Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.5/dft/simd/common/n1fv_9.c @ 42:2cd0e3b3e1fd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Current fftw source
author Chris Cannam
date Tue, 18 Oct 2016 13:40:26 +0100
parents
children
comparison
equal deleted inserted replaced
41:481f5f8c5634 42:2cd0e3b3e1fd
1 /*
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
3 * Copyright (c) 2003, 2007-14 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 Sat Jul 30 16:38:39 EDT 2016 */
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 9 -name n1fv_9 -include n1f.h */
29
30 /*
31 * This function contains 46 FP additions, 38 FP multiplications,
32 * (or, 12 additions, 4 multiplications, 34 fused multiply/add),
33 * 68 stack variables, 19 constants, and 18 memory accesses
34 */
35 #include "n1f.h"
36
37 static void n1fv_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP939692620, +0.939692620785908384054109277324731469936208134);
40 DVK(KP826351822, +0.826351822333069651148283373230685203999624323);
41 DVK(KP879385241, +0.879385241571816768108218554649462939872416269);
42 DVK(KP984807753, +0.984807753012208059366743024589523013670643252);
43 DVK(KP666666666, +0.666666666666666666666666666666666666666666667);
44 DVK(KP852868531, +0.852868531952443209628250963940074071936020296);
45 DVK(KP907603734, +0.907603734547952313649323976213898122064543220);
46 DVK(KP420276625, +0.420276625461206169731530603237061658838781920);
47 DVK(KP673648177, +0.673648177666930348851716626769314796000375677);
48 DVK(KP898197570, +0.898197570222573798468955502359086394667167570);
49 DVK(KP347296355, +0.347296355333860697703433253538629592000751354);
50 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
51 DVK(KP439692620, +0.439692620785908384054109277324731469936208134);
52 DVK(KP203604859, +0.203604859554852403062088995281827210665664861);
53 DVK(KP152703644, +0.152703644666139302296566746461370407999248646);
54 DVK(KP586256827, +0.586256827714544512072145703099641959914944179);
55 DVK(KP968908795, +0.968908795874236621082202410917456709164223497);
56 DVK(KP726681596, +0.726681596905677465811651808188092531873167623);
57 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
58 {
59 INT i;
60 const R *xi;
61 R *xo;
62 xi = ri;
63 xo = ro;
64 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(18, is), MAKE_VOLATILE_STRIDE(18, os)) {
65 V T1, T2, T3, T6, Tb, T7, T8, Tc, Td, Tv, T4;
66 T1 = LD(&(xi[0]), ivs, &(xi[0]));
67 T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
68 T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
69 T6 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
70 Tb = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
71 T7 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
72 T8 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
73 Tc = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
74 Td = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
75 Tv = VSUB(T3, T2);
76 T4 = VADD(T2, T3);
77 {
78 V Tl, T9, Tm, Te, Tj, T5;
79 Tl = VSUB(T7, T8);
80 T9 = VADD(T7, T8);
81 Tm = VSUB(Td, Tc);
82 Te = VADD(Tc, Td);
83 Tj = VFNMS(LDK(KP500000000), T4, T1);
84 T5 = VADD(T1, T4);
85 {
86 V Tn, Ta, Tk, Tf;
87 Tn = VFNMS(LDK(KP500000000), T9, T6);
88 Ta = VADD(T6, T9);
