Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n1fv_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 07:36:52 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 n1fv_12 -include n1f.h */
29
30 /*
31 * This function contains 48 FP additions, 20 FP multiplications,
32 * (or, 30 additions, 2 multiplications, 18 fused multiply/add),
33 * 49 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "n1f.h"
36
37 static void n1fv_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 ST(&(xo[0]), VADD(TL, TM), ovs, &(xo[0]));
111 ST(&(xo[WS(os, 6)]), VSUB(TL, TM), ovs, &(xo[0]));
112 ST(&(xo[WS(os, 3)]), VFMAI(TK, TH), ovs, &(xo[WS(os, 1)]));
113 ST(&(xo[WS(os, 9)]), VFNMSI(TK, TH), ovs, &(xo[WS(os, 1)]));
114 ST(&(xo[WS(os, 4)]), VFMAI(TE, TD), ovs, &(xo[0]));
115 ST(&(xo[WS(os, 8)]), VFNMSI(TE, TD), ovs, &(xo[0]));
116 ST(&(xo[WS(os, 10)]), VFNMSI(TC, Tz), ovs, &(xo[0]));
117 ST(&(xo[WS(os, 2)]), VFMAI(TC, Tz), ovs, &(xo[0]));
118 ST(&(xo[WS(os, 5)]), VFNMSI(Tw, Tv), ovs, &(xo[WS(os, 1)]));
119 ST(&(xo[WS(os, 7)]), VFMAI(Tw, Tv), ovs, &(xo[WS(os, 1)]));
120 ST(&(xo[WS(os, 11)]), VFMAI(Tu, Tj), ovs, &(xo[WS(os, 1)]));
121 ST(&(xo[WS(os, 1)]), VFNMSI(Tu, Tj), ovs, &(xo[WS(os, 1)]));
122 }
123 }
124 }
125 }
126 }
127 }
128 }
129 VLEAVE();
130 }
131
132 static const kdft_desc desc = { 12, XSIMD_STRING("n1fv_12"), {30, 2, 18, 0}, &GENUS, 0, 0, 0, 0 };
133
134 void XSIMD(codelet_n1fv_12) (planner *p) {
135 X(kdft_register) (p, n1fv_12, &desc);
136 }
137
138 #else /* HAVE_FMA */
139
140 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n1fv_12 -include n1f.h */
141
142 /*
143 * This function contains 48 FP additions, 8 FP multiplications,
144 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
145 * 27 stack variables, 2 constants, and 24 memory accesses
146 */
147 #include "n1f.h"
148
149 static void n1fv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
150 {
151 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
152 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
153 {
154 INT i;
155 const R *xi;
156 R *xo;
157 xi = ri;
158 xo = ro;
159 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)) {
160 V T5, Ta, TJ, Ty, Tq, Tp, Tg, Tl, TI, TA, Tz, Tu;
161 {
162 V T1, T6, T4, Tw, T9, Tx;
163 T1 = LD(&(xi[0]), ivs, &(xi[0]));
164 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
165 {
166 V T2, T3, T7, T8;
167 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
168 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
169 T4 = VADD(T2, T3);
170 Tw = VSUB(T3, T2);
171 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
172 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
173 T9 = VADD(T7, T8);
174 Tx = VSUB(T8, T7);
175 }
176 T5 = VADD(T1, T4);
177 Ta = VADD(T6, T9);
178 TJ = VADD(Tw, Tx);
179 Ty = VMUL(LDK(KP866025403), VSUB(Tw, Tx));
180 Tq = VFNMS(LDK(KP500000000), T9, T6);
181 Tp = VFNMS(LDK(KP500000000), T4, T1);
182 }
183 {
184 V Tc, Th, Tf, Ts, Tk, Tt;
185 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
186 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
187 {
188 V Td, Te, Ti, Tj;
189 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
190 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
191 Tf = VADD(Td, Te);
192 Ts = VSUB(Te, Td);
193 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
194 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
195 Tk = VADD(Ti, Tj);
196 Tt = VSUB(Tj, Ti);
197 }
198 Tg = VADD(Tc, Tf);
199 Tl = VADD(Th, Tk);
200 TI = VADD(Ts, Tt);
201 TA = VFNMS(LDK(KP500000000), Tk, Th);
202 Tz = VFNMS(LDK(KP500000000), Tf, Tc);
203 Tu = VMUL(LDK(KP866025403), VSUB(Ts, Tt));
204 }
205 {
206 V Tb, Tm, Tn, To;
207 Tb = VSUB(T5, Ta);
208 Tm = VBYI(VSUB(Tg, Tl));
209 ST(&(xo[WS(os, 9)]), VSUB(Tb, Tm), ovs, &(xo[WS(os, 1)]));
210 ST(&(xo[WS(os, 3)]), VADD(Tb, Tm), ovs, &(xo[WS(os, 1)]));
211 Tn = VADD(T5, Ta);
212 To = VADD(Tg, Tl);
213 ST(&(xo[WS(os, 6)]), VSUB(Tn, To), ovs, &(xo[0]));
214 ST(&(xo[0]), VADD(Tn, To), ovs, &(xo[0]));
215 }
216 {
217 V Tv, TE, TC, TD, Tr, TB;
218 Tr = VSUB(Tp, Tq);
219 Tv = VSUB(Tr, Tu);
220 TE = VADD(Tr, Tu);
221 TB = VSUB(Tz, TA);
222 TC = VBYI(VADD(Ty, TB));
223 TD = VBYI(VSUB(Ty, TB));
224 ST(&(xo[WS(os, 5)]), VSUB(Tv, TC), ovs, &(xo[WS(os, 1)]));
225 ST(&(xo[WS(os, 11)]), VSUB(TE, TD), ovs, &(xo[WS(os, 1)]));
226 ST(&(xo[WS(os, 7)]), VADD(TC, Tv), ovs, &(xo[WS(os, 1)]));
227 ST(&(xo[WS(os, 1)]), VADD(TD, TE), ovs, &(xo[WS(os, 1)]));
228 }
229 {
230 V TK, TM, TH, TL, TF, TG;
231 TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
232 TM = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
233 TF = VADD(Tp, Tq);
234 TG = VADD(Tz, TA);
235 TH = VSUB(TF, TG);
236 TL = VADD(TF, TG);
237 ST(&(xo[WS(os, 10)]), VSUB(TH, TK), ovs, &(xo[0]));
238 ST(&(xo[WS(os, 4)]), VADD(TL, TM), ovs, &(xo[0]));
239 ST(&(xo[WS(os, 2)]), VADD(TH, TK), ovs, &(xo[0]));
240 ST(&(xo[WS(os, 8)]), VSUB(TL, TM), ovs, &(xo[0]));
241 }
242 }
243 }
244 VLEAVE();
245 }
246
247 static const kdft_desc desc = { 12, XSIMD_STRING("n1fv_12"), {44, 4, 4, 0}, &GENUS, 0, 0, 0, 0 };
248
249 void XSIMD(codelet_n1fv_12) (planner *p) {
250 X(kdft_register) (p, n1fv_12, &desc);
251 }
252
253 #endif /* HAVE_FMA */