Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cb/r2cb_16.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 08:07:29 EDT 2018 */
23
24 #include "rdft/codelet-rdft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_r2cb.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -name r2cb_16 -include rdft/scalar/r2cb.h */
29
30 /*
31 * This function contains 58 FP additions, 32 FP multiplications,
32 * (or, 26 additions, 0 multiplications, 32 fused multiply/add),
33 * 31 stack variables, 4 constants, and 32 memory accesses
34 */
35 #include "rdft/scalar/r2cb.h"
36
37 static void r2cb_16(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
38 {
39 DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
40 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
41 DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
42 DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
43 {
44 INT i;
45 for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
46 E T5, TL, Tj, TD, T8, TM, To, TE, Tc, TP, Tf, TQ, Tu, Tz, TR;
47 E TO, TH, TG;
48 {
49 E T4, Ti, T3, Th, T1, T2;
50 T4 = Cr[WS(csr, 4)];
51 Ti = Ci[WS(csi, 4)];
52 T1 = Cr[0];
53 T2 = Cr[WS(csr, 8)];
54 T3 = T1 + T2;
55 Th = T1 - T2;
56 T5 = FMA(KP2_000000000, T4, T3);
57 TL = FNMS(KP2_000000000, T4, T3);
58 Tj = FNMS(KP2_000000000, Ti, Th);
59 TD = FMA(KP2_000000000, Ti, Th);
60 }
61 {
62 E T6, T7, Tk, Tl, Tm, Tn;
63 T6 = Cr[WS(csr, 2)];
64 T7 = Cr[WS(csr, 6)];
65 Tk = T6 - T7;
66 Tl = Ci[WS(csi, 2)];
67 Tm = Ci[WS(csi, 6)];
68 Tn = Tl + Tm;
69 T8 = T6 + T7;
70 TM = Tl - Tm;
71 To = Tk - Tn;
72 TE = Tk + Tn;
73 }
74 {
75 E Tq, Ty, Tv, Tt;
76 {
77 E Ta, Tb, Tw, Tx;
78 Ta = Cr[WS(csr, 1)];
79 Tb = Cr[WS(csr, 7)];
80 Tc = Ta + Tb;
81 Tq = Ta - Tb;
82 Tw = Ci[WS(csi, 1)];
83 Tx = Ci[WS(csi, 7)];
84 Ty = Tw + Tx;
85 TP = Tw - Tx;
86 }
87 {
88 E Td, Te, Tr, Ts;
89 Td = Cr[WS(csr, 5)];
90 Te = Cr[WS(csr, 3)];
91 Tf = Td + Te;
92 Tv = Td - Te;
93 Tr = Ci[WS(csi, 5)];
94 Ts = Ci[WS(csi, 3)];
95 Tt = Tr + Ts;
96 TQ = Tr - Ts;
97 }
98 Tu = Tq - Tt;
99 Tz = Tv + Ty;
100 TR = TP - TQ;
101 TO = Tc - Tf;
102 TH = Tq + Tt;
103 TG = Ty - Tv;
104 }
105 {
106 E T9, Tg, TT, TU;
107 T9 = FMA(KP2_000000000, T8, T5);
108 Tg = Tc + Tf;
109 R0[WS(rs, 4)] = FNMS(KP2_000000000, Tg, T9);
110 R0[0] = FMA(KP2_000000000, Tg, T9);
111 TT = FMA(KP2_000000000, TM, TL);
112 TU = TO + TR;
113 R0[WS(rs, 3)] = FNMS(KP1_414213562, TU, TT);
114 R0[WS(rs, 7)] = FMA(KP1_414213562, TU, TT);
115 }
116 {
117 E TV, TW, Tp, TA;
118 TV = FNMS(KP2_000000000, T8, T5);
119 TW = TQ + TP;
120 R0[WS(rs, 2)] = FNMS(KP2_000000000, TW, TV);
121 R0[WS(rs, 6)] = FMA(KP2_000000000, TW, TV);
122 Tp = FMA(KP1_414213562, To, Tj);
123 TA = FNMS(KP414213562, Tz, Tu);
124 R1[WS(rs, 4)] = FNMS(KP1_847759065, TA, Tp);
125 R1[0] = FMA(KP1_847759065, TA, Tp);
126 }
127 {
128 E TB, TC, TJ, TK;
129 TB = FNMS(KP1_414213562, To, Tj);
130 TC = FMA(KP414213562, Tu, Tz);
131 R1[WS(rs, 2)] = FNMS(KP1_847759065, TC, TB);
132 R1[WS(rs, 6)] = FMA(KP1_847759065, TC, TB);
133 TJ = FMA(KP1_414213562, TE, TD);
134 TK = FMA(KP414213562, TG, TH);
135 R1[WS(rs, 3)] = FNMS(KP1_847759065, TK, TJ);
136 R1[WS(rs, 7)] = FMA(KP1_847759065, TK, TJ);
137 }
138 {
139 E TN, TS, TF, TI;
140 TN = FNMS(KP2_000000000, TM, TL);
141 TS = TO - TR;
142 R0[WS(rs, 5)] = FNMS(KP1_414213562, TS, TN);
143 R0[WS(rs, 1)] = FMA(KP1_414213562, TS, TN);
144 TF = FNMS(KP1_414213562, TE, TD);
145 TI = FNMS(KP414213562, TH, TG);
146 R1[WS(rs, 1)] = FNMS(KP1_847759065, TI, TF);
147 R1[WS(rs, 5)] = FMA(KP1_847759065, TI, TF);
148 }
149 }
150 }
151 }
152
