Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t1_6.c @ 95:89f5e221ed7b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam <cannam@all-day-breakfast.com>
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
94:d278df1123f9 95:89f5e221ed7b
1 /*
2 * Copyright (c) 2003, 2007-11 Matteo Frigo
3 * Copyright (c) 2003, 2007-11 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 Sun Nov 25 07:35:48 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 6 -name t1_6 -include t.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 "t.h"
36
37 static void t1_6(R *ri, R *ii, 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 * 10); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 10, MAKE_VOLATILE_STRIDE(12, rs)) {
44 E TY, TU, T10, TZ;
45 {
46 E T1, TX, TW, T7, Tn, Tq, TJ, TR, TB, Tl, To, TK, Tt, Tw, Ts;
47 E Tp, Tv;
48 T1 = ri[0];
49 TX = ii[0];
50 {
51 E T3, T6, T2, T5;
52 T3 = ri[WS(rs, 3)];
53 T6 = ii[WS(rs, 3)];
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 = ri[WS(rs, 2)];
59 Td = ii[WS(rs, 2)];
60 TV = T2 * T6;
61 T4 = T2 * T3;
62 T9 = W[2];
63 Tg = ri[WS(rs, 5)];
64 TW = FNMS(T5, T3, TV);
65 T7 = FMA(T5, T6, T4);
66 TF = T9 * Td;
67 Tb = T9 * Ta;
68 Tj = ii[WS(rs, 5)];
69 Tf = W[8];
70 Tc = W[3];
71 Ti = W[9];
72 {
73 E TG, Te, TI, Tk, TH, Th, Tm;
74 Tn = ri[WS(rs, 4)];
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 = ii[WS(rs, 4)];
82 Tm = W[6];
83 TJ = TG - TI;
84 TR = TG + TI;
85 TB = Te + Tk;
86 Tl = Te - Tk;
87 To = Tm * Tn;
88 TK = Tm * Tq;
89 }
90 Tt = ri[WS(rs, 1)];
91 Tw = ii[WS(rs, 1)];
92 Ts = W[0];
93 Tp = W[7];
94 Tv = W[1];
95 }
96 }
97 {
98 E TA, T8, TL, Tr, TN, Tx, T11, 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 T11 = TX - TW;
108 TY = TW + TX;
109 {
110 E TP, TT, TD, TE, TQ, Tz, T14, T13;
111 {
112 E TO, TS, TC, Ty, T12;
113 TO = TL - TN;
114 TS = TL + TN;
115 TC = Tr + Tx;
116 Ty = Tr - Tx;
117 T12 = TJ + TO;
118 TP = TJ - TO;
119 TT = TR - TS;
120 TU = TR + TS;
121 Tz = Tl + Ty;
122 T14 = Ty - Tl;
123 ii[WS(rs, 3)] = T12 + T11;
124 T13 = FNMS(KP500000000, T12, T11);
125 T10 = TC - TB;
126 TD = TB + TC;
127 }
128 ri[WS(rs, 3)] = T8 + Tz;
129 TE = FNMS(KP500000000, Tz, T8);
130 ii[WS(rs, 5)] = FNMS(KP866025403, T14, T13);
131 ii[WS(rs, 1)] = FMA(KP866025403, T14, T13);
132 TQ = FNMS(KP500000000, TD, TA);
133 ri[WS(rs, 5)] = FNMS(KP866025403, TP, TE);
134 ri[WS(rs, 1)] = FMA(KP866025403, TP, TE);
135 ri[0] = TA + TD;
136 ri[WS(rs, 4)] = FMA(KP866025403, TT, TQ);
137 ri[WS(rs, 2)] = FNMS(KP866025403, TT, TQ);
138 }
139 }
140 }
141 ii[0] = TU + TY;
142 TZ = FNMS(KP500000000, TU, TY);
143 ii[WS(rs, 2)] = FNMS(KP866025403, T10, TZ);
144 ii[WS(rs, 4)] = FMA(KP866025403, T10, TZ);
145 }
146 }
147 }
148
149 static const tw_instr twinstr[] = {
150 {TW_FULL, 0, 6},
151 {TW_NEXT, 1, 0}
152 };
153
154 static const ct_desc desc = { 6, "t1_6", twinstr, &GENUS, {24, 10, 22, 0}, 0, 0, 0 };
155
156 void X(codelet_t1_6) (planner *p) {
157 X(kdft_dit_register) (p, t1_6, &desc);
158 }
159 #else /* HAVE_FMA */
160
161 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 6 -name t1_6 -include t.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 "t.h"
169
170 static void t1_6(R *ri, R *ii, 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 * 10); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 10, MAKE_VOLATILE_STRIDE(12, rs)) {
177 E T7, TS, Tv, TO, Tt, TJ, Tx, TF, Ti, TI, Tw, TC;
178 {
179 E T1, TN, T6, TM;
180 T1 = ri[0];
181 TN = ii[0];
182 {
183 E T3, T5, T2, T4;
184 T3 = ri[WS(rs, 3)];
185 T5 = ii[WS(rs, 3)];
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 = ri[WS(rs, 4)];
201 Tm = ii[WS(rs, 4)];
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 = ri[WS(rs, 1)];
210 Tr = ii[WS(rs, 1)];
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 = TD + TE;
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 = ri[WS(rs, 2)];
226 Tb = ii[WS(rs, 2)];
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 = ri[WS(rs, 5)];
235 Tg = ii[WS(rs, 5)];
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 TG, Tu, Tz, TR, TT, TU;
248 TG = KP866025403 * (TC - TF);
249 Tu = Ti + Tt;
250 Tz = FNMS(KP500000000, Tu, T7);
251 ri[WS(rs, 3)] = T7 + Tu;
252 ri[WS(rs, 1)] = Tz + TG;
253 ri[WS(rs, 5)] = Tz - TG;
254 TR = KP866025403 * (Tt - Ti);
255 TT = TC + TF;
256 TU = FNMS(KP500000000, TT, TS);
257 ii[WS(rs, 1)] = TR + TU;
258 ii[WS(rs, 3)] = TT + TS;
259 ii[WS(rs, 5)] = TU - TR;
260 }
261 {
262 E TK, Ty, TH, TQ, TL, TP;
263 TK = KP866025403 * (TI - TJ);
264 Ty = Tw + Tx;
265 TH = FNMS(KP500000000, Ty, Tv);
266 ri[0] = Tv + Ty;
267 ri[WS(rs, 4)] = TH + TK;
268 ri[WS(rs, 2)] = TH - TK;
269 TQ = KP866025403 * (Tx - Tw);
270 TL = TI + TJ;
271 TP = FNMS(KP500000000, TL, TO);
272 ii[0] = TL + TO;
273 ii[WS(rs, 4)] = TQ + TP;
274 ii[WS(rs, 2)] = TP - TQ;
275 }
276 }
277 }
278 }
279
280 static const tw_instr twinstr[] = {
281 {TW_FULL, 0, 6},
282 {TW_NEXT, 1, 0}
283 };
284
285 static const ct_desc desc = { 6, "t1_6", twinstr, &GENUS, {32, 14, 14, 0}, 0, 0, 0 };
286
287 void X(codelet_t1_6) (planner *p) {
288 X(kdft_dit_register) (p, t1_6, &desc);
289 }
290 #endif /* HAVE_FMA */