Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2fv_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 07:37:22 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 -n 12 -name n2fv_12 -with-ostride 2 -include n2f.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 "n2f.h"
36
37 static void n2fv_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 = ri;
46 xo = ro;
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, Tk, Tn, Tc, Td, Tf, Tr, T4, Ts, T9, Tg, Te, Tl;
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 Tk = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
58 Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
59 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
60 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
61 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
62 Tr = VSUB(T3, T2);
63 T4 = VADD(T2, T3);
64 Ts = VSUB(T8, T7);
65 T9 = VADD(T7, T8);
66 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
67 }
68 Te = VSUB(Tc, Td);
69 Tl = VADD(Td, Tc);
70 {
71 V T5, TF, TB, Tt, Ta, TG, Th, To, Tm, TI;
72 T5 = VFNMS(LDK(KP500000000), T4, T1);
73 TF = VADD(T1, T4);
74 TB = VADD(Tr, Ts);
75 Tt = VSUB(Tr, Ts);
76 Ta = VFNMS(LDK(KP500000000), T9, T6);
77 TG = VADD(T6, T9);
78 Th = VSUB(Tf, Tg);
79 To = VADD(Tf, Tg);
80 Tm = VFNMS(LDK(KP500000000), Tl, Tk);
81 TI = VADD(Tk, Tl);
82 {
83 V TH, TL, Tb, Tx, TJ, Tp, Ti, TA;
84 TH = VSUB(TF, TG);
85 TL = VADD(TF, TG);
86 Tb = VSUB(T5, Ta);
87 Tx = VADD(T5, Ta);
88 TJ = VADD(Tn, To);
89 Tp = VFNMS(LDK(KP500000000), To, Tn);
90 Ti = VADD(Te, Th);
91 TA = VSUB(Te, Th);
92 {
93 V Tq, Ty, TK, TM;
94 Tq = VSUB(Tm, Tp);
95 Ty = VADD(Tm, Tp);
96 TK = VSUB(TI, TJ);
97 TM = VADD(TI, TJ);
98 {
99 V TC, TE, Tj, Tv;
100 TC = VMUL(LDK(KP866025403), VSUB(TA, TB));
101 TE = VMUL(LDK(KP866025403), VADD(TB, TA));
102 Tj = VFMA(LDK(KP866025403), Ti, Tb);
103 Tv = VFNMS(LDK(KP866025403), Ti, Tb);
104 {
105 V Tz, TD, Tu, Tw;
106 Tz = VSUB(Tx, Ty);
107 TD = VADD(Tx, Ty);
108 Tu = VFNMS(LDK(KP866025403), Tt, Tq);
109 Tw = VFMA(LDK(KP866025403), Tt, Tq);
110 {
111 V TN, TO, TP, TQ;
112 TN = VADD(TL, TM);
113 STM2(&(xo[0]), TN, ovs, &(xo[0]));
114 TO = VSUB(TL, TM);
115 STM2(&(xo[12]), TO, ovs, &(xo[0]));
116 TP = VFMAI(TK, TH);
117 STM2(&(xo[6]), TP, ovs, &(xo[2]));
118 TQ = VFNMSI(TK, TH);
119 STM2(&(xo[18]), TQ, ovs, &(xo[2]));
120 {
121 V TR, TS, TT, TU;
122 TR = VFMAI(TE, TD);
123 STM2(&(xo[8]), TR, ovs, &(xo[0]));
124 TS = VFNMSI(TE, TD);
125 STM2(&(xo[16]), TS, ovs, &(xo[0]));
126 STN2(&(xo[16]), TS, TQ, ovs);
127 TT = VFNMSI(TC, Tz);
128 STM2(&(xo[20]), TT, ovs, &(xo[0]));
129 TU = VFMAI(TC, Tz);
130 STM2(&(xo[4]), TU, ovs, &(xo[0]));
131 STN2(&(xo[4]), TU, TP, ovs);
132 {
133 V TV, TW, TX, TY;
134 TV = VFNMSI(Tw, Tv);
135 STM2(&(xo[10]), TV, ovs, &(xo[2]));
136 STN2(&(xo[8]), TR, TV, ovs);
137 TW = VFMAI(Tw, Tv);
138 STM2(&(xo[14]), TW, ovs, &(xo[2]));
139 STN2(&(xo[12]), TO, TW, ovs);
140 TX = VFMAI(Tu, Tj);
141 STM2(&(xo[22]), TX, ovs, &(xo[2]));
142 STN2(&(xo[20]), TT, TX, ovs);
143 TY = VFNMSI(Tu, Tj);
144 STM2(&(xo[2]), TY, ovs, &(xo[2]));
145 STN2(&(xo[0]), TN, TY, ovs);
146 }
147 }
148 }
149 }
150 }
151 }
152 }
153 }
154 }
155 }
156 VLEAVE();
157 }
158
159 static const kdft_desc desc = { 12, XSIMD_STRING("n2fv_12"), {30, 2, 18, 0}, &GENUS, 0, 2, 0, 0 };
160
161 void XSIMD(codelet_n2fv_12) (planner *p) {
162 X(kdft_register) (p, n2fv_12, &desc);
163 }
164
165 #else /* HAVE_FMA */
166
167 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n2fv_12 -with-ostride 2 -include n2f.h -store-multiple 2 */
168
169 /*
170 * This function contains 48 FP additions, 8 FP multiplications,
171 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
172 * 33 stack variables, 2 constants, and 30 memory accesses
