Mercurial > hg > js-dsp-test

comparison fft/fftw/fftw-3.3.4/rdft/scalar/r2cb/hc2cb_6.c @ 19:26056e866c29

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW to comparison table
author Chris Cannam
date Tue, 06 Oct 2015 13:08:39 +0100
parents
children
comparison
equal deleted inserted replaced
18:8db794ca3e0b 19:26056e866c29
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 Tue Mar 4 13:50:37 EST 2014 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_hc2c.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hc2cb_6 -include hc2cb.h */
29
30 /*
31 * This function contains 46 FP additions, 32 FP multiplications,
32 * (or, 24 additions, 10 multiplications, 22 fused multiply/add),
33 * 45 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "hc2cb.h"
36
37 static void hc2cb_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT m;
43 for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 10, MAKE_VOLATILE_STRIDE(24, rs)) {
44 E TK, TR, TB, TM, TL, TS;
45 {
46 E Td, TN, TO, TJ, Tn, Tk, TC, T3, Tr, T7, T8, T4, T5;
47 {
48 E TI, Tj, Tg, TH, Te, Tf, T1, T2;
49 {
50 E Tb, Tc, Th, Ti;
51 Tb = Ip[0];
52 Tc = Im[WS(rs, 2)];
53 Th = Ip[WS(rs, 1)];
54 Ti = Im[WS(rs, 1)];
55 Te = Ip[WS(rs, 2)];
56 Td = Tb - Tc;
57 TN = Tb + Tc;
58 Tf = Im[0];
59 TI = Th + Ti;
60 Tj = Th - Ti;
61 }
62 Tg = Te - Tf;
63 TH = Te + Tf;
64 T1 = Rp[0];
65 T2 = Rm[WS(rs, 2)];
66 TO = TH - TI;
67 TJ = TH + TI;
68 Tn = Tj - Tg;
69 Tk = Tg + Tj;
70 TC = T1 - T2;
71 T3 = T1 + T2;
72 Tr = FNMS(KP500000000, Tk, Td);
73 T7 = Rm[WS(rs, 1)];
74 T8 = Rp[WS(rs, 1)];
75 T4 = Rp[WS(rs, 2)];
76 T5 = Rm[0];
77 }
78 {
79 E Tl, Tq, TQ, Ts, Ta, T10, TG;
80 Rm[0] = Td + Tk;
81 {
82 E T9, TE, T6, TD, TF;
83 T9 = T7 + T8;
84 TE = T7 - T8;
85 T6 = T4 + T5;
86 TD = T4 - T5;
87 Tl = W[2];
88 Tq = W[3];
89 TQ = TD - TE;
90 TF = TD + TE;
91 Ts = T6 - T9;
92 Ta = T6 + T9;
93 T10 = TC + TF;
94 TG = FNMS(KP500000000, TF, TC);
95 }
96 {
97 E T13, TP, Tz, TZ, Tw, T14, Tv, Ty;
98 {
99 E Tt, T12, T11, Tp, Tm, To, Tu;
100 T13 = TN + TO;
101 TP = FNMS(KP500000000, TO, TN);
102 Rp[0] = T3 + Ta;
103 Tm = FNMS(KP500000000, Ta, T3);
104 Tz = FMA(KP866025403, Ts, Tr);
105 Tt = FNMS(KP866025403, Ts, Tr);
106 TZ = W[4];
107 To = FNMS(KP866025403, Tn, Tm);
108 Tw = FMA(KP866025403, Tn, Tm);
109 Tu = Tl * Tt;
110 T12 = W[5];
111 T11 = TZ * T10;
112 Tp = Tl * To;
113 Rm[WS(rs, 1)] = FMA(Tq, To, Tu);
114 T14 = T12 * T10;
115 Ip[WS(rs, 1)] = FNMS(T12, T13, T11);
116 Rp[WS(rs, 1)] = FNMS(Tq, Tt, Tp);
117 }
118 Im[WS(rs, 1)] = FMA(TZ, T13, T14);
119 Tv = W[6];
120 Ty = W[7];
121 {
122 E TX, TT, TW, TV, TY, TU, TA, Tx;
123 TK = FNMS(KP866025403, TJ, TG);
124 TU = FMA(KP866025403, TJ, TG);
125 TA = Tv * Tz;
126 Tx = Tv * Tw;
127 TX = FNMS(KP866025403, TQ, TP);
128 TR = FMA(KP866025403, TQ, TP);
129 Rm[WS(rs, 2)] = FMA(Ty, Tw, TA);
130 Rp[WS(rs, 2)] = FNMS(Ty, Tz, Tx);
131 TT = W[8];
132 TW = W[9];
133 TB = W[0];
134 TV = TT * TU;
135 TY = TW * TU;
136 TM = W[1];
137 TL = TB * TK;
138 Ip[WS(rs, 2)] = FNMS(TW, TX, TV);
139 Im[WS(rs, 2)] = FMA(TT, TX, TY);
140 }
141 }
142 }
143 }
144 Ip[0] = FNMS(TM, TR, TL);
145 TS = TM * TK;
146 Im[0] = FMA(TB, TR, TS);
147 }
148 }
149 }
150
151 static const tw_instr twinstr[] = {
152 {TW_FULL, 1, 6},
153 {TW_NEXT, 1, 0}
