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 07:06:09 EST 2020 */
23
24 #include "rdft/codelet-rdft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_hc2c.native -fma -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cf_8 -include rdft/scalar/hc2cf.h */
29
30 /*
31 * This function contains 66 FP additions, 36 FP multiplications,
32 * (or, 44 additions, 14 multiplications, 22 fused multiply/add),
33 * 34 stack variables, 1 constants, and 32 memory accesses
34 */
35 #include "rdft/scalar/hc2cf.h"
36
hc2cf_8(R * Rp,R * Ip,R * Rm,R * Im,const R * W,stride rs,INT mb,INT me,INT ms)37 static void hc2cf_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT m;
42 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
43 E T1, T1m, T7, T1l, Tk, TS, Te, TQ, TF, T14, TL, T16, T12, T17, Ts;
44 E TX, Ty, TZ, TV, T10;
45 T1 = Rp[0];
46 T1m = Rm[0];
47 {
48 E T3, T6, T4, T1k, T2, T5;
49 T3 = Rp[WS(rs, 2)];
50 T6 = Rm[WS(rs, 2)];
51 T2 = W[6];
52 T4 = T2 * T3;
53 T1k = T2 * T6;
54 T5 = W[7];
55 T7 = FMA(T5, T6, T4);
56 T1l = FNMS(T5, T3, T1k);
57 }
58 {
59 E Tg, Tj, Th, TR, Tf, Ti;
60 Tg = Rp[WS(rs, 3)];
61 Tj = Rm[WS(rs, 3)];
62 Tf = W[10];
63 Th = Tf * Tg;
64 TR = Tf * Tj;
65 Ti = W[11];
66 Tk = FMA(Ti, Tj, Th);
67 TS = FNMS(Ti, Tg, TR);
68 }
69 {
70 E Ta, Td, Tb, TP, T9, Tc;
71 Ta = Rp[WS(rs, 1)];
72 Td = Rm[WS(rs, 1)];
73 T9 = W[2];
74 Tb = T9 * Ta;
75 TP = T9 * Td;
76 Tc = W[3];
77 Te = FMA(Tc, Td, Tb);
78 TQ = FNMS(Tc, Ta, TP);
79 }
80 {
81 E TB, TE, TC, T13, TH, TK, TI, T15, TA, TG, TD, TJ;
82 TB = Ip[WS(rs, 3)];
83 TE = Im[WS(rs, 3)];
84 TA = W[12];
85 TC = TA * TB;
86 T13 = TA * TE;
87 TH = Ip[WS(rs, 1)];
88 TK = Im[WS(rs, 1)];
89 TG = W[4];
90 TI = TG * TH;
91 T15 = TG * TK;
92 TD = W[13];
93 TF = FMA(TD, TE, TC);
94 T14 = FNMS(TD, TB, T13);
95 TJ = W[5];
96 TL = FMA(TJ, TK, TI);
97 T16 = FNMS(TJ, TH, T15);
98 T12 = TF - TL;
99 T17 = T14 - T16;
100 }
101 {
102 E To, Tr, Tp, TW, Tu, Tx, Tv, TY, Tn, Tt, Tq, Tw;
103 To = Ip[0];
104 Tr = Im[0];
105 Tn = W[0];
106 Tp = Tn * To;
107 TW = Tn * Tr;
108 Tu = Ip[WS(rs, 2)];
109 Tx = Im[WS(rs, 2)];
110 Tt = W[8];
111 Tv = Tt * Tu;
112 TY = Tt * Tx;
113 Tq = W[1];
114 Ts = FMA(Tq, Tr, Tp);
115 TX = FNMS(Tq, To, TW);
116 Tw = W[9];
117 Ty = FMA(Tw, Tx, Tv);
118 TZ = FNMS(Tw, Tu, TY);
119 TV = Ts - Ty;
120 T10 = TX - TZ;
121 }
122 {
123 E TU, T1a, T1t, T1v, T19, T1w, T1d, T1u;
124 {
125 E TO, TT, T1r, T1s;
126 TO = T1 - T7;
127 TT = TQ - TS;
128 TU = TO + TT;
129 T1a = TO - TT;
130 T1r = T1m - T1l;
131 T1s = Te - Tk;
132 T1t = T1r - T1s;
133 T1v = T1s + T1r;
134 }
135 {
136 E T11, T18, T1b, T1c;
137 T11 = TV + T10;
138 T18 = T12 - T17;
139 T19 = T11 + T18;
