Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n1fv_12.c @ 167:bd3cc4d1df30

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam <cannam@all-day-breakfast.com>
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
166:cbd6d7e562c7 167:bd3cc4d1df30
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 Thu May 24 08:04:51 EDT 2018 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n1fv_12 -include dft/simd/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 * 27 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "dft/simd/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 T5, Ta, TG, TF, TB, Tt, Ti, Tm, TJ, TI, TA, Tp;
49 {
50 V T1, T6, T4, Tr, T9, Ts;
51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
52 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
53 {
54 V T2, T3, T7, T8;
55 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
56 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
57 T4 = VADD(T2, T3);
58 Tr = VSUB(T3, T2);
59 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
60 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
61 T9 = VADD(T7, T8);
62 Ts = VSUB(T8, T7);
63 }
64 T5 = VFNMS(LDK(KP500000000), T4, T1);
65 Ta = VFNMS(LDK(KP500000000), T9, T6);
66 TG = VADD(T6, T9);
67 TF = VADD(T1, T4);
68 TB = VADD(Tr, Ts);
69 Tt = VSUB(Tr, Ts);
70 }
71 {
72 V Tk, Tn, Te, Tl, Th, To;
73 Tk = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
74 Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
75 {
76 V Tc, Td, Tf, Tg;
77 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
78 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
79 Te = VSUB(Tc, Td);
80 Tl = VADD(Td, Tc);
81 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
82 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
83 Th = VSUB(Tf, Tg);
84 To = VADD(Tf, Tg);
85 }
86 Ti = VADD(Te, Th);
87 Tm = VFNMS(LDK(KP500000000), Tl, Tk);
88 TJ = VADD(Tn, To);
89 TI = VADD(Tk, Tl);
90 TA = VSUB(Te, Th);
91 Tp = VFNMS(LDK(KP500000000), To, Tn);
92 }
93 {
94 V TH, TK, TL, TM;
95 TH = VSUB(TF, TG);
96 TK = VSUB(TI, TJ);
97 ST(&(xo[WS(os, 9)]), VFNMSI(TK, TH), ovs, &(xo[WS(os, 1)]));
98 ST(&(xo[WS(os, 3)]), VFMAI(TK, TH), ovs, &(xo[WS(os, 1)]));
99 TL = VADD(TF, TG);
100 TM = VADD(TI, TJ);
101 ST(&(xo[WS(os, 6)]), VSUB(TL, TM), ovs, &(xo[0]));
102 ST(&(xo[0]), VADD(TL, TM), ovs, &(xo[0]));
103 }
104 {
105 V Tj, Tv, Tu, Tw, Tb, Tq;
106 Tb = VSUB(T5, Ta);
107 Tj = VFMA(LDK(KP866025403), Ti, Tb);
108 Tv = VFNMS(LDK(KP866025403), Ti, Tb);
109 Tq = VSUB(Tm, Tp);
110 Tu = VFNMS(LDK(KP866025403), Tt, Tq);
111 Tw = VFMA(LDK(KP866025403), Tt, Tq);
112 ST(&(xo[WS(os, 1)]), VFNMSI(Tu, Tj), ovs, &(xo[WS(os, 1)]));
113 ST(&(xo[WS(os, 7)]), VFMAI(Tw, Tv), ovs, &(xo[WS(os, 1)]));
114 ST(&(xo[WS(os, 11)]), VFMAI(Tu, Tj), ovs, &(xo[WS(os, 1)]));
115 ST(&(xo[WS(os, 5)]), VFNMSI(Tw, Tv), ovs, &(xo[WS(os, 1)]));
116 }
117 {
118 V TC, TE, Tz, TD, Tx, Ty;
119 TC = VMUL(LDK(KP866025403), VSUB(TA, TB));
120 TE = VMUL(LDK(KP866025403), VADD(TB, TA));
121 Tx = VADD(T5, Ta);
122 Ty = VADD(Tm, Tp);
123 Tz = VSUB(Tx, Ty);
124 TD = VADD(Tx, Ty);
125 ST(&(xo[WS(os, 2)]), VFMAI(TC, Tz), ovs, &(xo[0]));
126 ST(&(xo[WS(os, 8)]), VFNMSI(TE, TD), ovs, &(xo[0]));
127 ST(&(xo[WS(os, 10)]), VFNMSI(TC, Tz), ovs, &(xo[0]));
128 ST(&(xo[WS(os, 4)]), VFMAI(TE, TD), ovs, &(xo[0]));
129 }
130 }
131 }
132 VLEAVE();
133 }
134
135 static const kdft_desc desc = { 12, XSIMD_STRING("n1fv_12"), {30, 2, 18, 0}, &GENUS, 0, 0, 0, 0 };
136
137 void XSIMD(codelet_n1fv_12) (planner *p) {
138 X(kdft_register) (p, n1fv_12, &desc);
139 }
140
141 #else
142
