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 10 -name n1_10 -include dft/scalar/n.h */
29
30 /*
31 * This function contains 84 FP additions, 36 FP multiplications,
32 * (or, 48 additions, 0 multiplications, 36 fused multiply/add),
33 * 41 stack variables, 4 constants, and 40 memory accesses
34 */
35 #include "dft/scalar/n.h"
36
n1_10(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
41 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
42 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 {
44 INT i;
45 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) {
46 E T3, Tj, TN, T1b, TU, TV, T1j, T1i, Tm, Tp, Tq, Ta, Th, Ti, TA;
47 E TH, T17, T14, T1c, T1d, T1e, TO, TP, TQ;
48 {
49 E T1, T2, TL, TM;
50 T1 = ri[0];
51 T2 = ri[WS(is, 5)];
52 T3 = T1 - T2;
53 Tj = T1 + T2;
54 TL = ii[0];
55 TM = ii[WS(is, 5)];
56 TN = TL - TM;
57 T1b = TL + TM;
58 }
59 {
60 E T6, Tk, Tg, To, T9, Tl, Td, Tn;
61 {
62 E T4, T5, Te, Tf;
63 T4 = ri[WS(is, 2)];
64 T5 = ri[WS(is, 7)];
65 T6 = T4 - T5;
66 Tk = T4 + T5;
67 Te = ri[WS(is, 6)];
68 Tf = ri[WS(is, 1)];
69 Tg = Te - Tf;
70 To = Te + Tf;
71 }
72 {
73 E T7, T8, Tb, Tc;
74 T7 = ri[WS(is, 8)];
75 T8 = ri[WS(is, 3)];
76 T9 = T7 - T8;
77 Tl = T7 + T8;
78 Tb = ri[WS(is, 4)];
79 Tc = ri[WS(is, 9)];
80 Td = Tb - Tc;
81 Tn = Tb + Tc;
82 }
83 TU = T6 - T9;
84 TV = Td - Tg;
85 T1j = Tk - Tl;
86 T1i = Tn - To;
87 Tm = Tk + Tl;
88 Tp = Tn + To;
89 Tq = Tm + Tp;
90 Ta = T6 + T9;
91 Th = Td + Tg;
92 Ti = Ta + Th;
93 }
94 {
95 E Tw, T15, TG, T13, Tz, T16, TD, T12;
96 {
97 E Tu, Tv, TE, TF;
98 Tu = ii[WS(is, 2)];
99 Tv = ii[WS(is, 7)];
100 Tw = Tu - Tv;
101 T15 = Tu + Tv;
102 TE = ii[WS(is, 6)];
103 TF = ii[WS(is, 1)];
104 TG = TE - TF;
105 T13 = TE + TF;
106 }
107 {
108 E Tx, Ty, TB, TC;
109 Tx = ii[WS(is, 8)];
110 Ty = ii[WS(is, 3)];
111 Tz = Tx - Ty;
112 T16 = Tx + Ty;
113 TB = ii[WS(is, 4)];
114 TC = ii[WS(is, 9)];
115 TD = TB - TC;
116 T12 = TB + TC;
117 }
118 TA = Tw - Tz;
119 TH = TD - TG;
120 T17 = T15 - T16;
121 T14 = T12 - T13;
122 T1c = T15 + T16;
123 T1d = T12 + T13;
124 T1e = T1c + T1d;
125 TO = Tw + Tz;
126 TP = TD + TG;
127 TQ = TO + TP;
128 }
129 ro[WS(os, 5)] = T3 + Ti;
130 io[WS(os, 5)] = TN + TQ;
131 ro[0] = Tj + Tq;
132 io[0] = T1b + T1e;
133 {
134 E TI, TK, Tt, TJ, Tr, Ts;
135 TI = FMA(KP618033988, TH, TA);
136 TK = FNMS(KP618033988, TA, TH);
137 Tr = FNMS(KP250000000, Ti, T3);
138 Ts = Ta - Th;
139 Tt = FMA(KP559016994, Ts, Tr);
140 TJ = FNMS(KP559016994, Ts, Tr);
141 ro[WS(os, 9)] = FNMS(KP951056516, TI, Tt);
142 ro[WS(os, 3)] = FMA(KP951056516, TK, TJ);
143 ro[WS(os, 1)] = FMA(KP951056516, TI, Tt);
144 ro[WS(os, 7)] = FNMS(KP951056516, TK, TJ);
145 }
146 {
147 E TW, TY, TT, TX, TR, TS;
148 TW = FMA(KP618033988, TV, TU);
149 TY = FNMS(KP618033988, TU, TV);
150 TR = FNMS(KP250000000, TQ, TN);
151 TS = TO - TP;
