Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n2sv_8.c @ 82:d0c2a83c1364

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
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
81:7029a4916348 82:d0c2a83c1364
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 08:05:19 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_notw.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2sv_8 -with-ostride 1 -include dft/simd/n2s.h -store-multiple 4 */
29
30 /*
31 * This function contains 52 FP additions, 8 FP multiplications,
32 * (or, 44 additions, 0 multiplications, 8 fused multiply/add),
33 * 34 stack variables, 1 constants, and 36 memory accesses
34 */
35 #include "dft/simd/n2s.h"
36
37 static void n2sv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT i;
42 for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
43 V T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu;
44 V TG;
45 {
46 V T1, T2, Tj, Tk;
47 T1 = LD(&(ri[0]), ivs, &(ri[0]));
48 T2 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
49 T3 = VADD(T1, T2);
50 Tn = VSUB(T1, T2);
51 {
52 V Tg, Th, T4, T5;
53 Tg = LD(&(ii[0]), ivs, &(ii[0]));
54 Th = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
55 Ti = VADD(Tg, Th);
56 TC = VSUB(Tg, Th);
57 T4 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
58 T5 = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
59 T6 = VADD(T4, T5);
60 TB = VSUB(T4, T5);
61 }
62 Tj = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
63 Tk = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
64 Tl = VADD(Tj, Tk);
65 To = VSUB(Tj, Tk);
66 {
67 V Tb, Tc, Tv, Tw, Tx, Ty;
68 Tb = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
69 Tc = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
70 Tv = VSUB(Tb, Tc);
71 Tw = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
72 Tx = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
73 Ty = VSUB(Tw, Tx);
74 Td = VADD(Tb, Tc);
75 TN = VADD(Tw, Tx);
76 Tz = VSUB(Tv, Ty);
77 TH = VADD(Tv, Ty);
78 }
79 {
80 V T8, T9, Tq, Tr, Ts, Tt;
81 T8 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
82 T9 = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
83 Tq = VSUB(T8, T9);
84 Tr = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
85 Ts = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
86 Tt = VSUB(Tr, Ts);
87 Ta = VADD(T8, T9);
88 TM = VADD(Tr, Ts);
89 Tu = VADD(Tq, Tt);
90 TG = VSUB(Tt, Tq);
91 }
92 }
93 {
94 V TR, TS, TT, TU, TV, TW, TX, TY;
95 {
96 V T7, Te, TP, TQ;
97 T7 = VADD(T3, T6);
98 Te = VADD(Ta, Td);
99 TR = VSUB(T7, Te);
100 STM4(&(ro[4]), TR, ovs, &(ro[0]));
101 TS = VADD(T7, Te);
102 STM4(&(ro[0]), TS, ovs, &(ro[0]));
103 TP = VADD(Ti, Tl);
104 TQ = VADD(TM, TN);
105 TT = VSUB(TP, TQ);
106 STM4(&(io[4]), TT, ovs, &(io[0]));
107 TU = VADD(TP, TQ);
108 STM4(&(io[0]), TU, ovs, &(io[0]));
109 }
110 {
111 V Tf, Tm, TL, TO;
112 Tf = VSUB(Td, Ta);
113 Tm = VSUB(Ti, Tl);
114 TV = VADD(Tf, Tm);
115 STM4(&(io[2]), TV, ovs, &(io[0]));
116 TW = VSUB(Tm, Tf);
117 STM4(&(io[6]), TW, ovs, &(io[0]));
