Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/t1_6.c @ 167:bd3cc4d1df30

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam <cannam@all-day-breakfast.com>
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
166:cbd6d7e562c7 167:bd3cc4d1df30
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 May 24 08:04:13 EDT 2018 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_twiddle.native -fma -compact -variables 4 -pipeline-latency 4 -n 6 -name t1_6 -include dft/scalar/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 * 31 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "dft/scalar/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 T1, TX, T7, TW, Tl, TR, TB, TJ, Ty, TS, TC, TO;
45 T1 = ri[0];
46 TX = ii[0];
47 {
48 E T3, T6, T4, TV, T2, T5;
49 T3 = ri[WS(rs, 3)];
50 T6 = ii[WS(rs, 3)];
51 T2 = W[4];
52 T4 = T2 * T3;
53 TV = T2 * T6;
54 T5 = W[5];
55 T7 = FMA(T5, T6, T4);
56 TW = FNMS(T5, T3, TV);
57 }
58 {
59 E Ta, Td, Tb, TF, Tg, Tj, Th, TH, T9, Tf;
60 Ta = ri[WS(rs, 2)];
61 Td = ii[WS(rs, 2)];
62 T9 = W[2];
63 Tb = T9 * Ta;
64 TF = T9 * Td;
65 Tg = ri[WS(rs, 5)];
66 Tj = ii[WS(rs, 5)];
67 Tf = W[8];
68 Th = Tf * Tg;
69 TH = Tf * Tj;
70 {
71 E Te, TG, Tk, TI, Tc, Ti;
72 Tc = W[3];
73 Te = FMA(Tc, Td, Tb);
74 TG = FNMS(Tc, Ta, TF);
75 Ti = W[9];
76 Tk = FMA(Ti, Tj, Th);
77 TI = FNMS(Ti, Tg, TH);
78 Tl = Te - Tk;
79 TR = TG + TI;
80 TB = Te + Tk;
81 TJ = TG - TI;
82 }
83 }
84 {
85 E Tn, Tq, To, TK, Tt, Tw, Tu, TM, Tm, Ts;
86 Tn = ri[WS(rs, 4)];
87 Tq = ii[WS(rs, 4)];
88 Tm = W[6];
89 To = Tm * Tn;
90 TK = Tm * Tq;
91 Tt = ri[WS(rs, 1)];
92 Tw = ii[WS(rs, 1)];
93 Ts = W[0];
94 Tu = Ts * Tt;
95 TM = Ts * Tw;
96 {
97 E Tr, TL, Tx, TN, Tp, Tv;
98 Tp = W[7];
99 Tr = FMA(Tp, Tq, To);
100 TL = FNMS(Tp, Tn, TK);
101 Tv = W[1];
102 Tx = FMA(Tv, Tw, Tu);
103 TN = FNMS(Tv, Tt, TM);
104 Ty = Tr - Tx;
105 TS = TL + TN;
106 TC = Tr + Tx;
107 TO = TL - TN;
108 }
109 }
110 {
111 E TP, T8, Tz, TE;
112 TP = TJ - TO;
113 T8 = T1 - T7;
114 Tz = Tl + Ty;
115 TE = FNMS(KP500000000, Tz, T8);
116 ri[WS(rs, 3)] = T8 + Tz;
117 ri[WS(rs, 1)] = FMA(KP866025403, TP, TE);
118 ri[WS(rs, 5)] = FNMS(KP866025403, TP, TE);
119 }
120 {
121 E T14, T11, T12, T13;
122 T14 = Ty - Tl;
123 T11 = TX - TW;
124 T12 = TJ + TO;
125 T13 = FNMS(KP500000000, T12, T11);
126 ii[WS(rs, 1)] = FMA(KP866025403, T14, T13);
127 ii[WS(rs, 3)] = T12 + T11;
128 ii[WS(rs, 5)] = FNMS(KP866025403, T14, T13);
129 }
130 {
131 E TT, TA, TD, TQ;
132 TT = TR - TS;
133 TA = T1 + T7;
134 TD = TB + TC;
135 TQ = FNMS(KP500000000, TD, TA);
136 ri[0] = TA + TD;
137 ri[WS(rs, 4)] = FMA(KP866025403, TT, TQ);
138 ri[WS(rs, 2)] = FNMS(KP866025403, TT, TQ);
139 }
140 {
141 E T10, TU, TY, TZ;
142 T10 = TC - TB;
143 TU = TR + TS;
144 TY = TW + TX;
145 TZ = FNMS(KP500000000, TU, TY);
146 ii[0] = TU + TY;
147 ii[WS(rs, 4)] = FMA(KP866025403, T10, TZ);
148 ii[WS(rs, 2)] = FNMS(KP866025403, T10, TZ);
149 }
150 }
151 }
152 }
153
154 static const tw_instr twinstr[] = {
155 {TW_FULL, 0, 6},
156 {TW_NEXT, 1, 0}
157 };
158
159 static const ct_desc desc = { 6, "t1_6", twinstr, &GENUS, {24, 10, 22, 0}, 0, 0, 0 };