143 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 12 -name n1fv_12 -include dft/simd/n1f.h */
144
145 /*
146 * This function contains 48 FP additions, 8 FP multiplications,
147 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
148 * 27 stack variables, 2 constants, and 24 memory accesses
149 */
150 #include "dft/simd/n1f.h"
151
152 static void n1fv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
153 {
154 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
155 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
156 {
157 INT i;
158 const R *xi;
159 R *xo;
160 xi = ri;
161 xo = ro;
162 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)) {
163 V T5, Ta, TJ, Ty, Tq, Tp, Tg, Tl, TI, TA, Tz, Tu;
164 {
165 V T1, T6, T4, Tw, T9, Tx;
166 T1 = LD(&(xi[0]), ivs, &(xi[0]));
167 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
168 {
169 V T2, T3, T7, T8;
170 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
171 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
172 T4 = VADD(T2, T3);
173 Tw = VSUB(T3, T2);
174 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
175 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
176 T9 = VADD(T7, T8);
177 Tx = VSUB(T8, T7);
178 }
179 T5 = VADD(T1, T4);
180 Ta = VADD(T6, T9);
181 TJ = VADD(Tw, Tx);
182 Ty = VMUL(LDK(KP866025403), VSUB(Tw, Tx));
183 Tq = VFNMS(LDK(KP500000000), T9, T6);
184 Tp = VFNMS(LDK(KP500000000), T4, T1);
185 }
186 {
187 V Tc, Th, Tf, Ts, Tk, Tt;
188 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
189 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
190 {
191 V Td, Te, Ti, Tj;
192 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
193 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
194 Tf = VADD(Td, Te);
195 Ts = VSUB(Te, Td);
196 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
197 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
198 Tk = VADD(Ti, Tj);
199 Tt = VSUB(Tj, Ti);
200 }
201 Tg = VADD(Tc, Tf);
202 Tl = VADD(Th, Tk);
203 TI = VADD(Ts, Tt);
204 TA = VFNMS(LDK(KP500000000), Tk, Th);
205 Tz = VFNMS(LDK(KP500000000), Tf, Tc);
206 Tu = VMUL(LDK(KP866025403), VSUB(Ts, Tt));
207 }
208 {
209 V Tb, Tm, Tn, To;
210 Tb = VSUB(T5, Ta);
211 Tm = VBYI(VSUB(Tg, Tl));
212 ST(&(xo[WS(os, 9)]), VSUB(Tb, Tm), ovs, &(xo[WS(os, 1)]));
213 ST(&(xo[WS(os, 3)]), VADD(Tb, Tm), ovs, &(xo[WS(os, 1)]));
214 Tn = VADD(T5, Ta);
215 To = VADD(Tg, Tl);
216 ST(&(xo[WS(os, 6)]), VSUB(Tn, To), ovs, &(xo[0]));
217 ST(&(xo[0]), VADD(Tn, To), ovs, &(xo[0]));
218 }
219 {
220 V Tv, TE, TC, TD, Tr, TB;
221 Tr = VSUB(Tp, Tq);
222 Tv = VSUB(Tr, Tu);
223 TE = VADD(Tr, Tu);
224 TB = VSUB(Tz, TA);
225 TC = VBYI(VADD(Ty, TB));
226 TD = VBYI(VSUB(Ty, TB));
227 ST(&(xo[WS(os, 5)]), VSUB(Tv, TC), ovs, &(xo[WS(os, 1)]));
228 ST(&(xo[WS(os, 11)]), VSUB(TE, TD), ovs, &(xo[WS(os, 1)]));
229 ST(&(xo[WS(os, 7)]), VADD(TC, Tv), ovs, &(xo[WS(os, 1)]));
230 ST(&(xo[WS(os, 1)]), VADD(TD, TE), ovs, &(xo[WS(os, 1)]));
231 }
232 {
233 V TK, TM, TH, TL, TF, TG;
234 TK = VBYI(VMUL(LDK(KP866025403), VSUB(TI, TJ)));
235 TM = VBYI(VMUL(LDK(KP866025403), VADD(TJ, TI)));
236 TF = VADD(Tp, Tq);
237 TG = VADD(Tz, TA);
238 TH = VSUB(TF, TG);
239 TL = VADD(TF, TG);
240 ST(&(xo[WS(os, 10)]), VSUB(TH, TK), ovs, &(xo[0]));
241 ST(&(xo[WS(os, 4)]), VADD(TL, TM), ovs, &(xo[0]));
242 ST(&(xo[WS(os, 2)]), VADD(TH, TK), ovs, &(xo[0]));
243 ST(&(xo[WS(os, 8)]), VSUB(TL, TM), ovs, &(xo[0]));
244 }
245 }
246 }
247 VLEAVE();
248 }
249
250 static const kdft_desc desc = { 12, XSIMD_STRING("n1fv_12"), {44, 4, 4, 0}, &GENUS, 0, 0, 0, 0 };
251
252 void XSIMD(codelet_n1fv_12) (planner *p) {
253 X(kdft_register) (p, n1fv_12, &desc);
254 }
255
256 #endif