Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2bv_12.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 07:37:30 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 12 -name n2bv_12 -with-ostride 2 -include n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 48 FP additions, 20 FP multiplications,
32 * (or, 30 additions, 2 multiplications, 18 fused multiply/add),
33 * 61 stack variables, 2 constants, and 30 memory accesses
34 */
35 #include "n2b.h"
36
37 static void n2bv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT i;
43 const R *xi;
44 R *xo;
45 xi = ii;
46 xo = io;
47 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(24, is), MAKE_VOLATILE_STRIDE(24, os)) {
48 V T1, T6, Tc, Th, Td, Te, Ti, Tz, T4, TA, T9, Tj, Tf, Tw;
49 {
50 V T2, T3, T7, T8;
51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
52 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
53 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
54 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
55 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
56 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
57 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
58 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
59 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
60 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
61 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
62 Tz = VSUB(T2, T3);
63 T4 = VADD(T2, T3);
64 TA = VSUB(T7, T8);
65 T9 = VADD(T7, T8);
66 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
67 }
68 Tf = VADD(Td, Te);
69 Tw = VSUB(Td, Te);
70 {
71 V T5, Tp, TJ, TB, Ta, Tq, Tk, Tx, Tg, Ts;
72 T5 = VADD(T1, T4);
73 Tp = VFNMS(LDK(KP500000000), T4, T1);
74 TJ = VSUB(Tz, TA);
75 TB = VADD(Tz, TA);
76 Ta = VADD(T6, T9);
77 Tq = VFNMS(LDK(KP500000000), T9, T6);
78 Tk = VADD(Ti, Tj);
79 Tx = VSUB(Tj, Ti);
80 Tg = VADD(Tc, Tf);
81 Ts = VFNMS(LDK(KP500000000), Tf, Tc);
82 {
83 V Tr, TF, Tb, Tn, TG, Ty, Tl, Tt;
84 Tr = VADD(Tp, Tq);
85 TF = VSUB(Tp, Tq);
86 Tb = VSUB(T5, Ta);
87 Tn = VADD(T5, Ta);
88 TG = VADD(Tw, Tx);
89 Ty = VSUB(Tw, Tx);
90 Tl = VADD(Th, Tk);
91 Tt = VFNMS(LDK(KP500000000), Tk, Th);
92 {
93 V TC, TE, TH, TL, Tu, TI, Tm, To;
94 TC = VMUL(LDK(KP866025403), VSUB(Ty, TB));
95 TE = VMUL(LDK(KP866025403), VADD(TB, Ty));
96 TH = VFNMS(LDK(KP866025403), TG, TF);
97 TL = VFMA(LDK(KP866025403), TG, TF);
98 Tu = VADD(Ts, Tt);
99 TI = VSUB(Ts, Tt);
100 Tm = VSUB(Tg, Tl);
101 To = VADD(Tg, Tl);
102 {
103 V TK, TM, Tv, TD;
104 TK = VFMA(LDK(KP866025403), TJ, TI);
105 TM = VFNMS(LDK(KP866025403), TJ, TI);
106 Tv = VSUB(Tr, Tu);
107 TD = VADD(Tr, Tu);
108 {
109 V TN, TO, TP, TQ;
110 TN = VADD(Tn, To);
111 STM2(&(xo[0]), TN, ovs, &(xo[0]));
112 TO = VSUB(Tn, To);
113 STM2(&(xo[12]), TO, ovs, &(xo[0]));
114 TP = VFMAI(Tm, Tb);
115 STM2(&(xo[18]), TP, ovs, &(xo[2]));
116 TQ = VFNMSI(Tm, Tb);
117 STM2(&(xo[6]), TQ, ovs, &(xo[2]));
118 {
119 V TR, TS, TT, TU;
120 TR = VFMAI(TM, TL);
121 STM2(&(xo[10]), TR, ovs, &(xo[2]));
122 TS = VFNMSI(TM, TL);
123 STM2(&(xo[14]), TS, ovs, &(xo[2]));
124 STN2(&(xo[12]), TO, TS, ovs);
125 TT = VFNMSI(TK, TH);
126 STM2(&(xo[22]), TT, ovs, &(xo[2]));
127 TU = VFMAI(TK, TH);
128 STM2(&(xo[2]), TU, ovs, &(xo[2]));
129 STN2(&(xo[0]), TN, TU, ovs);
130 {
131 V TV, TW, TX, TY;
132 TV = VFNMSI(TE, TD);
133 STM2(&(xo[16]), TV, ovs, &(xo[0]));
134 STN2(&(xo[16]), TV, TP, ovs);
135 TW = VFMAI(TE, TD);
136 STM2(&(xo[8]), TW, ovs, &(xo[0]));
137 STN2(&(xo[8]), TW, TR, ovs);
138 TX = VFMAI(TC, Tv);
139 STM2(&(xo[4]), TX, ovs, &(xo[0]));
140 STN2(&(xo[4]), TX, TQ, ovs);
141 TY = VFNMSI(TC, Tv);
142 STM2(&(xo[20]), TY, ovs, &(xo[0]));
143 STN2(&(xo[20]), TY, TT, ovs);
144 }
145 }
146 }
147 }
148 }
149 }
150 }
151 }
152 }
153 VLEAVE();
154 }
155
156 static const kdft_desc desc = { 12, XSIMD_STRING("n2bv_12"), {30, 2, 18, 0}, &GENUS, 0, 2, 0, 0 };
157
158 void XSIMD(codelet_n2bv_12) (planner *p) {
159 X(kdft_register) (p, n2bv_12, &desc);
160 }
161
162 #else /* HAVE_FMA */
163
