Mercurial > hg > js-dsp-test

comparison fft/fftw/fftw-3.3.4/rdft/scalar/r2cf/r2cfII_16.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 13:49:19 EST 2014 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 16 -name r2cfII_16 -dft-II -include r2cfII.h */
29
30 /*
31 * This function contains 66 FP additions, 48 FP multiplications,
32 * (or, 18 additions, 0 multiplications, 48 fused multiply/add),
33 * 54 stack variables, 7 constants, and 32 memory accesses
34 */
35 #include "r2cfII.h"
36
37 static void r2cfII_16(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
38 {
39 DK(KP980785280, +0.980785280403230449126182236134239036973933731);
40 DK(KP198912367, +0.198912367379658006911597622644676228597850501);
41 DK(KP831469612, +0.831469612302545237078788377617905756738560812);
42 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
43 DK(KP668178637, +0.668178637919298919997757686523080761552472251);
44 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
45 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
46 {
47 INT i;
48 for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
49 E TN, TF, TX, TV, TO, TP, TY, TM, TQ, TW;
50 {
51 E TT, TZ, TB, T5, Tu, TK, TJ, Tr, T9, TC, T8, Tl, TH, TG, Ti;
52 E Ta;
53 {
54 E T1, TR, Tn, Ts, To, TS, T4, Tp, T2, T3;
55 T1 = R0[0];
56 TR = R0[WS(rs, 4)];
57 T2 = R0[WS(rs, 2)];
58 T3 = R0[WS(rs, 6)];
59 Tn = R1[WS(rs, 7)];
60 Ts = R1[WS(rs, 3)];
61 To = R1[WS(rs, 1)];
62 TS = T2 + T3;
63 T4 = T2 - T3;
64 Tp = R1[WS(rs, 5)];
65 {
66 E Te, Tj, Tf, Tg, Tt, Tq;
67 Te = R1[0];
68 TT = FMA(KP707106781, TS, TR);
69 TZ = FNMS(KP707106781, TS, TR);
70 TB = FMA(KP707106781, T4, T1);
71 T5 = FNMS(KP707106781, T4, T1);
72 Tt = To + Tp;
73 Tq = To - Tp;
74 Tj = R1[WS(rs, 4)];
75 Tf = R1[WS(rs, 2)];
76 Tu = FNMS(KP707106781, Tt, Ts);
77 TK = FMA(KP707106781, Tt, Ts);
78 TJ = FMS(KP707106781, Tq, Tn);
79 Tr = FMA(KP707106781, Tq, Tn);
80 Tg = R1[WS(rs, 6)];
81 {
82 E T6, T7, Tk, Th;
83 T6 = R0[WS(rs, 5)];
84 T7 = R0[WS(rs, 1)];
85 T9 = R0[WS(rs, 3)];
86 Tk = Tf + Tg;
87 Th = Tf - Tg;
88 TC = FNMS(KP414213562, T6, T7);
89 T8 = FMA(KP414213562, T7, T6);
90 Tl = FNMS(KP707106781, Tk, Tj);
91 TH = FMA(KP707106781, Tk, Tj);
92 TG = FMA(KP707106781, Th, Te);
93 Ti = FNMS(KP707106781, Th, Te);
94 Ta = R0[WS(rs, 7)];
95 }
96 }
97 }
98 {
99 E TE, TU, Ty, Tv, TI, TL;
100 Ty = FNMS(KP668178637, Tr, Tu);
101 Tv = FMA(KP668178637, Tu, Tr);
102 {
103 E Tw, T14, T12, TA, T11, T13, Tx, Td;
104 {
105 E Tz, Tm, TD, Tb, T10, Tc;
106 Tz = FNMS(KP668178637, Ti, Tl);
107 Tm = FMA(KP668178637, Tl, Ti);
108 TD = FMS(KP414213562, T9, Ta);
109 Tb = FMA(KP414213562, Ta, T9);
110 Tw = Tm - Tv;
111 T14 = Tm + Tv;
112 T10 = TD - TC;
113 TE = TC + TD;
114 Tc = T8 - Tb;
