Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2sv_8.c @ 10:37bf6b4a2645

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
9:c0fb53affa76 10:37bf6b4a2645
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 07:37:47 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2sv_8 -with-ostride 1 -include 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 * 58 stack variables, 1 constants, and 36 memory accesses
34 */
35 #include "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 TF, TJ, TD, TR, TS, TT, TU, TV, TW, TE, TX, TY, TK, TI, TZ;
44 V T10, T11, T12;
45 {
46 V Tb, Tn, T3, TC, Ti, TB, T6, To, Tl, Tc, Tw, Tx, T8, T9, Tr;
47 V Ts;
48 {
49 V T1, T2, Tg, Th, T4, T5, Tj, Tk;
50 T1 = LD(&(ri[0]), ivs, &(ri[0]));
51 T2 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
52 Tg = LD(&(ii[0]), ivs, &(ii[0]));
53 Th = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
54 T4 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
55 T5 = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
56 Tj = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
57 Tk = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
58 Tb = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
59 Tn = VSUB(T1, T2);
60 T3 = VADD(T1, T2);
61 TC = VSUB(Tg, Th);
62 Ti = VADD(Tg, Th);
63 TB = VSUB(T4, T5);
64 T6 = VADD(T4, T5);
65 To = VSUB(Tj, Tk);
66 Tl = VADD(Tj, Tk);
67 Tc = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
68 Tw = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
69 Tx = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
70 T8 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
71 T9 = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
72 Tr = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
73 Ts = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
74 }
75 {
76 V TL, T7, TP, Tm, Tz, TH, Te, Tf, TO, TQ, TG, Tu, Tp, TA;
77 {
78 V Td, Tv, TN, Ty, Ta, Tq, TM, Tt;
79 TL = VSUB(T3, T6);
80 T7 = VADD(T3, T6);
81 Td = VADD(Tb, Tc);
82 Tv = VSUB(Tb, Tc);
83 TN = VADD(Tw, Tx);
84 Ty = VSUB(Tw, Tx);
85 Ta = VADD(T8, T9);
86 Tq = VSUB(T8, T9);
87 TM = VADD(Tr, Ts);
88 Tt = VSUB(Tr, Ts);
89 TP = VADD(Ti, Tl);
90 Tm = VSUB(Ti, Tl);
91 Tz = VSUB(Tv, Ty);
92 TH = VADD(Tv, Ty);
93 Te = VADD(Ta, Td);
94 Tf = VSUB(Td, Ta);
95 TO = VSUB(TM, TN);
96 TQ = VADD(TM, TN);
97 TG = VSUB(Tt, Tq);
98 Tu = VADD(Tq, Tt);
99 }
100 TF = VSUB(Tn, To);
101 Tp = VADD(Tn, To);
102 TJ = VSUB(TC, TB);
103 TD = VADD(TB, TC);
104 TR = VSUB(Tm, Tf);
105 STM4(&(io[6]), TR, ovs, &(io[0]));
106 TS = VADD(Tf, Tm);
107 STM4(&(io[2]), TS, ovs, &(io[0]));
108 TT = VADD(T7, Te);
109 STM4(&(ro[0]), TT, ovs, &(ro[0]));
110 TU = VSUB(T7, Te);
111 STM4(&(ro[4]), TU, ovs, &(ro[0]));
112 TV = VADD(TP, TQ);
113 STM4(&(io[0]), TV, ovs, &(io[0]));
114 TW = VSUB(TP, TQ);
115 STM4(&(io[4]), TW, ovs, &(io[0]));
116 TE = VSUB(Tz, Tu);
117 TA = VADD(Tu, Tz);