152 TT = FMA(KP559016994, TS, TR);
153 TX = FNMS(KP559016994, TS, TR);
154 io[WS(os, 1)] = FNMS(KP951056516, TW, TT);
155 io[WS(os, 7)] = FMA(KP951056516, TY, TX);
156 io[WS(os, 9)] = FMA(KP951056516, TW, TT);
157 io[WS(os, 3)] = FNMS(KP951056516, TY, TX);
158 }
159 {
160 E T18, T1a, T11, T19, TZ, T10;
161 T18 = FNMS(KP618033988, T17, T14);
162 T1a = FMA(KP618033988, T14, T17);
163 TZ = FNMS(KP250000000, Tq, Tj);
164 T10 = Tm - Tp;
165 T11 = FNMS(KP559016994, T10, TZ);
166 T19 = FMA(KP559016994, T10, TZ);
167 ro[WS(os, 2)] = FNMS(KP951056516, T18, T11);
168 ro[WS(os, 6)] = FMA(KP951056516, T1a, T19);
169 ro[WS(os, 8)] = FMA(KP951056516, T18, T11);
170 ro[WS(os, 4)] = FNMS(KP951056516, T1a, T19);
171 }
172 {
173 E T1k, T1m, T1h, T1l, T1f, T1g;
174 T1k = FNMS(KP618033988, T1j, T1i);
175 T1m = FMA(KP618033988, T1i, T1j);
176 T1f = FNMS(KP250000000, T1e, T1b);
177 T1g = T1c - T1d;
178 T1h = FNMS(KP559016994, T1g, T1f);
179 T1l = FMA(KP559016994, T1g, T1f);
180 io[WS(os, 2)] = FMA(KP951056516, T1k, T1h);
181 io[WS(os, 6)] = FNMS(KP951056516, T1m, T1l);
182 io[WS(os, 8)] = FNMS(KP951056516, T1k, T1h);
183 io[WS(os, 4)] = FMA(KP951056516, T1m, T1l);
184 }
185 }
186 }
187 }
188
189 static const kdft_desc desc = { 10, "n1_10", { 48, 0, 36, 0 }, &GENUS, 0, 0, 0, 0 };
190
X(codelet_n1_10)191 void X(codelet_n1_10) (planner *p) { X(kdft_register) (p, n1_10, &desc);
192 }
193
194 #else
195
196 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 10 -name n1_10 -include dft/scalar/n.h */
197
198 /*
199 * This function contains 84 FP additions, 24 FP multiplications,
200 * (or, 72 additions, 12 multiplications, 12 fused multiply/add),
201 * 41 stack variables, 4 constants, and 40 memory accesses
202 */
203 #include "dft/scalar/n.h"
204
n1_10(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)205 static void n1_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
206 {
207 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
208 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
209 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
210 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
211 {
212 INT i;
213 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) {
214 E T3, Tj, TQ, T1e, TU, TV, T1c, T1b, Tm, Tp, Tq, Ta, Th, Ti, TA;
215 E TH, T17, T14, T1f, T1g, T1h, TL, TM, TR;
216 {
217 E T1, T2, TO, TP;
218 T1 = ri[0];
219 T2 = ri[WS(is, 5)];
220 T3 = T1 - T2;
221 Tj = T1 + T2;
222 TO = ii[0];
223 TP = ii[WS(is, 5)];
224 TQ = TO - TP;
225 T1e = TO + TP;
226 }
227 {
228 E T6, Tk, Tg, To, T9, Tl, Td, Tn;
229 {
230 E T4, T5, Te, Tf;
231 T4 = ri[WS(is, 2)];
232 T5 = ri[WS(is, 7)];
233 T6 = T4 - T5;
234 Tk = T4 + T5;
235 Te = ri[WS(is, 6)];
236 Tf = ri[WS(is, 1)];
237 Tg = Te - Tf;
238 To = Te + Tf;
239 }
240 {
241 E T7, T8, Tb, Tc;
242 T7 = ri[WS(is, 8)];
243 T8 = ri[WS(is, 3)];