115 TU = T8 + Tb;
116 T12 = Tz + Ty;
117 TA = Ty - Tz;
118 T11 = FMA(KP923879532, T10, TZ);
119 T13 = FNMS(KP923879532, T10, TZ);
120 Tx = FNMS(KP923879532, Tc, T5);
121 Td = FMA(KP923879532, Tc, T5);
122 }
123 Ci[WS(csi, 2)] = -(FMA(KP831469612, T14, T13));
124 Ci[WS(csi, 5)] = FNMS(KP831469612, T14, T13);
125 Cr[WS(csr, 1)] = FMA(KP831469612, Tw, Td);
126 Cr[WS(csr, 6)] = FNMS(KP831469612, Tw, Td);
127 Cr[WS(csr, 5)] = FNMS(KP831469612, TA, Tx);
128 Ci[WS(csi, 1)] = FMA(KP831469612, T12, T11);
129 Cr[WS(csr, 2)] = FMA(KP831469612, TA, Tx);
130 Ci[WS(csi, 6)] = FMS(KP831469612, T12, T11);
131 }
132 TN = FNMS(KP923879532, TE, TB);
133 TF = FMA(KP923879532, TE, TB);
134 TX = FNMS(KP923879532, TU, TT);
135 TV = FMA(KP923879532, TU, TT);
136 TO = FMA(KP198912367, TG, TH);
137 TI = FNMS(KP198912367, TH, TG);
138 TL = FMA(KP198912367, TK, TJ);
139 TP = FNMS(KP198912367, TJ, TK);
140 TY = TL - TI;
141 TM = TI + TL;
142 }
143 }
144 Ci[WS(csi, 4)] = FMS(KP980785280, TY, TX);
145 Ci[WS(csi, 3)] = FMA(KP980785280, TY, TX);
146 Cr[0] = FMA(KP980785280, TM, TF);
147 Cr[WS(csr, 7)] = FNMS(KP980785280, TM, TF);
148 TQ = TO - TP;
149 TW = TO + TP;
150 Ci[0] = -(FMA(KP980785280, TW, TV));
151 Ci[WS(csi, 7)] = FNMS(KP980785280, TW, TV);
152 Cr[WS(csr, 3)] = FMA(KP980785280, TQ, TN);
153 Cr[WS(csr, 4)] = FNMS(KP980785280, TQ, TN);
154 }
155 }
156 }
157
158 static const kr2c_desc desc = { 16, "r2cfII_16", {18, 0, 48, 0}, &GENUS };
159
160 void X(codelet_r2cfII_16) (planner *p) {
161 X(kr2c_register) (p, r2cfII_16, &desc);
162 }
163
164 #else /* HAVE_FMA */
165
166 /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 16 -name r2cfII_16 -dft-II -include r2cfII.h */
167
168 /*
169 * This function contains 66 FP additions, 30 FP multiplications,
170 * (or, 54 additions, 18 multiplications, 12 fused multiply/add),
171 * 32 stack variables, 7 constants, and 32 memory accesses
172 */
173 #include "r2cfII.h"
174
175 static void r2cfII_16(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
176 {
177 DK(KP555570233, +0.555570233019602224742830813948532874374937191);
178 DK(KP831469612, +0.831469612302545237078788377617905756738560812);
179 DK(KP980785280, +0.980785280403230449126182236134239036973933731);
180 DK(KP195090322, +0.195090322016128267848284868477022240927691618);
181 DK(KP382683432, +0.382683432365089771728459984030398866761344562);
182 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
183 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
184 {
185 INT i;
186 for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
187 E T5, T11, TB, TV, Tr, TK, Tu, TJ, Ti, TH, Tl, TG, Tc, T10, TE;
188 E TS;
189 {
190 E T1, TU, T4, TT, T2, T3;
191 T1 = R0[0];
192 TU = R0[WS(rs, 4)];
193 T2 = R0[WS(rs, 2)];