118 TL = VSUB(T3, T6);
119 TO = VSUB(TM, TN);
120 TX = VSUB(TL, TO);
121 STM4(&(ro[6]), TX, ovs, &(ro[0]));
122 TY = VADD(TL, TO);
123 STM4(&(ro[2]), TY, ovs, &(ro[0]));
124 }
125 {
126 V TZ, T10, T11, T12;
127 {
128 V Tp, TA, TJ, TK;
129 Tp = VADD(Tn, To);
130 TA = VADD(Tu, Tz);
131 TZ = VFNMS(LDK(KP707106781), TA, Tp);
132 STM4(&(ro[5]), TZ, ovs, &(ro[1]));
133 T10 = VFMA(LDK(KP707106781), TA, Tp);
134 STM4(&(ro[1]), T10, ovs, &(ro[1]));
135 TJ = VSUB(TC, TB);
136 TK = VADD(TG, TH);
137 T11 = VFNMS(LDK(KP707106781), TK, TJ);
138 STM4(&(io[5]), T11, ovs, &(io[1]));
139 T12 = VFMA(LDK(KP707106781), TK, TJ);
140 STM4(&(io[1]), T12, ovs, &(io[1]));
141 }
142 {
143 V TD, TE, T13, T14;
144 TD = VADD(TB, TC);
145 TE = VSUB(Tz, Tu);
146 T13 = VFNMS(LDK(KP707106781), TE, TD);
147 STM4(&(io[7]), T13, ovs, &(io[1]));
148 STN4(&(io[4]), TT, T11, TW, T13, ovs);
149 T14 = VFMA(LDK(KP707106781), TE, TD);
150 STM4(&(io[3]), T14, ovs, &(io[1]));
151 STN4(&(io[0]), TU, T12, TV, T14, ovs);
152 }
153 {
154 V TF, TI, T15, T16;
155 TF = VSUB(Tn, To);
156 TI = VSUB(TG, TH);
157 T15 = VFNMS(LDK(KP707106781), TI, TF);
158 STM4(&(ro[7]), T15, ovs, &(ro[1]));
159 STN4(&(ro[4]), TR, TZ, TX, T15, ovs);
160 T16 = VFMA(LDK(KP707106781), TI, TF);
161 STM4(&(ro[3]), T16, ovs, &(ro[1]));
162 STN4(&(ro[0]), TS, T10, TY, T16, ovs);
163 }
164 }
165 }
166 }
167 }
168 VLEAVE();
169 }
170
171 static const kdft_desc desc = { 8, XSIMD_STRING("n2sv_8"), {44, 0, 8, 0}, &GENUS, 0, 1, 0, 0 };
172
173 void XSIMD(codelet_n2sv_8) (planner *p) {
174 X(kdft_register) (p, n2sv_8, &desc);
175 }
176
177 #else
178
179 /* Generated by: ../../../genfft/gen_notw.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2sv_8 -with-ostride 1 -include dft/simd/n2s.h -store-multiple 4 */
180
181 /*
182 * This function contains 52 FP additions, 4 FP multiplications,
183 * (or, 52 additions, 4 multiplications, 0 fused multiply/add),
184 * 34 stack variables, 1 constants, and 36 memory accesses
185 */
186 #include "dft/simd/n2s.h"
187
188 static void n2sv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
189 {
190 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
191 {
192 INT i;
193 for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
194 V T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu;
195 V TG;
196 {
197 V T1, T2, Tj, Tk;
198 T1 = LD(&(ri[0]), ivs, &(ri[0]));
199 T2 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
200 T3 = VADD(T1, T2);
201 Tn = VSUB(T1, T2);
202 {
203 V Tg, Th, T4, T5;
204 Tg = LD(&(ii[0]), ivs, &(ii[0]));
205 Th = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
206 Ti = VADD(Tg, Th);
207 TC = VSUB(Tg, Th);
208 T4 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
209 T5 = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
210 T6 = VADD(T4, T5);
211 TB = VSUB(T4, T5);
212 }
213 Tj = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
214 Tk = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
215 Tl = VADD(Tj, Tk);
216 To = VSUB(Tj, Tk);
