Mercurial > hg > js-dsp-test

comparison fft/fftw/fftw-3.3.4/rdft/scalar/r2cf/hc2cf_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 13:49:21 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 -n 6 -dit -name hc2cf_6 -include hc2cf.h */
29
30 /*
31 * This function contains 46 FP additions, 32 FP multiplications,
32 * (or, 24 additions, 10 multiplications, 22 fused multiply/add),
33 * 47 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "hc2cf.h"
36
37 static void hc2cf_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 TY, TU, T10, TZ;
45 {
46 E T1, TX, TW, T7, Tn, Tq, TJ, TS, TB, Tl, To, TK, Tt, Tw, Ts;
47 E Tp, Tv;
48 T1 = Rp[0];
49 TX = Rm[0];
50 {
51 E T3, T6, T2, T5;
52 T3 = Ip[WS(rs, 1)];
53 T6 = Im[WS(rs, 1)];
54 T2 = W[4];
55 T5 = W[5];
56 {
57 E Ta, Td, Tg, TF, Tb, Tj, Tf, Tc, Ti, TV, T4, T9;
58 Ta = Rp[WS(rs, 1)];
59 Td = Rm[WS(rs, 1)];
60 TV = T2 * T6;
61 T4 = T2 * T3;
62 T9 = W[2];
63 Tg = Ip[WS(rs, 2)];
64 TW = FNMS(T5, T3, TV);
65 T7 = FMA(T5, T6, T4);
66 TF = T9 * Td;
67 Tb = T9 * Ta;
68 Tj = Im[WS(rs, 2)];
69 Tf = W[8];
70 Tc = W[3];
71 Ti = W[9];
72 {
73 E TG, Te, TI, Tk, TH, Th, Tm;
74 Tn = Rp[WS(rs, 2)];
75 TH = Tf * Tj;
76 Th = Tf * Tg;
77 TG = FNMS(Tc, Ta, TF);
78 Te = FMA(Tc, Td, Tb);
79 TI = FNMS(Ti, Tg, TH);
80 Tk = FMA(Ti, Tj, Th);
81 Tq = Rm[WS(rs, 2)];
82 Tm = W[6];
83 TJ = TG + TI;
84 TS = TI - TG;
85 TB = Te + Tk;
86 Tl = Te - Tk;
87 To = Tm * Tn;
88 TK = Tm * Tq;
89 }
90 Tt = Ip[0];
91 Tw = Im[0];
92 Ts = W[0];
93 Tp = W[7];
94 Tv = W[1];
95 }
96 }
97 {
98 E TA, T8, TL, Tr, TN, Tx, T12, TM, Tu;
99 TA = T1 + T7;
100 T8 = T1 - T7;
101 TM = Ts * Tw;
102 Tu = Ts * Tt;
103 TL = FNMS(Tp, Tn, TK);
104 Tr = FMA(Tp, Tq, To);
105 TN = FNMS(Tv, Tt, TM);
106 Tx = FMA(Tv, Tw, Tu);
107 T12 = TX - TW;
108 TY = TW + TX;
109 {
110 E TP, TT, TD, TQ, TE, Tz, T14, T13;
111 {
112 E TO, TR, TC, Ty, T11;
113 TO = TL + TN;
114 TR = TN - TL;
115 TC = Tr + Tx;
116 Ty = Tr - Tx;
117 TP = TJ - TO;
118 TU = TJ + TO;
119 TT = TR - TS;
120 T11 = TS + TR;
121 Tz = Tl + Ty;
122 T14 = Ty - Tl;
123 Im[WS(rs, 2)] = T11 - T12;
124 T13 = FMA(KP500000000, T11, T12);
125 T10 = TB - TC;
126 TD = TB + TC;
127 }
128 Rm[WS(rs, 2)] = T8 + Tz;
129 TQ = FNMS(KP500000000, Tz, T8);
130 Im[0] = FMS(KP866025403, T14, T13);
131 Ip[WS(rs, 1)] = FMA(KP866025403, T14, T13);
132 TE = FNMS(KP500000000, TD, TA);
133 Rm[0] = FNMS(KP866025403, TT, TQ);
134 Rp[WS(rs, 1)] = FMA(KP866025403, TT, TQ);
135 Rp[0] = TA + TD;
136 Rm[WS(rs, 1)] = FMA(KP866025403, TP, TE);
137 Rp[WS(rs, 2)] = FNMS(KP866025403, TP, TE);
138 }
139 }
140 }
141 Ip[0] = TU + TY;
142 TZ = FNMS(KP500000000, TU, TY);
143 Im[WS(rs, 1)] = FMS(KP866025403, T10, TZ);
144 Ip[WS(rs, 2)] = FMA(KP866025403, T10, TZ);
145 }
146 }
147 }
148
149 static const tw_instr twinstr[] = {
150 {TW_FULL, 1, 6},
151 {TW_NEXT, 1, 0}
152 };
153
154 static const hc2c_desc desc = { 6, "hc2cf_6", twinstr, &GENUS, {24, 10, 22, 0} };
155
156 void X(codelet_hc2cf_6) (planner *p) {
157 X(khc2c_register) (p, hc2cf_6, &desc, HC2C_VIA_RDFT);