153 static const kr2c_desc desc = { 16, "r2cb_16", {26, 0, 32, 0}, &GENUS };
154
155 void X(codelet_r2cb_16) (planner *p) {
156 X(kr2c_register) (p, r2cb_16, &desc);
157 }
158
159 #else
160
161 /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 16 -name r2cb_16 -include rdft/scalar/r2cb.h */
162
163 /*
164 * This function contains 58 FP additions, 18 FP multiplications,
165 * (or, 54 additions, 14 multiplications, 4 fused multiply/add),
166 * 31 stack variables, 4 constants, and 32 memory accesses
167 */
168 #include "rdft/scalar/r2cb.h"
169
170 static void r2cb_16(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
171 {
172 DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
173 DK(KP765366864, +0.765366864730179543456919968060797733522689125);
174 DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
175 DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
176 {
177 INT i;
178 for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(64, rs), MAKE_VOLATILE_STRIDE(64, csr), MAKE_VOLATILE_STRIDE(64, csi)) {
179 E T9, TS, Tl, TG, T6, TR, Ti, TD, Td, Tq, Tg, Tt, Tn, Tu, TV;
180 E TU, TN, TK;
181 {
182 E T7, T8, TE, Tj, Tk, TF;
183 T7 = Cr[WS(csr, 2)];
184 T8 = Cr[WS(csr, 6)];
185 TE = T7 - T8;
186 Tj = Ci[WS(csi, 2)];
187 Tk = Ci[WS(csi, 6)];
188 TF = Tj + Tk;
189 T9 = KP2_000000000 * (T7 + T8);
190 TS = KP1_414213562 * (TE + TF);
191 Tl = KP2_000000000 * (Tj - Tk);
192 TG = KP1_414213562 * (TE - TF);
193 }
194 {
195 E T5, TC, T3, TA;
196 {
197 E T4, TB, T1, T2;
198 T4 = Cr[WS(csr, 4)];
199 T5 = KP2_000000000 * T4;
200 TB = Ci[WS(csi, 4)];
201 TC = KP2_000000000 * TB;
202 T1 = Cr[0];
203 T2 = Cr[WS(csr, 8)];
204 T3 = T1 + T2;
205 TA = T1 - T2;
206 }
207 T6 = T3 + T5;
208 TR = TA + TC;
209 Ti = T3 - T5;
210 TD = TA - TC;
211 }
212 {
213 E TI, TM, TL, TJ;
214 {
215 E Tb, Tc, To, Tp;
216 Tb = Cr[WS(csr, 1)];
217 Tc = Cr[WS(csr, 7)];
218 Td = Tb + Tc;
219 TI = Tb - Tc;
220 To = Ci[WS(csi, 1)];
221 Tp = Ci[WS(csi, 7)];
222 Tq = To - Tp;
223 TM = To + Tp;
224 }
225 {
226 E Te, Tf, Tr, Ts;
227 Te = Cr[WS(csr, 5)];
228 Tf = Cr[WS(csr, 3)];
229 Tg = Te + Tf;
230 TL = Te - Tf;
231 Tr = Ci[WS(csi, 5)];
232 Ts = Ci[WS(csi, 3)];
233 Tt = Tr - Ts;
234 TJ = Tr + Ts;
235 }
236 Tn = Td - Tg;
237 Tu = Tq - Tt;
238 TV = TM - TL;
239 TU = TI + TJ;
240 TN = TL + TM;
241 TK = TI - TJ;
242 }
243 {
244 E Ta, Th, TT, TW;
245 Ta = T6 + T9;
246 Th = KP2_000000000 * (Td + Tg);
247 R0[WS(rs, 4)] = Ta - Th;
248 R0[0] = Ta + Th;
249 TT = TR - TS;
250 TW = FNMS(KP1_847759065, TV, KP765366864 * TU);
251 R1[WS(rs, 5)] = TT - TW;
252 R1[WS(rs, 1)] = TT + TW;
253 }
254 {
255 E TX, TY, Tm, Tv;
256 TX = TR + TS;
257 TY = FMA(KP1_847759065, TU, KP765366864 * TV);
258 R1[WS(rs, 3)] = TX - TY;
259 R1[WS(rs, 7)] = TX + TY;
260 Tm = Ti - Tl;
261 Tv = KP1_414213562 * (Tn - Tu);
262 R0[WS(rs, 5)] = Tm - Tv;
263 R0[WS(rs, 1)] = Tm + Tv;
264 }
265 {
266 E Tw, Tx, TH, TO;
267 Tw = Ti + Tl;
268 Tx = KP1_414213562 * (Tn + Tu);
269 R0[WS(rs, 3)] = Tw - Tx;
270 R0[WS(rs, 7)] = Tw + Tx;
271 TH = TD + TG;
272 TO = FNMS(KP765366864, TN, KP1_847759065 * TK);
273 R1[WS(rs, 4)] = TH - TO;
274 R1[0] = TH + TO;
275 }
276 {
277 E TP, TQ, Ty, Tz;
278 TP = TD - TG;
279 TQ = FMA(KP765366864, TK, KP1_847759065 * TN);
280 R1[WS(rs, 2)] = TP - TQ;
281 R1[WS(rs, 6)] = TP + TQ;
282 Ty = T6 - T9;
283 Tz = KP2_000000000 * (Tt + Tq);
284 R0[WS(rs, 2)] = Ty - Tz;
285 R0[WS(rs, 6)] = Ty + Tz;
286 }
287 }
288 }
289 }
290
291 static const kr2c_desc desc = { 16, "r2cb_16", {54, 14, 4, 0}, &GENUS };
292
293 void X(codelet_r2cb_16) (planner *p) {
294 X(kr2c_register) (p, r2cb_16, &desc);
295 }
296
297 #endif