154 };
155
156 static const hc2c_desc desc = { 6, "hc2cb_6", twinstr, &GENUS, {24, 10, 22, 0} };
157
158 void X(codelet_hc2cb_6) (planner *p) {
159 X(khc2c_register) (p, hc2cb_6, &desc, HC2C_VIA_RDFT);
160 }
161 #else /* HAVE_FMA */
162
163 /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hc2cb_6 -include hc2cb.h */
164
165 /*
166 * This function contains 46 FP additions, 28 FP multiplications,
167 * (or, 32 additions, 14 multiplications, 14 fused multiply/add),
168 * 25 stack variables, 2 constants, and 24 memory accesses
169 */
170 #include "hc2cb.h"
171
172 static void hc2cb_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
173 {
174 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
175 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
176 {
177 INT m;
178 for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 10, MAKE_VOLATILE_STRIDE(24, rs)) {
179 E T3, Ty, Td, TE, Ta, TO, Tr, TB, Tk, TL, Tn, TH;
180 {
181 E T1, T2, Tb, Tc;
182 T1 = Rp[0];
183 T2 = Rm[WS(rs, 2)];
184 T3 = T1 + T2;
185 Ty = T1 - T2;
186 Tb = Ip[0];
187 Tc = Im[WS(rs, 2)];
188 Td = Tb - Tc;
189 TE = Tb + Tc;
190 }
191 {
192 E T6, Tz, T9, TA;
193 {
194 E T4, T5, T7, T8;
195 T4 = Rp[WS(rs, 2)];
196 T5 = Rm[0];
197 T6 = T4 + T5;
198 Tz = T4 - T5;
199 T7 = Rm[WS(rs, 1)];
200 T8 = Rp[WS(rs, 1)];
201 T9 = T7 + T8;
202 TA = T7 - T8;
203 }
204 Ta = T6 + T9;
205 TO = KP866025403 * (Tz - TA);
206 Tr = KP866025403 * (T6 - T9);
207 TB = Tz + TA;
208 }
209 {
210 E Tg, TG, Tj, TF;
211 {
212 E Te, Tf, Th, Ti;
213 Te = Ip[WS(rs, 2)];
214 Tf = Im[0];
215 Tg = Te - Tf;
216 TG = Te + Tf;
217 Th = Ip[WS(rs, 1)];
218 Ti = Im[WS(rs, 1)];
219 Tj = Th - Ti;
220 TF = Th + Ti;
221 }
222 Tk = Tg + Tj;
223 TL = KP866025403 * (TG + TF);
224 Tn = KP866025403 * (Tj - Tg);
225 TH = TF - TG;
226 }
227 Rp[0] = T3 + Ta;
228 Rm[0] = Td + Tk;
229 {
230 E TC, TI, Tx, TD;
231 TC = Ty + TB;
232 TI = TE - TH;
233 Tx = W[4];
234 TD = W[5];
235 Ip[WS(rs, 1)] = FNMS(TD, TI, Tx * TC);
236 Im[WS(rs, 1)] = FMA(TD, TC, Tx * TI);
237 }
238 {
239 E To, Tu, Ts, Tw, Tm, Tq;
240 Tm = FNMS(KP500000000, Ta, T3);
241 To = Tm - Tn;
242 Tu = Tm + Tn;
243 Tq = FNMS(KP500000000, Tk, Td);
244 Ts = Tq - Tr;
245 Tw = Tr + Tq;
246 {
247 E Tl, Tp, Tt, Tv;
248 Tl = W[2];
249 Tp = W[3];
250 Rp[WS(rs, 1)] = FNMS(Tp, Ts, Tl * To);
251 Rm[WS(rs, 1)] = FMA(Tl, Ts, Tp * To);
252 Tt = W[6];
253 Tv = W[7];
254 Rp[WS(rs, 2)] = FNMS(Tv, Tw, Tt * Tu);
255 Rm[WS(rs, 2)] = FMA(Tt, Tw, Tv * Tu);
256 }
257 }
258 {
259 E TM, TS, TQ, TU, TK, TP;
260 TK = FNMS(KP500000000, TB, Ty);
261 TM = TK - TL;
262 TS = TK + TL;
263 TP = FMA(KP500000000, TH, TE);
264 TQ = TO + TP;
265 TU = TP - TO;
266 {
267 E TJ, TN, TR, TT;
268 TJ = W[0];
269 TN = W[1];
270 Ip[0] = FNMS(TN, TQ, TJ * TM);
271 Im[0] = FMA(TN, TM, TJ * TQ);
272 TR = W[8];
273 TT = W[9];
274 Ip[WS(rs, 2)] = FNMS(TT, TU, TR * TS);
275 Im[WS(rs, 2)] = FMA(TT, TS, TR * TU);
276 }
277 }
278 }
279 }
280 }
281
282 static const tw_instr twinstr[] = {
283 {TW_FULL, 1, 6},
284 {TW_NEXT, 1, 0}
285 };
286
287 static const hc2c_desc desc = { 6, "hc2cb_6", twinstr, &GENUS, {32, 14, 14, 0} };
288
289 void X(codelet_hc2cb_6) (planner *p) {
290 X(khc2c_register) (p, hc2cb_6, &desc, HC2C_VIA_RDFT);
291 }
292 #endif /* HAVE_FMA */