89 Tk = VFNMS(LDK(KP500000000), Te, Tb);
90 Tf = VADD(Tb, Te);
91 {
92 V Ty, TC, To, TB, Tx, Ts, Tg, Ti;
93 Ty = VFNMS(LDK(KP726681596), Tl, Tn);
94 TC = VFMA(LDK(KP968908795), Tn, Tl);
95 To = VFNMS(LDK(KP586256827), Tn, Tm);
96 TB = VFNMS(LDK(KP152703644), Tm, Tk);
97 Tx = VFMA(LDK(KP203604859), Tk, Tm);
98 Ts = VFNMS(LDK(KP439692620), Tl, Tk);
99 Tg = VADD(Ta, Tf);
100 Ti = VMUL(LDK(KP866025403), VSUB(Tf, Ta));
101 {
102 V Tz, TI, TF, TD, Tt, Th, Tq, Tp;
103 Tp = VFNMS(LDK(KP347296355), To, Tl);
104 Tz = VFMA(LDK(KP898197570), Ty, Tx);
105 TI = VFNMS(LDK(KP898197570), Ty, Tx);
106 TF = VFNMS(LDK(KP673648177), TC, TB);
107 TD = VFMA(LDK(KP673648177), TC, TB);
108 Tt = VFNMS(LDK(KP420276625), Ts, Tm);
109 ST(&(xo[0]), VADD(T5, Tg), ovs, &(xo[0]));
110 Th = VFNMS(LDK(KP500000000), Tg, T5);
111 Tq = VFNMS(LDK(KP907603734), Tp, Tk);
112 {
113 V TA, TJ, TE, TG, Tu, Tr, TK, TH, Tw;
114 TA = VFMA(LDK(KP852868531), Tz, Tj);
115 TJ = VFMA(LDK(KP666666666), TD, TI);
116 TE = VMUL(LDK(KP984807753), VFNMS(LDK(KP879385241), Tv, TD));
117 TG = VFNMS(LDK(KP500000000), Tz, TF);
118 Tu = VFNMS(LDK(KP826351822), Tt, Tn);
119 ST(&(xo[WS(os, 6)]), VFNMSI(Ti, Th), ovs, &(xo[0]));
120 ST(&(xo[WS(os, 3)]), VFMAI(Ti, Th), ovs, &(xo[WS(os, 1)]));
121 Tr = VFNMS(LDK(KP939692620), Tq, Tj);
122 TK = VMUL(LDK(KP866025403), VFMA(LDK(KP852868531), TJ, Tv));
123 ST(&(xo[WS(os, 8)]), VFMAI(TE, TA), ovs, &(xo[0]));
124 ST(&(xo[WS(os, 1)]), VFNMSI(TE, TA), ovs, &(xo[WS(os, 1)]));
125 TH = VFMA(LDK(KP852868531), TG, Tj);
126 Tw = VMUL(LDK(KP984807753), VFMA(LDK(KP879385241), Tv, Tu));
127 ST(&(xo[WS(os, 4)]), VFMAI(TK, TH), ovs, &(xo[0]));
128 ST(&(xo[WS(os, 5)]), VFNMSI(TK, TH), ovs, &(xo[WS(os, 1)]));
129 ST(&(xo[WS(os, 7)]), VFMAI(Tw, Tr), ovs, &(xo[WS(os, 1)]));
130 ST(&(xo[WS(os, 2)]), VFNMSI(Tw, Tr), ovs, &(xo[0]));
131 }
132 }
133 }
134 }
135 }
136 }
137 }
138 VLEAVE();
139 }
140
141 static const kdft_desc desc = { 9, XSIMD_STRING("n1fv_9"), {12, 4, 34, 0}, &GENUS, 0, 0, 0, 0 };
142
143 void XSIMD(codelet_n1fv_9) (planner *p) {
144 X(kdft_register) (p, n1fv_9, &desc);
145 }
146
147 #else /* HAVE_FMA */
148
149 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 9 -name n1fv_9 -include n1f.h */
150
151 /*
152 * This function contains 46 FP additions, 26 FP multiplications,
153 * (or, 30 additions, 10 multiplications, 16 fused multiply/add),
154 * 41 stack variables, 14 constants, and 18 memory accesses
155 */
156 #include "n1f.h"
157
158 static void n1fv_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
159 {
160 DVK(KP342020143, +0.342020143325668733044099614682259580763083368);
161 DVK(KP813797681, +0.813797681349373692844693217248393223289101568);
162 DVK(KP939692620, +0.939692620785908384054109277324731469936208134);
163 DVK(KP296198132, +0.296198132726023843175338011893050938967728390);
164 DVK(KP642787609, +0.642787609686539326322643409907263432907559884);
165 DVK(KP663413948, +0.663413948168938396205421319635891297216863310);
166 DVK(KP556670399, +0.556670399226419366452912952047023132968291906);
167 DVK(KP766044443, +0.766044443118978035202392650555416673935832457);
168 DVK(KP984807753, +0.984807753012208059366743024589523013670643252);
169 DVK(KP150383733, +0.150383733180435296639271897612501926072238258);