118 TX = VADD(TL, TO);
119 STM4(&(ro[2]), TX, ovs, &(ro[0]));
120 TY = VSUB(TL, TO);
121 STM4(&(ro[6]), TY, ovs, &(ro[0]));
122 TK = VADD(TG, TH);
123 TI = VSUB(TG, TH);
124 TZ = VFMA(LDK(KP707106781), TA, Tp);
125 STM4(&(ro[1]), TZ, ovs, &(ro[1]));
126 T10 = VFNMS(LDK(KP707106781), TA, Tp);
127 STM4(&(ro[5]), T10, ovs, &(ro[1]));
128 }
129 }
130 T11 = VFMA(LDK(KP707106781), TK, TJ);
131 STM4(&(io[1]), T11, ovs, &(io[1]));
132 T12 = VFNMS(LDK(KP707106781), TK, TJ);
133 STM4(&(io[5]), T12, ovs, &(io[1]));
134 {
135 V T13, T14, T15, T16;
136 T13 = VFMA(LDK(KP707106781), TE, TD);
137 STM4(&(io[3]), T13, ovs, &(io[1]));
138 STN4(&(io[0]), TV, T11, TS, T13, ovs);
139 T14 = VFNMS(LDK(KP707106781), TE, TD);
140 STM4(&(io[7]), T14, ovs, &(io[1]));
141 STN4(&(io[4]), TW, T12, TR, T14, ovs);
142 T15 = VFMA(LDK(KP707106781), TI, TF);
143 STM4(&(ro[3]), T15, ovs, &(ro[1]));
144 STN4(&(ro[0]), TT, TZ, TX, T15, ovs);
145 T16 = VFNMS(LDK(KP707106781), TI, TF);
146 STM4(&(ro[7]), T16, ovs, &(ro[1]));
147 STN4(&(ro[4]), TU, T10, TY, T16, ovs);
148 }
149 }
150 }
151 VLEAVE();
152 }
153
154 static const kdft_desc desc = { 8, XSIMD_STRING("n2sv_8"), {44, 0, 8, 0}, &GENUS, 0, 1, 0, 0 };
155
156 void XSIMD(codelet_n2sv_8) (planner *p) {
157 X(kdft_register) (p, n2sv_8, &desc);
158 }
159
160 #else /* HAVE_FMA */
161
162 /* Generated by: ../../../genfft/gen_notw.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2sv_8 -with-ostride 1 -include n2s.h -store-multiple 4 */
163
164 /*
165 * This function contains 52 FP additions, 4 FP multiplications,
166 * (or, 52 additions, 4 multiplications, 0 fused multiply/add),
167 * 34 stack variables, 1 constants, and 36 memory accesses
168 */
169 #include "n2s.h"
170
171 static void n2sv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
172 {
173 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
174 {
175 INT i;
176 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)) {
177 V T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu;
178 V TG;
179 {
180 V T1, T2, Tj, Tk;
181 T1 = LD(&(ri[0]), ivs, &(ri[0]));
182 T2 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
183 T3 = VADD(T1, T2);
184 Tn = VSUB(T1, T2);
185 {
186 V Tg, Th, T4, T5;
187 Tg = LD(&(ii[0]), ivs, &(ii[0]));
188 Th = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
189 Ti = VADD(Tg, Th);
190 TC = VSUB(Tg, Th);
191 T4 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
192 T5 = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
193 T6 = VADD(T4, T5);
194 TB = VSUB(T4, T5);
195 }
196 Tj = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
197 Tk = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
198 Tl = VADD(Tj, Tk);
199 To = VSUB(Tj, Tk);
200 {
201 V Tb, Tc, Tv, Tw, Tx, Ty;
202 Tb = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
203 Tc = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