158 }
159 #else /* HAVE_FMA */
160
161 /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -n 6 -dit -name hc2cf_6 -include hc2cf.h */
162
163 /*
164 * This function contains 46 FP additions, 28 FP multiplications,
165 * (or, 32 additions, 14 multiplications, 14 fused multiply/add),
166 * 23 stack variables, 2 constants, and 24 memory accesses
167 */
168 #include "hc2cf.h"
169
170 static void hc2cf_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
171 {
172 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
173 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
174 {
175 INT m;
176 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)) {
177 E T7, TS, Tv, TO, Tt, TJ, Tx, TF, Ti, TI, Tw, TC;
178 {
179 E T1, TN, T6, TM;
180 T1 = Rp[0];
181 TN = Rm[0];
182 {
183 E T3, T5, T2, T4;
184 T3 = Ip[WS(rs, 1)];
185 T5 = Im[WS(rs, 1)];
186 T2 = W[4];
187 T4 = W[5];
188 T6 = FMA(T2, T3, T4 * T5);
189 TM = FNMS(T4, T3, T2 * T5);
190 }
191 T7 = T1 - T6;
192 TS = TN - TM;
193 Tv = T1 + T6;
194 TO = TM + TN;
195 }
196 {
197 E Tn, TD, Ts, TE;
198 {
199 E Tk, Tm, Tj, Tl;
200 Tk = Rp[WS(rs, 2)];
201 Tm = Rm[WS(rs, 2)];
202 Tj = W[6];
203 Tl = W[7];
204 Tn = FMA(Tj, Tk, Tl * Tm);
205 TD = FNMS(Tl, Tk, Tj * Tm);
206 }
207 {
208 E Tp, Tr, To, Tq;
209 Tp = Ip[0];
210 Tr = Im[0];
211 To = W[0];
212 Tq = W[1];
213 Ts = FMA(To, Tp, Tq * Tr);
214 TE = FNMS(Tq, Tp, To * Tr);
215 }
216 Tt = Tn - Ts;
217 TJ = TE - TD;
218 Tx = Tn + Ts;
219 TF = TD + TE;
220 }
221 {
222 E Tc, TA, Th, TB;
223 {
224 E T9, Tb, T8, Ta;
225 T9 = Rp[WS(rs, 1)];
226 Tb = Rm[WS(rs, 1)];
227 T8 = W[2];
228 Ta = W[3];
229 Tc = FMA(T8, T9, Ta * Tb);
230 TA = FNMS(Ta, T9, T8 * Tb);
231 }
232 {
233 E Te, Tg, Td, Tf;
234 Te = Ip[WS(rs, 2)];
235 Tg = Im[WS(rs, 2)];
236 Td = W[8];
237 Tf = W[9];
238 Th = FMA(Td, Te, Tf * Tg);
239 TB = FNMS(Tf, Te, Td * Tg);
240 }
241 Ti = Tc - Th;
242 TI = TA - TB;
243 Tw = Tc + Th;
244 TC = TA + TB;
245 }
246 {
247 E TK, Tu, TH, TT, TR, TU;
248 TK = KP866025403 * (TI + TJ);
249 Tu = Ti + Tt;
250 TH = FNMS(KP500000000, Tu, T7);
251 Rm[WS(rs, 2)] = T7 + Tu;
252 Rp[WS(rs, 1)] = TH + TK;
253 Rm[0] = TH - TK;
254 TT = KP866025403 * (Tt - Ti);
255 TR = TJ - TI;
256 TU = FMA(KP500000000, TR, TS);
257 Im[WS(rs, 2)] = TR - TS;
258 Ip[WS(rs, 1)] = TT + TU;
259 Im[0] = TT - TU;
260 }
261 {
262 E TG, Ty, Tz, TP, TL, TQ;
263 TG = KP866025403 * (TC - TF);
264 Ty = Tw + Tx;
265 Tz = FNMS(KP500000000, Ty, Tv);
266 Rp[0] = Tv + Ty;
267 Rm[WS(rs, 1)] = Tz + TG;
268 Rp[WS(rs, 2)] = Tz - TG;
269 TP = KP866025403 * (Tw - Tx);
270 TL = TC + TF;
271 TQ = FNMS(KP500000000, TL, TO);
272 Ip[0] = TL + TO;
273 Ip[WS(rs, 2)] = TP + TQ;
274 Im[WS(rs, 1)] = TP - TQ;
275 }
276 }
277 }
278 }
279
280 static const tw_instr twinstr[] = {
281 {TW_FULL, 1, 6},
282 {TW_NEXT, 1, 0}
283 };
284
285 static const hc2c_desc desc = { 6, "hc2cf_6", twinstr, &GENUS, {32, 14, 14, 0} };
286
287 void X(codelet_hc2cf_6) (planner *p) {
288 X(khc2c_register) (p, hc2cf_6, &desc, HC2C_VIA_RDFT);
289 }
290 #endif /* HAVE_FMA */