166 E TY, T11, T14, T15;
167 TY = FMA(KP176326980, TX, TW);
168 T11 = FNMS(KP363970234, T10, TZ);
169 T12 = FNMS(KP954188894, T11, TY);
170 T1a = FMA(KP954188894, T11, TY);
171 T14 = FNMS(KP176326980, TW, TX);
172 T15 = FMA(KP363970234, TZ, T10);
173 T16 = FNMS(KP954188894, T15, T14);
174 T18 = FMA(KP954188894, T15, T14);
175 }
176 ro[WS(os, 2)] = FMA(KP984807753, T12, TV);
177 io[WS(os, 2)] = FNMS(KP984807753, T18, T17);
178 T13 = FNMS(KP492403876, T12, TV);
179 ro[WS(os, 5)] = FNMS(KP852868531, T16, T13);
180 ro[WS(os, 8)] = FMA(KP852868531, T16, T13);
181 T19 = FMA(KP492403876, T18, T17);
182 io[WS(os, 5)] = FNMS(KP852868531, T1a, T19);
183 io[WS(os, 8)] = FMA(KP852868531, T1a, T19);
184 }
185 }
186 }
187 }
188
189 static const kdft_desc desc = { 9, "n1_9", { 24, 0, 56, 0 }, &GENUS, 0, 0, 0, 0 };
190
X(codelet_n1_9)191 void X(codelet_n1_9) (planner *p) { X(kdft_register) (p, n1_9, &desc);
192 }
193
194 #else
195
196 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 9 -name n1_9 -include dft/scalar/n.h */
197
198 /*
199 * This function contains 80 FP additions, 40 FP multiplications,
200 * (or, 60 additions, 20 multiplications, 20 fused multiply/add),
201 * 39 stack variables, 8 constants, and 36 memory accesses
202 */
203 #include "dft/scalar/n.h"
204
n1_9(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)205 static void n1_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
206 {
207 DK(KP939692620, +0.939692620785908384054109277324731469936208134);
208 DK(KP342020143, +0.342020143325668733044099614682259580763083368);
209 DK(KP984807753, +0.984807753012208059366743024589523013670643252);
210 DK(KP173648177, +0.173648177666930348851716626769314796000375677);
211 DK(KP642787609, +0.642787609686539326322643409907263432907559884);
212 DK(KP766044443, +0.766044443118978035202392650555416673935832457);
213 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
214 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
215 {
216 INT i;
217 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(36, is), MAKE_VOLATILE_STRIDE(36, os)) {
218 E T5, TO, Th, Tk, T1g, TR, Ta, T1c, Tq, TW, Tv, TX, Tf, T1d, TB;
219 E T10, TG, TZ;
220 {
221 E T1, T2, T3, T4;
222 T1 = ri[0];
223 T2 = ri[WS(is, 3)];
224 T3 = ri[WS(is, 6)];
225 T4 = T2 + T3;
226 T5 = T1 + T4;
227 TO = KP866025403 * (T3 - T2);
228 Th = FNMS(KP500000000, T4, T1);
229 }
230 {
231 E TP, Ti, Tj, TQ;
232 TP = ii[0];
233 Ti = ii[WS(is, 3)];
234 Tj = ii[WS(is, 6)];
235 TQ = Ti + Tj;
236 Tk = KP866025403 * (Ti - Tj);
237 T1g = TP + TQ;
238 TR = FNMS(KP500000000, TQ, TP);
239 }
240 {
241 E T6, Ts, T9, Tr, Tp, Tt, Tm, Tu;
242 T6 = ri[WS(is, 1)];
243 Ts = ii[WS(is, 1)];
244 {
245 E T7, T8, Tn, To;
246 T7 = ri[WS(is, 4)];
247 T8 = ri[WS(is, 7)];
248 T9 = T7 + T8;
249 Tr = KP866025403 * (T8 - T7);