140 T1w = T18 - T11;
141 T1b = T10 - TV;
142 T1c = T12 + T17;
143 T1d = T1b - T1c;
144 T1u = T1b + T1c;
145 }
146 Rm[WS(rs, 2)] = FNMS(KP707106781, T19, TU);
147 Im[WS(rs, 2)] = FMS(KP707106781, T1u, T1t);
148 Rp[WS(rs, 1)] = FMA(KP707106781, T19, TU);
149 Ip[WS(rs, 1)] = FMA(KP707106781, T1u, T1t);
150 Rm[0] = FNMS(KP707106781, T1d, T1a);
151 Im[0] = FMS(KP707106781, T1w, T1v);
152 Rp[WS(rs, 3)] = FMA(KP707106781, T1d, T1a);
153 Ip[WS(rs, 3)] = FMA(KP707106781, T1w, T1v);
154 }
155 {
156 E Tm, T1e, T1o, T1q, TN, T1p, T1h, T1i;
157 {
158 E T8, Tl, T1j, T1n;
159 T8 = T1 + T7;
160 Tl = Te + Tk;
161 Tm = T8 + Tl;
162 T1e = T8 - Tl;
163 T1j = TQ + TS;
164 T1n = T1l + T1m;
165 T1o = T1j + T1n;
166 T1q = T1n - T1j;
167 }
168 {
169 E Tz, TM, T1f, T1g;
170 Tz = Ts + Ty;
171 TM = TF + TL;
172 TN = Tz + TM;
173 T1p = TM - Tz;
174 T1f = TX + TZ;
175 T1g = T14 + T16;
176 T1h = T1f - T1g;
177 T1i = T1f + T1g;
178 }
179 Rm[WS(rs, 3)] = Tm - TN;
180 Im[WS(rs, 3)] = T1i - T1o;
181 Rp[0] = Tm + TN;
182 Ip[0] = T1i + T1o;
183 Rm[WS(rs, 1)] = T1e - T1h;
184 Im[WS(rs, 1)] = T1p - T1q;
185 Rp[WS(rs, 2)] = T1e + T1h;
186 Ip[WS(rs, 2)] = T1p + T1q;
187 }
188 }
189 }
190 }
191
192 static const tw_instr twinstr[] = {
193 { TW_FULL, 1, 8 },
194 { TW_NEXT, 1, 0 }
195 };
196
197 static const hc2c_desc desc = { 8, "hc2cf_8", twinstr, &GENUS, { 44, 14, 22, 0 } };
198
X(codelet_hc2cf_8)199 void X(codelet_hc2cf_8) (planner *p) {
200 X(khc2c_register) (p, hc2cf_8, &desc, HC2C_VIA_RDFT);
201 }
202 #else
203
204 /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cf_8 -include rdft/scalar/hc2cf.h */
205
206 /*
207 * This function contains 66 FP additions, 32 FP multiplications,
208 * (or, 52 additions, 18 multiplications, 14 fused multiply/add),
209 * 28 stack variables, 1 constants, and 32 memory accesses
210 */
211 #include "rdft/scalar/hc2cf.h"
212
hc2cf_8(R * Rp,R * Ip,R * Rm,R * Im,const R * W,stride rs,INT mb,INT me,INT ms)213 static void hc2cf_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
214 {
215 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
216 {
217 INT m;
218 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
219 E T7, T1e, TH, T19, TF, T13, TR, TU, Ti, T1f, TK, T16, Tu, T12, TM;
220 E TP;
221 {
222 E T1, T18, T6, T17;
223 T1 = Rp[0];
224 T18 = Rm[0];
225 {
226 E T3, T5, T2, T4;
227 T3 = Rp[WS(rs, 2)];
228 T5 = Rm[WS(rs, 2)];
229 T2 = W[6];
230 T4 = W[7];
231 T6 = FMA(T2, T3, T4 * T5);
232 T17 = FNMS(T4, T3, T2 * T5);
233 }
234 T7 = T1 + T6;
235 T1e = T18 - T17;
236 TH = T1 - T6;
237 T19 = T17 + T18;
238 }
239 {
240 E Tz, TS, TE, TT;
241 {
242 E Tw, Ty, Tv, Tx;
243 Tw = Ip[WS(rs, 3)];
244 Ty = Im[WS(rs, 3)];
245 Tv = W[12];