194 T3 = R0[WS(rs, 6)];
195 T4 = KP707106781 * (T2 - T3);
196 TT = KP707106781 * (T2 + T3);
197 T5 = T1 + T4;
198 T11 = TU - TT;
199 TB = T1 - T4;
200 TV = TT + TU;
201 }
202 {
203 E Tq, Tt, Tp, Ts, Tn, To;
204 Tq = R1[WS(rs, 7)];
205 Tt = R1[WS(rs, 3)];
206 Tn = R1[WS(rs, 1)];
207 To = R1[WS(rs, 5)];
208 Tp = KP707106781 * (Tn - To);
209 Ts = KP707106781 * (Tn + To);
210 Tr = Tp - Tq;
211 TK = Tt - Ts;
212 Tu = Ts + Tt;
213 TJ = Tp + Tq;
214 }
215 {
216 E Te, Tk, Th, Tj, Tf, Tg;
217 Te = R1[0];
218 Tk = R1[WS(rs, 4)];
219 Tf = R1[WS(rs, 2)];
220 Tg = R1[WS(rs, 6)];
221 Th = KP707106781 * (Tf - Tg);
222 Tj = KP707106781 * (Tf + Tg);
223 Ti = Te + Th;
224 TH = Tk - Tj;
225 Tl = Tj + Tk;
226 TG = Te - Th;
227 }
228 {
229 E T8, TC, Tb, TD;
230 {
231 E T6, T7, T9, Ta;
232 T6 = R0[WS(rs, 1)];
233 T7 = R0[WS(rs, 5)];
234 T8 = FNMS(KP382683432, T7, KP923879532 * T6);
235 TC = FMA(KP382683432, T6, KP923879532 * T7);
236 T9 = R0[WS(rs, 3)];
237 Ta = R0[WS(rs, 7)];
238 Tb = FNMS(KP923879532, Ta, KP382683432 * T9);
239 TD = FMA(KP923879532, T9, KP382683432 * Ta);
240 }
241 Tc = T8 + Tb;
242 T10 = Tb - T8;
243 TE = TC - TD;
244 TS = TC + TD;
245 }
246 {
247 E Td, TW, Tw, TR, Tm, Tv;
248 Td = T5 - Tc;
249 TW = TS + TV;
250 Tm = FMA(KP195090322, Ti, KP980785280 * Tl);
251 Tv = FNMS(KP980785280, Tu, KP195090322 * Tr);
252 Tw = Tm + Tv;
253 TR = Tv - Tm;
254 Cr[WS(csr, 4)] = Td - Tw;
255 Ci[WS(csi, 7)] = TR + TW;
256 Cr[WS(csr, 3)] = Td + Tw;
257 Ci[0] = TR - TW;
258 }
259 {
260 E Tx, TY, TA, TX, Ty, Tz;
261 Tx = T5 + Tc;
262 TY = TV - TS;
263 Ty = FNMS(KP195090322, Tl, KP980785280 * Ti);
264 Tz = FMA(KP980785280, Tr, KP195090322 * Tu);
265 TA = Ty + Tz;
266 TX = Tz - Ty;
267 Cr[WS(csr, 7)] = Tx - TA;
268 Ci[WS(csi, 3)] = TX + TY;
269 Cr[0] = Tx + TA;
270 Ci[WS(csi, 4)] = TX - TY;
271 }
272 {
273 E TF, T12, TM, TZ, TI, TL;
274 TF = TB + TE;
275 T12 = T10 - T11;
276 TI = FMA(KP831469612, TG, KP555570233 * TH);
277 TL = FMA(KP831469612, TJ, KP555570233 * TK);
278 TM = TI - TL;
279 TZ = TI + TL;
280 Cr[WS(csr, 6)] = TF - TM;
281 Ci[WS(csi, 2)] = T12 - TZ;
282 Cr[WS(csr, 1)] = TF + TM;
283 Ci[WS(csi, 5)] = -(TZ + T12);
284 }
285 {
286 E TN, T14, TQ, T13, TO, TP;
287 TN = TB - TE;
288 T14 = T10 + T11;
289 TO = FNMS(KP555570233, TJ, KP831469612 * TK);
290 TP = FNMS(KP555570233, TG, KP831469612 * TH);
291 TQ = TO - TP;
292 T13 = TP + TO;
293 Cr[WS(csr, 5)] = TN - TQ;
294 Ci[WS(csi, 1)] = T13 + T14;
295 Cr[WS(csr, 2)] = TN + TQ;
296 Ci[WS(csi, 6)] = T13 - T14;
297 }
298 }
299 }
300 }
301
302 static const kr2c_desc desc = { 16, "r2cfII_16", {54, 18, 12, 0}, &GENUS };
303
304 void X(codelet_r2cfII_16) (planner *p) {
305 X(kr2c_register) (p, r2cfII_16, &desc);
306 }
307
308 #endif /* HAVE_FMA */