217 {
218 V Tb, Tc, Tv, Tw, Tx, Ty;
219 Tb = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
220 Tc = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
221 Tv = VSUB(Tb, Tc);
222 Tw = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
223 Tx = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
224 Ty = VSUB(Tw, Tx);
225 Td = VADD(Tb, Tc);
226 TN = VADD(Tw, Tx);
227 Tz = VSUB(Tv, Ty);
228 TH = VADD(Tv, Ty);
229 }
230 {
231 V T8, T9, Tq, Tr, Ts, Tt;
232 T8 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
233 T9 = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
234 Tq = VSUB(T8, T9);
235 Tr = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
236 Ts = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
237 Tt = VSUB(Tr, Ts);
238 Ta = VADD(T8, T9);
239 TM = VADD(Tr, Ts);
240 Tu = VADD(Tq, Tt);
241 TG = VSUB(Tt, Tq);
242 }
243 }
244 {
245 V TR, TS, TT, TU, TV, TW, TX, TY;
246 {
247 V T7, Te, TP, TQ;
248 T7 = VADD(T3, T6);
249 Te = VADD(Ta, Td);
250 TR = VSUB(T7, Te);
251 STM4(&(ro[4]), TR, ovs, &(ro[0]));
252 TS = VADD(T7, Te);
253 STM4(&(ro[0]), TS, ovs, &(ro[0]));
254 TP = VADD(Ti, Tl);
255 TQ = VADD(TM, TN);
256 TT = VSUB(TP, TQ);
257 STM4(&(io[4]), TT, ovs, &(io[0]));
258 TU = VADD(TP, TQ);
259 STM4(&(io[0]), TU, ovs, &(io[0]));
260 }
261 {
262 V Tf, Tm, TL, TO;
263 Tf = VSUB(Td, Ta);
264 Tm = VSUB(Ti, Tl);
265 TV = VADD(Tf, Tm);
266 STM4(&(io[2]), TV, ovs, &(io[0]));
267 TW = VSUB(Tm, Tf);
268 STM4(&(io[6]), TW, ovs, &(io[0]));
269 TL = VSUB(T3, T6);
270 TO = VSUB(TM, TN);
271 TX = VSUB(TL, TO);
272 STM4(&(ro[6]), TX, ovs, &(ro[0]));
273 TY = VADD(TL, TO);
274 STM4(&(ro[2]), TY, ovs, &(ro[0]));
275 }
276 {
277 V TZ, T10, T11, T12;
278 {
279 V Tp, TA, TJ, TK;
280 Tp = VADD(Tn, To);
281 TA = VMUL(LDK(KP707106781), VADD(Tu, Tz));
282 TZ = VSUB(Tp, TA);
283 STM4(&(ro[5]), TZ, ovs, &(ro[1]));
284 T10 = VADD(Tp, TA);
285 STM4(&(ro[1]), T10, ovs, &(ro[1]));
286 TJ = VSUB(TC, TB);
287 TK = VMUL(LDK(KP707106781), VADD(TG, TH));
288 T11 = VSUB(TJ, TK);
289 STM4(&(io[5]), T11, ovs, &(io[1]));
290 T12 = VADD(TJ, TK);
291 STM4(&(io[1]), T12, ovs, &(io[1]));
292 }
293 {
294 V TD, TE, T13, T14;
295 TD = VADD(TB, TC);
296 TE = VMUL(LDK(KP707106781), VSUB(Tz, Tu));
297 T13 = VSUB(TD, TE);
298 STM4(&(io[7]), T13, ovs, &(io[1]));
299 STN4(&(io[4]), TT, T11, TW, T13, ovs);
300 T14 = VADD(TD, TE);
301 STM4(&(io[3]), T14, ovs, &(io[1]));
302 STN4(&(io[0]), TU, T12, TV, T14, ovs);
303 }
304 {
305 V TF, TI, T15, T16;
306 TF = VSUB(Tn, To);
307 TI = VMUL(LDK(KP707106781), VSUB(TG, TH));
308 T15 = VSUB(TF, TI);
309 STM4(&(ro[7]), T15, ovs, &(ro[1]));
310 STN4(&(ro[4]), TR, TZ, TX, T15, ovs);
311 T16 = VADD(TF, TI);
312 STM4(&(ro[3]), T16, ovs, &(ro[1]));
313 STN4(&(ro[0]), TS, T10, TY, T16, ovs);
314 }
315 }
316 }
317 }
318 }
319 VLEAVE();
320 }
321
322 static const kdft_desc desc = { 8, XSIMD_STRING("n2sv_8"), {52, 4, 0, 0}, &GENUS, 0, 1, 0, 0 };
323
324 void XSIMD(codelet_n2sv_8) (planner *p) {
325 X(kdft_register) (p, n2sv_8, &desc);
326 }
327
328 #endif