173 */
174 #include "n2f.h"
175
176 static void n2fv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
177 {
178 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
179 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
180 {
181 INT i;
182 const R *xi;
183 R *xo;
184 xi = ri;
185 xo = ro;
186 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)) {
187 V T5, Ta, TJ, Ty, Tq, Tp, Tg, Tl, TI, TA, Tz, Tu;
188 {
189 V T1, T6, T4, Tw, T9, Tx;
190 T1 = LD(&(xi[0]), ivs, &(xi[0]));
191 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
192 {
193 V T2, T3, T7, T8;
194 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
195 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
196 T4 = VADD(T2, T3);
197 Tw = VSUB(T3, T2);
198 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
199 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
200 T9 = VADD(T7, T8);
201 Tx = VSUB(T8, T7);
202 }
203 T5 = VADD(T1, T4);
204 Ta = VADD(T6, T9);
205 TJ = VADD(Tw, Tx);
206 Ty = VMUL(LDK(KP866025403), VSUB(Tw, Tx));
207 Tq = VFNMS(LDK(KP500000000), T9, T6);
208 Tp = VFNMS(LDK(KP500000000), T4, T1);
209 }
210 {
211 V Tc, Th, Tf, Ts, Tk, Tt;
212 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
213 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
214 {
215 V Td, Te, Ti, Tj;
216 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
217 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
218 Tf = VADD(Td, Te);
219 Ts = VSUB(Te, Td);
220 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
221 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
222 Tk = VADD(Ti, Tj);
223 Tt = VSUB(Tj, Ti);
224 }
225 Tg = VADD(Tc, Tf);
226 Tl = VADD(Th, Tk);
227 TI = VADD(Ts, Tt);
228 TA = VFNMS(LDK(KP500000000), Tk, Th);
229 Tz = VFNMS(LDK(KP500000000), Tf, Tc);
230 Tu = VMUL(LDK(KP866025403), VSUB(Ts, Tt));
231 }
232 {
233 V TN, TO, TP, TQ, TR, TS;
234 {
235 V Tb, Tm, Tn, To;
236 Tb = VSUB(T5, Ta);
237 Tm = VBYI(VSUB(Tg, Tl));
238 TN = VSUB(Tb, Tm);
239 STM2(&(xo[18]), TN, ovs, &(xo[2]));
240 TO = VADD(Tb, Tm);
241 STM2(&(xo[6]), TO, ovs, &(xo[2]));
242 Tn = VADD(T5, Ta);
243 To = VADD(Tg, Tl);
244 TP = VSUB(Tn, To);
245 STM2(&(xo[12]), TP, ovs, &(xo[0]));
246 TQ = VADD(Tn, To);
247 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
248 }
249 {
250 V Tv, TE, TC, TD, Tr, TB, TT, TU;
251 Tr = VSUB(Tp, Tq);
252 Tv = VSUB(Tr, Tu);
253 TE = VADD(Tr, Tu);
254 TB = VSUB(Tz, TA);
255 TC = VBYI(VADD(Ty, TB));
256 TD = VBYI(VSUB(Ty, TB));
257 TR = VSUB(Tv, TC);
258 STM2(&(xo[10]), TR, ovs, &(xo[2]));
259 TS = VSUB(TE, TD);
260 STM2(&(xo[22]), TS, ovs, &(xo[2]));
261 TT = VADD(TC, Tv);
262 STM2(&(xo[14]), TT, ovs, &(xo[2]));
263 STN2(&(xo[12]), TP, TT, ovs);
264 TU = VADD(TD, TE);
265 STM2(&(xo[2]), TU, ovs, &(xo[2]));
266 STN2(&(xo[0]), TQ, TU, ovs);
267 }
268 {
269 V TK, TM, TH, TL, TF, TG;
270 TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
271 TM = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
272 TF = VADD(Tp, Tq);
273 TG = VADD(Tz, TA);
274 TH = VSUB(TF, TG);
275 TL = VADD(TF, TG);
276 {
277 V TV, TW, TX, TY;
278 TV = VSUB(TH, TK);
279 STM2(&(xo[20]), TV, ovs, &(xo[0]));
280 STN2(&(xo[20]), TV, TS, ovs);
281 TW = VADD(TL, TM);
282 STM2(&(xo[8]), TW, ovs, &(xo[0]));
283 STN2(&(xo[8]), TW, TR, ovs);
284 TX = VADD(TH, TK);
285 STM2(&(xo[4]), TX, ovs, &(xo[0]));
286 STN2(&(xo[4]), TX, TO, ovs);
287 TY = VSUB(TL, TM);
288 STM2(&(xo[16]), TY, ovs, &(xo[0]));
289 STN2(&(xo[16]), TY, TN, ovs);
290 }
291 }
292 }
293 }
294 }
295 VLEAVE();
296 }
297
298 static const kdft_desc desc = { 12, XSIMD_STRING("n2fv_12"), {44, 4, 4, 0}, &GENUS, 0, 2, 0, 0 };
299
300 void XSIMD(codelet_n2fv_12) (planner *p) {
301 X(kdft_register) (p, n2fv_12, &desc);
302 }
303
304 #endif /* HAVE_FMA */