Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2bv_10.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 07:37:29 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 10 -name n2bv_10 -with-ostride 2 -include n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 42 FP additions, 22 FP multiplications,
32 * (or, 24 additions, 4 multiplications, 18 fused multiply/add),
33 * 53 stack variables, 4 constants, and 25 memory accesses
34 */
35 #include "n2b.h"
36
37 static void n2bv_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
43 {
44 INT i;
45 const R *xi;
46 R *xo;
47 xi = ii;
48 xo = io;
49 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
50 V Tb, Tr, T3, Ts, T6, Tw, Tg, Tt, T9, Tc, T1, T2;
51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
52 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
53 {
54 V T4, T5, Te, Tf, T7, T8;
55 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
56 T5 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
57 Te = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
58 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
59 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
60 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
61 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
62 Tr = VADD(T1, T2);
63 T3 = VSUB(T1, T2);
64 Ts = VADD(T4, T5);
65 T6 = VSUB(T4, T5);
66 Tw = VADD(Te, Tf);
67 Tg = VSUB(Te, Tf);
68 Tt = VADD(T7, T8);
69 T9 = VSUB(T7, T8);
70 Tc = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
71 }
72 {
73 V TD, Tu, Tm, Ta, Td, Tv;
74 TD = VSUB(Ts, Tt);
75 Tu = VADD(Ts, Tt);
76 Tm = VSUB(T6, T9);
77 Ta = VADD(T6, T9);
78 Td = VSUB(Tb, Tc);
79 Tv = VADD(Tb, Tc);
80 {
81 V TC, Tx, Tn, Th;
82 TC = VSUB(Tv, Tw);
83 Tx = VADD(Tv, Tw);
84 Tn = VSUB(Td, Tg);
85 Th = VADD(Td, Tg);
86 {
87 V Ty, TA, TE, TG, Ti, Tk, To, Tq;
88 Ty = VADD(Tu, Tx);
89 TA = VSUB(Tu, Tx);
90 TE = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TD, TC));
91 TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TC, TD));
92 Ti = VADD(Ta, Th);
93 Tk = VSUB(Ta, Th);
94 To = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tn, Tm));
95 Tq = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tm, Tn));
96 {
97 V Tz, TH, Tj, TI;
98 Tz = VFNMS(LDK(KP250000000), Ty, Tr);
99 TH = VADD(Tr, Ty);
100 STM2(&(xo[0]), TH, ovs, &(xo[0]));
101 Tj = VFNMS(LDK(KP250000000), Ti, T3);
102 TI = VADD(T3, Ti);
103 STM2(&(xo[10]), TI, ovs, &(xo[2]));
104 {
105 V TB, TF, Tl, Tp;
106 TB = VFNMS(LDK(KP559016994), TA, Tz);
107 TF = VFMA(LDK(KP559016994), TA, Tz);
108 Tl = VFMA(LDK(KP559016994), Tk, Tj);
109 Tp = VFNMS(LDK(KP559016994), Tk, Tj);
110 {
111 V TJ, TK, TL, TM;
112 TJ = VFNMSI(TG, TF);
113 STM2(&(xo[8]), TJ, ovs, &(xo[0]));
114 STN2(&(xo[8]), TJ, TI, ovs);
115 TK = VFMAI(TG, TF);
116 STM2(&(xo[12]), TK, ovs, &(xo[0]));
117 TL = VFMAI(TE, TB);
118 STM2(&(xo[16]), TL, ovs, &(xo[0]));
119 TM = VFNMSI(TE, TB);
120 STM2(&(xo[4]), TM, ovs, &(xo[0]));
121 {
122 V TN, TO, TP, TQ;
123 TN = VFMAI(Tq, Tp);
124 STM2(&(xo[6]), TN, ovs, &(xo[2]));
125 STN2(&(xo[4]), TM, TN, ovs);
126 TO = VFNMSI(Tq, Tp);
127 STM2(&(xo[14]), TO, ovs, &(xo[2]));
128 STN2(&(xo[12]), TK, TO, ovs);
129 TP = VFNMSI(To, Tl);
130 STM2(&(xo[18]), TP, ovs, &(xo[2]));
131 STN2(&(xo[16]), TL, TP, ovs);
132 TQ = VFMAI(To, Tl);
133 STM2(&(xo[2]), TQ, ovs, &(xo[2]));
134 STN2(&(xo[0]), TH, TQ, ovs);
135 }
136 }
137 }
138 }
139 }
140 }
141 }
142 }
143 }
144 VLEAVE();
145 }
146
147 static const kdft_desc desc = { 10, XSIMD_STRING("n2bv_10"), {24, 4, 18, 0}, &GENUS, 0, 2, 0, 0 };
148
149 void XSIMD(codelet_n2bv_10) (planner *p) {
150 X(kdft_register) (p, n2bv_10, &desc);
151 }
152
153 #else /* HAVE_FMA */
154