160
161 void X(codelet_t1_6) (planner *p) {
162 X(kdft_dit_register) (p, t1_6, &desc);
163 }
164 #else
165
166 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 6 -name t1_6 -include dft/scalar/t.h */
167
168 /*
169 * This function contains 46 FP additions, 28 FP multiplications,
170 * (or, 32 additions, 14 multiplications, 14 fused multiply/add),
171 * 23 stack variables, 2 constants, and 24 memory accesses
172 */
173 #include "dft/scalar/t.h"
174
175 static void t1_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
176 {
177 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
178 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
179 {
180 INT m;
181 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)) {
182 E T7, TS, Tv, TO, Tt, TJ, Tx, TF, Ti, TI, Tw, TC;
183 {
184 E T1, TN, T6, TM;
185 T1 = ri[0];
186 TN = ii[0];
187 {
188 E T3, T5, T2, T4;
189 T3 = ri[WS(rs, 3)];
190 T5 = ii[WS(rs, 3)];
191 T2 = W[4];
192 T4 = W[5];
193 T6 = FMA(T2, T3, T4 * T5);
194 TM = FNMS(T4, T3, T2 * T5);
195 }
196 T7 = T1 - T6;
197 TS = TN - TM;
198 Tv = T1 + T6;
199 TO = TM + TN;
200 }
201 {
202 E Tn, TD, Ts, TE;
203 {
204 E Tk, Tm, Tj, Tl;
205 Tk = ri[WS(rs, 4)];
206 Tm = ii[WS(rs, 4)];
207 Tj = W[6];
208 Tl = W[7];
209 Tn = FMA(Tj, Tk, Tl * Tm);
210 TD = FNMS(Tl, Tk, Tj * Tm);
211 }
212 {
213 E Tp, Tr, To, Tq;
214 Tp = ri[WS(rs, 1)];
215 Tr = ii[WS(rs, 1)];
216 To = W[0];
217 Tq = W[1];
218 Ts = FMA(To, Tp, Tq * Tr);
219 TE = FNMS(Tq, Tp, To * Tr);
220 }
221 Tt = Tn - Ts;
222 TJ = TD + TE;
223 Tx = Tn + Ts;
224 TF = TD - TE;
225 }
226 {
227 E Tc, TA, Th, TB;
228 {
229 E T9, Tb, T8, Ta;
230 T9 = ri[WS(rs, 2)];
231 Tb = ii[WS(rs, 2)];
232 T8 = W[2];
233 Ta = W[3];
234 Tc = FMA(T8, T9, Ta * Tb);
235 TA = FNMS(Ta, T9, T8 * Tb);
236 }
237 {
238 E Te, Tg, Td, Tf;
239 Te = ri[WS(rs, 5)];
240 Tg = ii[WS(rs, 5)];
241 Td = W[8];
242 Tf = W[9];
243 Th = FMA(Td, Te, Tf * Tg);
244 TB = FNMS(Tf, Te, Td * Tg);
245 }
246 Ti = Tc - Th;
247 TI = TA + TB;
248 Tw = Tc + Th;
249 TC = TA - TB;
250 }
251 {
252 E TG, Tu, Tz, TR, TT, TU;
253 TG = KP866025403 * (TC - TF);
254 Tu = Ti + Tt;
255 Tz = FNMS(KP500000000, Tu, T7);
256 ri[WS(rs, 3)] = T7 + Tu;
257 ri[WS(rs, 1)] = Tz + TG;
258 ri[WS(rs, 5)] = Tz - TG;
259 TR = KP866025403 * (Tt - Ti);
260 TT = TC + TF;
261 TU = FNMS(KP500000000, TT, TS);
262 ii[WS(rs, 1)] = TR + TU;
263 ii[WS(rs, 3)] = TT + TS;
264 ii[WS(rs, 5)] = TU - TR;
265 }
266 {
267 E TK, Ty, TH, TQ, TL, TP;
268 TK = KP866025403 * (TI - TJ);
269 Ty = Tw + Tx;
270 TH = FNMS(KP500000000, Ty, Tv);
271 ri[0] = Tv + Ty;
272 ri[WS(rs, 4)] = TH + TK;
273 ri[WS(rs, 2)] = TH - TK;
274 TQ = KP866025403 * (Tx - Tw);
275 TL = TI + TJ;
276 TP = FNMS(KP500000000, TL, TO);
277 ii[0] = TL + TO;
278 ii[WS(rs, 4)] = TQ + TP;
279 ii[WS(rs, 2)] = TP - TQ;
280 }
281 }
282 }
283 }
284
285 static const tw_instr twinstr[] = {
286 {TW_FULL, 0, 6},
287 {TW_NEXT, 1, 0}
288 };
289
290 static const ct_desc desc = { 6, "t1_6", twinstr, &GENUS, {32, 14, 14, 0}, 0, 0, 0 };
291
292 void X(codelet_t1_6) (planner *p) {
293 X(kdft_dit_register) (p, t1_6, &desc);
294 }
295 #endif