250 Tn = ii[WS(is, 4)];
251 To = ii[WS(is, 7)];
252 Tp = KP866025403 * (Tn - To);
253 Tt = Tn + To;
254 }
255 Ta = T6 + T9;
256 T1c = Ts + Tt;
257 Tm = FNMS(KP500000000, T9, T6);
258 Tq = Tm + Tp;
259 TW = Tm - Tp;
260 Tu = FNMS(KP500000000, Tt, Ts);
261 Tv = Tr + Tu;
262 TX = Tu - Tr;
263 }
264 {
265 E Tb, TD, Te, TC, TA, TE, Tx, TF;
266 Tb = ri[WS(is, 2)];
267 TD = ii[WS(is, 2)];
268 {
269 E Tc, Td, Ty, Tz;
270 Tc = ri[WS(is, 5)];
271 Td = ri[WS(is, 8)];
272 Te = Tc + Td;
273 TC = KP866025403 * (Td - Tc);
274 Ty = ii[WS(is, 5)];
275 Tz = ii[WS(is, 8)];
276 TA = KP866025403 * (Ty - Tz);
277 TE = Ty + Tz;
278 }
279 Tf = Tb + Te;
280 T1d = TD + TE;
281 Tx = FNMS(KP500000000, Te, Tb);
282 TB = Tx + TA;
283 T10 = Tx - TA;
284 TF = FNMS(KP500000000, TE, TD);
285 TG = TC + TF;
286 TZ = TF - TC;
287 }
288 {
289 E T1e, Tg, T1b, T1f, T1h, T1i;
290 T1e = KP866025403 * (T1c - T1d);
291 Tg = Ta + Tf;
292 T1b = FNMS(KP500000000, Tg, T5);
293 ro[0] = T5 + Tg;
294 ro[WS(os, 3)] = T1b + T1e;
295 ro[WS(os, 6)] = T1b - T1e;
296 T1f = KP866025403 * (Tf - Ta);
297 T1h = T1c + T1d;
298 T1i = FNMS(KP500000000, T1h, T1g);
299 io[WS(os, 3)] = T1f + T1i;
300 io[0] = T1g + T1h;
301 io[WS(os, 6)] = T1i - T1f;
302 }
303 {
304 E Tl, TS, TI, TN, TM, TT, TJ, TU;
305 Tl = Th + Tk;
306 TS = TO + TR;
307 {
308 E Tw, TH, TK, TL;
309 Tw = FMA(KP766044443, Tq, KP642787609 * Tv);
310 TH = FMA(KP173648177, TB, KP984807753 * TG);
311 TI = Tw + TH;
312 TN = KP866025403 * (TH - Tw);
313 TK = FNMS(KP642787609, Tq, KP766044443 * Tv);
314 TL = FNMS(KP984807753, TB, KP173648177 * TG);
315 TM = KP866025403 * (TK - TL);
316 TT = TK + TL;
317 }
318 ro[WS(os, 1)] = Tl + TI;
319 io[WS(os, 1)] = TS + TT;
320 TJ = FNMS(KP500000000, TI, Tl);
321 ro[WS(os, 7)] = TJ - TM;
322 ro[WS(os, 4)] = TJ + TM;
323 TU = FNMS(KP500000000, TT, TS);
324 io[WS(os, 4)] = TN + TU;
325 io[WS(os, 7)] = TU - TN;
326 }
327 {
328 E TV, T14, T12, T13, T17, T1a, T18, T19;
329 TV = Th - Tk;
330 T14 = TR - TO;
331 {
332 E TY, T11, T15, T16;
333 TY = FMA(KP173648177, TW, KP984807753 * TX);
334 T11 = FNMS(KP939692620, T10, KP342020143 * TZ);
335 T12 = TY + T11;
336 T13 = KP866025403 * (T11 - TY);
337 T15 = FNMS(KP984807753, TW, KP173648177 * TX);
338 T16 = FMA(KP342020143, T10, KP939692620 * TZ);
339 T17 = T15 - T16;
340 T1a = KP866025403 * (T15 + T16);
341 }
342 ro[WS(os, 2)] = TV + T12;
343 io[WS(os, 2)] = T14 + T17;
344 T18 = FNMS(KP500000000, T17, T14);
345 io[WS(os, 5)] = T13 + T18;
346 io[WS(os, 8)] = T18 - T13;
347 T19 = FNMS(KP500000000, T12, TV);
348 ro[WS(os, 8)] = T19 - T1a;
349 ro[WS(os, 5)] = T19 + T1a;
350 }
351 }
352 }
353 }
354
355 static const kdft_desc desc = { 9, "n1_9", { 60, 20, 20, 0 }, &GENUS, 0, 0, 0, 0 };
356
X(codelet_n1_9)357 void X(codelet_n1_9) (planner *p) { X(kdft_register) (p, n1_9, &desc);
358 }
359
360 #endif
361