164 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 12 -name n2bv_12 -with-ostride 2 -include n2b.h -store-multiple 2 */
165
166 /*
167 * This function contains 48 FP additions, 8 FP multiplications,
168 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
169 * 33 stack variables, 2 constants, and 30 memory accesses
170 */
171 #include "n2b.h"
172
173 static void n2bv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
174 {
175 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
176 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
177 {
178 INT i;
179 const R *xi;
180 R *xo;
181 xi = ii;
182 xo = io;
183 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(24, is), MAKE_VOLATILE_STRIDE(24, os)) {
184 V T5, Ta, TG, TF, Ty, Tm, Ti, Tp, TJ, TI, Tx, Ts;
185 {
186 V T1, T6, T4, Tk, T9, Tl;
187 T1 = LD(&(xi[0]), ivs, &(xi[0]));
188 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
189 {
190 V T2, T3, T7, T8;
191 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
192 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
193 T4 = VADD(T2, T3);
194 Tk = VSUB(T2, T3);
195 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
196 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
197 T9 = VADD(T7, T8);
198 Tl = VSUB(T7, T8);
199 }
200 T5 = VFNMS(LDK(KP500000000), T4, T1);
201 Ta = VFNMS(LDK(KP500000000), T9, T6);
202 TG = VADD(T6, T9);
203 TF = VADD(T1, T4);
204 Ty = VADD(Tk, Tl);
205 Tm = VMUL(LDK(KP866025403), VSUB(Tk, Tl));
206 }
207 {
208 V Tn, Tq, Te, To, Th, Tr;
209 Tn = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
210 Tq = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
211 {
212 V Tc, Td, Tf, Tg;
213 Tc = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
214 Td = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
215 Te = VSUB(Tc, Td);
216 To = VADD(Tc, Td);
217 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
218 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
219 Th = VSUB(Tf, Tg);
220 Tr = VADD(Tf, Tg);
221 }
222 Ti = VMUL(LDK(KP866025403), VSUB(Te, Th));
223 Tp = VFNMS(LDK(KP500000000), To, Tn);
224 TJ = VADD(Tq, Tr);
225 TI = VADD(Tn, To);
226 Tx = VADD(Te, Th);
227 Ts = VFNMS(LDK(KP500000000), Tr, Tq);
228 }
229 {
230 V TN, TO, TP, TQ, TR, TS;
231 {
232 V TH, TK, TL, TM;
233 TH = VSUB(TF, TG);
234 TK = VBYI(VSUB(TI, TJ));
235 TN = VSUB(TH, TK);
236 STM2(&(xo[6]), TN, ovs, &(xo[2]));
237 TO = VADD(TH, TK);
238 STM2(&(xo[18]), TO, ovs, &(xo[2]));
239 TL = VADD(TF, TG);
240 TM = VADD(TI, TJ);
241 TP = VSUB(TL, TM);
242 STM2(&(xo[12]), TP, ovs, &(xo[0]));
243 TQ = VADD(TL, TM);
244 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
245 }
246 {
247 V Tj, Tv, Tu, Tw, Tb, Tt, TT, TU;
248 Tb = VSUB(T5, Ta);
249 Tj = VSUB(Tb, Ti);
250 Tv = VADD(Tb, Ti);
251 Tt = VSUB(Tp, Ts);
252 Tu = VBYI(VADD(Tm, Tt));
253 Tw = VBYI(VSUB(Tt, Tm));
254 TR = VSUB(Tj, Tu);
255 STM2(&(xo[22]), TR, ovs, &(xo[2]));
256 TS = VADD(Tv, Tw);
257 STM2(&(xo[10]), TS, ovs, &(xo[2]));
258 TT = VADD(Tj, Tu);
259 STM2(&(xo[2]), TT, ovs, &(xo[2]));
260 STN2(&(xo[0]), TQ, TT, ovs);
261 TU = VSUB(Tv, Tw);
262 STM2(&(xo[14]), TU, ovs, &(xo[2]));
263 STN2(&(xo[12]), TP, TU, ovs);
264 }
265 {
266 V Tz, TD, TC, TE, TA, TB;
267 Tz = VBYI(VMUL(LDK(KP866025403), VSUB(Tx, Ty)));
268 TD = VBYI(VMUL(LDK(KP866025403), VADD(Ty, Tx)));
269 TA = VADD(T5, Ta);
270 TB = VADD(Tp, Ts);
271 TC = VSUB(TA, TB);
272 TE = VADD(TA, TB);
273 {
274 V TV, TW, TX, TY;
275 TV = VADD(Tz, TC);
276 STM2(&(xo[4]), TV, ovs, &(xo[0]));
277 STN2(&(xo[4]), TV, TN, ovs);
278 TW = VSUB(TE, TD);
279 STM2(&(xo[16]), TW, ovs, &(xo[0]));
280 STN2(&(xo[16]), TW, TO, ovs);
281 TX = VSUB(TC, Tz);
282 STM2(&(xo[20]), TX, ovs, &(xo[0]));
283 STN2(&(xo[20]), TX, TR, ovs);
284 TY = VADD(TD, TE);
285 STM2(&(xo[8]), TY, ovs, &(xo[0]));
286 STN2(&(xo[8]), TY, TS, ovs);
287 }
288 }
289 }
290 }
291 }
292 VLEAVE();
293 }
294
295 static const kdft_desc desc = { 12, XSIMD_STRING("n2bv_12"), {44, 4, 4, 0}, &GENUS, 0, 2, 0, 0 };
296
297 void XSIMD(codelet_n2bv_12) (planner *p) {
298 X(kdft_register) (p, n2bv_12, &desc);
299 }
300
301 #endif /* HAVE_FMA */