246 Tx = W[13];
247 Tz = FMA(Tv, Tw, Tx * Ty);
248 TS = FNMS(Tx, Tw, Tv * Ty);
249 }
250 {
251 E TB, TD, TA, TC;
252 TB = Ip[WS(rs, 1)];
253 TD = Im[WS(rs, 1)];
254 TA = W[4];
255 TC = W[5];
256 TE = FMA(TA, TB, TC * TD);
257 TT = FNMS(TC, TB, TA * TD);
258 }
259 TF = Tz + TE;
260 T13 = TS + TT;
261 TR = Tz - TE;
262 TU = TS - TT;
263 }
264 {
265 E Tc, TI, Th, TJ;
266 {
267 E T9, Tb, T8, Ta;
268 T9 = Rp[WS(rs, 1)];
269 Tb = Rm[WS(rs, 1)];
270 T8 = W[2];
271 Ta = W[3];
272 Tc = FMA(T8, T9, Ta * Tb);
273 TI = FNMS(Ta, T9, T8 * Tb);
274 }
275 {
276 E Te, Tg, Td, Tf;
277 Te = Rp[WS(rs, 3)];
278 Tg = Rm[WS(rs, 3)];
279 Td = W[10];
280 Tf = W[11];
281 Th = FMA(Td, Te, Tf * Tg);
282 TJ = FNMS(Tf, Te, Td * Tg);
283 }
284 Ti = Tc + Th;
285 T1f = Tc - Th;
286 TK = TI - TJ;
287 T16 = TI + TJ;
288 }
289 {
290 E To, TN, Tt, TO;
291 {
292 E Tl, Tn, Tk, Tm;
293 Tl = Ip[0];
294 Tn = Im[0];
295 Tk = W[0];
296 Tm = W[1];
297 To = FMA(Tk, Tl, Tm * Tn);
298 TN = FNMS(Tm, Tl, Tk * Tn);
299 }
300 {
301 E Tq, Ts, Tp, Tr;
302 Tq = Ip[WS(rs, 2)];
303 Ts = Im[WS(rs, 2)];
304 Tp = W[8];
305 Tr = W[9];
306 Tt = FMA(Tp, Tq, Tr * Ts);
307 TO = FNMS(Tr, Tq, Tp * Ts);
308 }
309 Tu = To + Tt;
310 T12 = TN + TO;
311 TM = To - Tt;
312 TP = TN - TO;
313 }
314 {
315 E Tj, TG, T1b, T1c;
316 Tj = T7 + Ti;
317 TG = Tu + TF;
318 Rm[WS(rs, 3)] = Tj - TG;
319 Rp[0] = Tj + TG;
320 {
321 E T15, T1a, T11, T14;
322 T15 = T12 + T13;
323 T1a = T16 + T19;
324 Im[WS(rs, 3)] = T15 - T1a;
325 Ip[0] = T15 + T1a;
326 T11 = T7 - Ti;
327 T14 = T12 - T13;
328 Rm[WS(rs, 1)] = T11 - T14;
329 Rp[WS(rs, 2)] = T11 + T14;
330 }
331 T1b = TF - Tu;
332 T1c = T19 - T16;
333 Im[WS(rs, 1)] = T1b - T1c;
334 Ip[WS(rs, 2)] = T1b + T1c;
335 {
336 E TX, T1g, T10, T1d, TY, TZ;
337 TX = TH - TK;
338 T1g = T1e - T1f;
339 TY = TP - TM;
340 TZ = TR + TU;
341 T10 = KP707106781 * (TY - TZ);
342 T1d = KP707106781 * (TY + TZ);
343 Rm[0] = TX - T10;
344 Ip[WS(rs, 1)] = T1d + T1g;
345 Rp[WS(rs, 3)] = TX + T10;
346 Im[WS(rs, 2)] = T1d - T1g;
347 }
348 {
349 E TL, T1i, TW, T1h, TQ, TV;
350 TL = TH + TK;
351 T1i = T1f + T1e;
352 TQ = TM + TP;
353 TV = TR - TU;
354 TW = KP707106781 * (TQ + TV);
355 T1h = KP707106781 * (TV - TQ);
356 Rm[WS(rs, 2)] = TL - TW;
357 Ip[WS(rs, 3)] = T1h + T1i;
358 Rp[WS(rs, 1)] = TL + TW;
359 Im[0] = T1h - T1i;
360 }
361 }
362 }
363 }
364 }
365
366 static const tw_instr twinstr[] = {
367 { TW_FULL, 1, 8 },
368 { TW_NEXT, 1, 0 }
369 };
370
371 static const hc2c_desc desc = { 8, "hc2cf_8", twinstr, &GENUS, { 52, 18, 14, 0 } };
372
X(codelet_hc2cf_8)373 void X(codelet_hc2cf_8) (planner *p) {
374 X(khc2c_register) (p, hc2cf_8, &desc, HC2C_VIA_RDFT);
375 }
376 #endif
377