204 Tv = VSUB(Tb, Tc);
205 Tw = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
206 Tx = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
207 Ty = VSUB(Tw, Tx);
208 Td = VADD(Tb, Tc);
209 TN = VADD(Tw, Tx);
210 Tz = VSUB(Tv, Ty);
211 TH = VADD(Tv, Ty);
212 }
213 {
214 V T8, T9, Tq, Tr, Ts, Tt;
215 T8 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
216 T9 = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
217 Tq = VSUB(T8, T9);
218 Tr = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
219 Ts = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
220 Tt = VSUB(Tr, Ts);
221 Ta = VADD(T8, T9);
222 TM = VADD(Tr, Ts);
223 Tu = VADD(Tq, Tt);
224 TG = VSUB(Tt, Tq);
225 }
226 }
227 {
228 V TR, TS, TT, TU, TV, TW, TX, TY;
229 {
230 V T7, Te, TP, TQ;
231 T7 = VADD(T3, T6);
232 Te = VADD(Ta, Td);
233 TR = VSUB(T7, Te);
234 STM4(&(ro[4]), TR, ovs, &(ro[0]));
235 TS = VADD(T7, Te);
236 STM4(&(ro[0]), TS, ovs, &(ro[0]));
237 TP = VADD(Ti, Tl);
238 TQ = VADD(TM, TN);
239 TT = VSUB(TP, TQ);
240 STM4(&(io[4]), TT, ovs, &(io[0]));
241 TU = VADD(TP, TQ);
242 STM4(&(io[0]), TU, ovs, &(io[0]));
243 }
244 {
245 V Tf, Tm, TL, TO;
246 Tf = VSUB(Td, Ta);
247 Tm = VSUB(Ti, Tl);
248 TV = VADD(Tf, Tm);
249 STM4(&(io[2]), TV, ovs, &(io[0]));
250 TW = VSUB(Tm, Tf);
251 STM4(&(io[6]), TW, ovs, &(io[0]));
252 TL = VSUB(T3, T6);
253 TO = VSUB(TM, TN);
254 TX = VSUB(TL, TO);
255 STM4(&(ro[6]), TX, ovs, &(ro[0]));
256 TY = VADD(TL, TO);
257 STM4(&(ro[2]), TY, ovs, &(ro[0]));
258 }
259 {
260 V TZ, T10, T11, T12;
261 {
262 V Tp, TA, TJ, TK;
263 Tp = VADD(Tn, To);
264 TA = VMUL(LDK(KP707106781), VADD(Tu, Tz));
265 TZ = VSUB(Tp, TA);
266 STM4(&(ro[5]), TZ, ovs, &(ro[1]));
267 T10 = VADD(Tp, TA);
268 STM4(&(ro[1]), T10, ovs, &(ro[1]));
269 TJ = VSUB(TC, TB);
270 TK = VMUL(LDK(KP707106781), VADD(TG, TH));
271 T11 = VSUB(TJ, TK);
272 STM4(&(io[5]), T11, ovs, &(io[1]));
273 T12 = VADD(TJ, TK);
274 STM4(&(io[1]), T12, ovs, &(io[1]));
275 }
276 {
277 V TD, TE, T13, T14;
278 TD = VADD(TB, TC);
279 TE = VMUL(LDK(KP707106781), VSUB(Tz, Tu));
280 T13 = VSUB(TD, TE);
281 STM4(&(io[7]), T13, ovs, &(io[1]));
282 STN4(&(io[4]), TT, T11, TW, T13, ovs);
283 T14 = VADD(TD, TE);
284 STM4(&(io[3]), T14, ovs, &(io[1]));
285 STN4(&(io[0]), TU, T12, TV, T14, ovs);
286 }
287 {
288 V TF, TI, T15, T16;
289 TF = VSUB(Tn, To);
290 TI = VMUL(LDK(KP707106781), VSUB(TG, TH));
291 T15 = VSUB(TF, TI);
292 STM4(&(ro[7]), T15, ovs, &(ro[1]));
293 STN4(&(ro[4]), TR, TZ, TX, T15, ovs);
294 T16 = VADD(TF, TI);
295 STM4(&(ro[3]), T16, ovs, &(ro[1]));
296 STN4(&(ro[0]), TS, T10, TY, T16, ovs);
297 }
298 }
299 }
300 }
301 }
302 VLEAVE();
303 }
304
305 static const kdft_desc desc = { 8, XSIMD_STRING("n2sv_8"), {52, 4, 0, 0}, &GENUS, 0, 1, 0, 0 };
306
307 void XSIMD(codelet_n2sv_8) (planner *p) {
308 X(kdft_register) (p, n2sv_8, &desc);
309 }
310
311 #endif /* HAVE_FMA */