244 T9 = T7 - T8;
245 Tl = T7 + T8;
246 Tb = ri[WS(is, 4)];
247 Tc = ri[WS(is, 9)];
248 Td = Tb - Tc;
249 Tn = Tb + Tc;
250 }
251 TU = T6 - T9;
252 TV = Td - Tg;
253 T1c = Tk - Tl;
254 T1b = Tn - To;
255 Tm = Tk + Tl;
256 Tp = Tn + To;
257 Tq = Tm + Tp;
258 Ta = T6 + T9;
259 Th = Td + Tg;
260 Ti = Ta + Th;
261 }
262 {
263 E Tw, T15, TG, T13, Tz, T16, TD, T12;
264 {
265 E Tu, Tv, TE, TF;
266 Tu = ii[WS(is, 2)];
267 Tv = ii[WS(is, 7)];
268 Tw = Tu - Tv;
269 T15 = Tu + Tv;
270 TE = ii[WS(is, 6)];
271 TF = ii[WS(is, 1)];
272 TG = TE - TF;
273 T13 = TE + TF;
274 }
275 {
276 E Tx, Ty, TB, TC;
277 Tx = ii[WS(is, 8)];
278 Ty = ii[WS(is, 3)];
279 Tz = Tx - Ty;
280 T16 = Tx + Ty;
281 TB = ii[WS(is, 4)];
282 TC = ii[WS(is, 9)];
283 TD = TB - TC;
284 T12 = TB + TC;
285 }
286 TA = Tw - Tz;
287 TH = TD - TG;
288 T17 = T15 - T16;
289 T14 = T12 - T13;
290 T1f = T15 + T16;
291 T1g = T12 + T13;
292 T1h = T1f + T1g;
293 TL = Tw + Tz;
294 TM = TD + TG;
295 TR = TL + TM;
296 }
297 ro[WS(os, 5)] = T3 + Ti;
298 io[WS(os, 5)] = TQ + TR;
299 ro[0] = Tj + Tq;
300 io[0] = T1e + T1h;
301 {
302 E TI, TK, Tt, TJ, Tr, Ts;
303 TI = FMA(KP951056516, TA, KP587785252 * TH);
304 TK = FNMS(KP587785252, TA, KP951056516 * TH);
305 Tr = KP559016994 * (Ta - Th);
306 Ts = FNMS(KP250000000, Ti, T3);
307 Tt = Tr + Ts;
308 TJ = Ts - Tr;
309 ro[WS(os, 9)] = Tt - TI;
310 ro[WS(os, 3)] = TJ + TK;
311 ro[WS(os, 1)] = Tt + TI;
312 ro[WS(os, 7)] = TJ - TK;
313 }
314 {
315 E TW, TY, TT, TX, TN, TS;
316 TW = FMA(KP951056516, TU, KP587785252 * TV);
317 TY = FNMS(KP587785252, TU, KP951056516 * TV);
318 TN = KP559016994 * (TL - TM);
319 TS = FNMS(KP250000000, TR, TQ);
320 TT = TN + TS;
321 TX = TS - TN;
322 io[WS(os, 1)] = TT - TW;
323 io[WS(os, 7)] = TY + TX;
324 io[WS(os, 9)] = TW + TT;
325 io[WS(os, 3)] = TX - TY;
326 }
327 {
328 E T18, T1a, T11, T19, TZ, T10;
329 T18 = FNMS(KP587785252, T17, KP951056516 * T14);
330 T1a = FMA(KP951056516, T17, KP587785252 * T14);
331 TZ = FNMS(KP250000000, Tq, Tj);
332 T10 = KP559016994 * (Tm - Tp);
333 T11 = TZ - T10;
334 T19 = T10 + TZ;
335 ro[WS(os, 2)] = T11 - T18;
336 ro[WS(os, 6)] = T19 + T1a;
337 ro[WS(os, 8)] = T11 + T18;
338 ro[WS(os, 4)] = T19 - T1a;
339 }
340 {
341 E T1d, T1l, T1k, T1m, T1i, T1j;
342 T1d = FNMS(KP587785252, T1c, KP951056516 * T1b);
343 T1l = FMA(KP951056516, T1c, KP587785252 * T1b);
344 T1i = FNMS(KP250000000, T1h, T1e);
345 T1j = KP559016994 * (T1f - T1g);
346 T1k = T1i - T1j;
347 T1m = T1j + T1i;
348 io[WS(os, 2)] = T1d + T1k;
349 io[WS(os, 6)] = T1m - T1l;
350 io[WS(os, 8)] = T1k - T1d;
351 io[WS(os, 4)] = T1l + T1m;
352 }
353 }
354 }
355 }
356
357 static const kdft_desc desc = { 10, "n1_10", { 72, 12, 12, 0 }, &GENUS, 0, 0, 0, 0 };
358
X(codelet_n1_10)359 void X(codelet_n1_10) (planner *p) { X(kdft_register) (p, n1_10, &desc);
360 }
361
362 #endif
363