170 DVK(KP852868531, +0.852868531952443209628250963940074071936020296);
171 DVK(KP173648177, +0.173648177666930348851716626769314796000375677);
172 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
173 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
174 {
175 INT i;
176 const R *xi;
177 R *xo;
178 xi = ri;
179 xo = ro;
180 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(18, is), MAKE_VOLATILE_STRIDE(18, os)) {
181 V T5, Ts, Tj, To, Tf, Tn, Tp, Tu, Tl, Ta, Tk, Tm, Tt;
182 {
183 V T1, T2, T3, T4;
184 T1 = LD(&(xi[0]), ivs, &(xi[0]));
185 T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
186 T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
187 T4 = VADD(T2, T3);
188 T5 = VADD(T1, T4);
189 Ts = VMUL(LDK(KP866025403), VSUB(T3, T2));
190 Tj = VFNMS(LDK(KP500000000), T4, T1);
191 }
192 {
193 V Tb, Te, Tc, Td;
194 Tb = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
195 Tc = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
196 Td = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
197 Te = VADD(Tc, Td);
198 To = VSUB(Td, Tc);
199 Tf = VADD(Tb, Te);
200 Tn = VFNMS(LDK(KP500000000), Te, Tb);
201 Tp = VFMA(LDK(KP173648177), Tn, VMUL(LDK(KP852868531), To));
202 Tu = VFNMS(LDK(KP984807753), Tn, VMUL(LDK(KP150383733), To));
203 }
204 {
205 V T6, T9, T7, T8;
206 T6 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
207 T7 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
208 T8 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
209 T9 = VADD(T7, T8);
210 Tl = VSUB(T8, T7);
211 Ta = VADD(T6, T9);
212 Tk = VFNMS(LDK(KP500000000), T9, T6);
213 Tm = VFMA(LDK(KP766044443), Tk, VMUL(LDK(KP556670399), Tl));
214 Tt = VFNMS(LDK(KP642787609), Tk, VMUL(LDK(KP663413948), Tl));
215 }
216 {
217 V Ti, Tg, Th, Tz, TA;
218 Ti = VBYI(VMUL(LDK(KP866025403), VSUB(Tf, Ta)));
219 Tg = VADD(Ta, Tf);
220 Th = VFNMS(LDK(KP500000000), Tg, T5);
221 ST(&(xo[0]), VADD(T5, Tg), ovs, &(xo[0]));
222 ST(&(xo[WS(os, 3)]), VADD(Th, Ti), ovs, &(xo[WS(os, 1)]));
223 ST(&(xo[WS(os, 6)]), VSUB(Th, Ti), ovs, &(xo[0]));
224 Tz = VFMA(LDK(KP173648177), Tk, VFNMS(LDK(KP296198132), To, VFNMS(LDK(KP939692620), Tn, VFNMS(LDK(KP852868531), Tl, Tj))));
225 TA = VBYI(VSUB(VFNMS(LDK(KP342020143), Tn, VFNMS(LDK(KP150383733), Tl, VFNMS(LDK(KP984807753), Tk, VMUL(LDK(KP813797681), To)))), Ts));
226 ST(&(xo[WS(os, 7)]), VSUB(Tz, TA), ovs, &(xo[WS(os, 1)]));
227 ST(&(xo[WS(os, 2)]), VADD(Tz, TA), ovs, &(xo[0]));
228 {
229 V Tr, Tx, Tw, Ty, Tq, Tv;
230 Tq = VADD(Tm, Tp);
231 Tr = VADD(Tj, Tq);
232 Tx = VFMA(LDK(KP866025403), VSUB(Tt, Tu), VFNMS(LDK(KP500000000), Tq, Tj));
233 Tv = VADD(Tt, Tu);
234 Tw = VBYI(VADD(Ts, Tv));
235 Ty = VBYI(VADD(Ts, VFNMS(LDK(KP500000000), Tv, VMUL(LDK(KP866025403), VSUB(Tp, Tm)))));
236 ST(&(xo[WS(os, 8)]), VSUB(Tr, Tw), ovs, &(xo[0]));
237 ST(&(xo[WS(os, 4)]), VADD(Tx, Ty), ovs, &(xo[0]));
238 ST(&(xo[WS(os, 1)]), VADD(Tw, Tr), ovs, &(xo[WS(os, 1)]));
239 ST(&(xo[WS(os, 5)]), VSUB(Tx, Ty), ovs, &(xo[WS(os, 1)]));
240 }
241 }
242 }
243 }
244 VLEAVE();
245 }
246
247 static const kdft_desc desc = { 9, XSIMD_STRING("n1fv_9"), {30, 10, 16, 0}, &GENUS, 0, 0, 0, 0 };
248
249 void XSIMD(codelet_n1fv_9) (planner *p) {
250 X(kdft_register) (p, n1fv_9, &desc);
251 }
252
253 #endif /* HAVE_FMA */