155 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 10 -name n2bv_10 -with-ostride 2 -include n2b.h -store-multiple 2 */
156
157 /*
158 * This function contains 42 FP additions, 12 FP multiplications,
159 * (or, 36 additions, 6 multiplications, 6 fused multiply/add),
160 * 36 stack variables, 4 constants, and 25 memory accesses
161 */
162 #include "n2b.h"
163
164 static void n2bv_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
165 {
166 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
167 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
168 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
169 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
170 {
171 INT i;
172 const R *xi;
173 R *xo;
174 xi = ii;
175 xo = io;
176 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
177 V Tl, Ty, T7, Te, Tw, Tt, Tz, TA, TB, Tg, Th, Tm, Tj, Tk;
178 Tj = LD(&(xi[0]), ivs, &(xi[0]));
179 Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
180 Tl = VSUB(Tj, Tk);
181 Ty = VADD(Tj, Tk);
182 {
183 V T3, Tr, Td, Tv, T6, Ts, Ta, Tu;
184 {
185 V T1, T2, Tb, Tc;
186 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
187 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
188 T3 = VSUB(T1, T2);
189 Tr = VADD(T1, T2);
190 Tb = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
191 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
192 Td = VSUB(Tb, Tc);
193 Tv = VADD(Tb, Tc);
194 }
195 {
196 V T4, T5, T8, T9;
197 T4 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
198 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
199 T6 = VSUB(T4, T5);
200 Ts = VADD(T4, T5);
201 T8 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
202 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
203 Ta = VSUB(T8, T9);
204 Tu = VADD(T8, T9);
205 }
206 T7 = VSUB(T3, T6);
207 Te = VSUB(Ta, Td);
208 Tw = VSUB(Tu, Tv);
209 Tt = VSUB(Tr, Ts);
210 Tz = VADD(Tr, Ts);
211 TA = VADD(Tu, Tv);
212 TB = VADD(Tz, TA);
213 Tg = VADD(T3, T6);
214 Th = VADD(Ta, Td);
215 Tm = VADD(Tg, Th);
216 }
217 {
218 V TH, TI, TK, TL, TM;
219 TH = VADD(Tl, Tm);
220 STM2(&(xo[10]), TH, ovs, &(xo[2]));
221 TI = VADD(Ty, TB);
222 STM2(&(xo[0]), TI, ovs, &(xo[0]));
223 {
224 V Tf, Tq, To, Tp, Ti, Tn, TJ;
225 Tf = VBYI(VFMA(LDK(KP951056516), T7, VMUL(LDK(KP587785252), Te)));
226 Tq = VBYI(VFNMS(LDK(KP951056516), Te, VMUL(LDK(KP587785252), T7)));
227 Ti = VMUL(LDK(KP559016994), VSUB(Tg, Th));
228 Tn = VFNMS(LDK(KP250000000), Tm, Tl);
229 To = VADD(Ti, Tn);
230 Tp = VSUB(Tn, Ti);
231 TJ = VADD(Tf, To);
232 STM2(&(xo[2]), TJ, ovs, &(xo[2]));
233 STN2(&(xo[0]), TI, TJ, ovs);
234 TK = VADD(Tq, Tp);
235 STM2(&(xo[14]), TK, ovs, &(xo[2]));
236 TL = VSUB(To, Tf);
237 STM2(&(xo[18]), TL, ovs, &(xo[2]));
238 TM = VSUB(Tp, Tq);
239 STM2(&(xo[6]), TM, ovs, &(xo[2]));
240 }
241 {
242 V Tx, TG, TE, TF, TC, TD;
243 Tx = VBYI(VFNMS(LDK(KP951056516), Tw, VMUL(LDK(KP587785252), Tt)));
244 TG = VBYI(VFMA(LDK(KP951056516), Tt, VMUL(LDK(KP587785252), Tw)));
245 TC = VFNMS(LDK(KP250000000), TB, Ty);
246 TD = VMUL(LDK(KP559016994), VSUB(Tz, TA));
247 TE = VSUB(TC, TD);
248 TF = VADD(TD, TC);
249 {
250 V TN, TO, TP, TQ;
251 TN = VADD(Tx, TE);
252 STM2(&(xo[4]), TN, ovs, &(xo[0]));
253 STN2(&(xo[4]), TN, TM, ovs);
254 TO = VADD(TG, TF);
255 STM2(&(xo[12]), TO, ovs, &(xo[0]));
256 STN2(&(xo[12]), TO, TK, ovs);
257 TP = VSUB(TE, Tx);
258 STM2(&(xo[16]), TP, ovs, &(xo[0]));
259 STN2(&(xo[16]), TP, TL, ovs);
260 TQ = VSUB(TF, TG);
261 STM2(&(xo[8]), TQ, ovs, &(xo[0]));
262 STN2(&(xo[8]), TQ, TH, ovs);
263 }
264 }
265 }
266 }
267 }
268 VLEAVE();
269 }
270
271 static const kdft_desc desc = { 10, XSIMD_STRING("n2bv_10"), {36, 6, 6, 0}, &GENUS, 0, 2, 0, 0 };
272
273 void XSIMD(codelet_n2bv_10) (planner *p) {
274 X(kdft_register) (p, n2bv_10, &desc);
275 }
276
277 #endif /* HAVE_FMA */