Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n2fv_10.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 08:05:07 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 10 -name n2fv_10 -with-ostride 2 -include dft/simd/n2f.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 * 36 stack variables, 4 constants, and 25 memory accesses
34 */
35 #include "dft/simd/n2f.h"
36
37 static void n2fv_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 = ri;
48 xo = ro;
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 T3, Tr, Tm, Tn, TD, TC, Tu, Tx, Ty, Ta, Th, Ti, T1, T2;
51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
52 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
53 T3 = VSUB(T1, T2);
54 Tr = VADD(T1, T2);
55 {
56 V T6, Ts, Tg, Tw, T9, Tt, Td, Tv;
57 {
58 V T4, T5, Te, Tf;
59 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
60 T5 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
61 T6 = VSUB(T4, T5);
62 Ts = VADD(T4, T5);
63 Te = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
64 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
65 Tg = VSUB(Te, Tf);
66 Tw = VADD(Te, Tf);
67 }
68 {
69 V T7, T8, Tb, Tc;
70 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
71 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
72 T9 = VSUB(T7, T8);
73 Tt = VADD(T7, T8);
74 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
75 Tc = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
76 Td = VSUB(Tb, Tc);
77 Tv = VADD(Tb, Tc);
78 }
79 Tm = VSUB(T6, T9);
80 Tn = VSUB(Td, Tg);
81 TD = VSUB(Ts, Tt);
82 TC = VSUB(Tv, Tw);
83 Tu = VADD(Ts, Tt);
84 Tx = VADD(Tv, Tw);
85 Ty = VADD(Tu, Tx);
86 Ta = VADD(T6, T9);
87 Th = VADD(Td, Tg);
88 Ti = VADD(Ta, Th);
89 }
90 {
91 V TH, TI, TK, TL, TM;
92 TH = VADD(T3, Ti);
93 STM2(&(xo[10]), TH, ovs, &(xo[2]));
94 TI = VADD(Tr, Ty);
95 STM2(&(xo[0]), TI, ovs, &(xo[0]));
96 {
97 V To, Tq, Tl, Tp, Tj, Tk, TJ;
98 To = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tn, Tm));
99 Tq = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tm, Tn));
100 Tj = VFNMS(LDK(KP250000000), Ti, T3);
101 Tk = VSUB(Ta, Th);
102 Tl = VFMA(LDK(KP559016994), Tk, Tj);
103 Tp = VFNMS(LDK(KP559016994), Tk, Tj);
104 TJ = VFNMSI(To, Tl);
105 STM2(&(xo[2]), TJ, ovs, &(xo[2]));
106 STN2(&(xo[0]), TI, TJ, ovs);
107 TK = VFMAI(Tq, Tp);
108 STM2(&(xo[14]), TK, ovs, &(xo[2]));
109 TL = VFMAI(To, Tl);
110 STM2(&(xo[18]), TL, ovs, &(xo[2]));
111 TM = VFNMSI(Tq, Tp);
112 STM2(&(xo[6]), TM, ovs, &(xo[2]));
113 }
114 {
115 V TE, TG, TB, TF, Tz, TA;
116 TE = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TD, TC));
117 TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TC, TD));
118 Tz = VFNMS(LDK(KP250000000), Ty, Tr);
119 TA = VSUB(Tu, Tx);
120 TB = VFNMS(LDK(KP559016994), TA, Tz);
121 TF = VFMA(LDK(KP559016994), TA, Tz);
122 {
123 V TN, TO, TP, TQ;
124 TN = VFMAI(TE, TB);
125 STM2(&(xo[4]), TN, ovs, &(xo[0]));
126 STN2(&(xo[4]), TN, TM, ovs);
127 TO = VFNMSI(TG, TF);
128 STM2(&(xo[12]), TO, ovs, &(xo[0]));
129 STN2(&(xo[12]), TO, TK, ovs);
130 TP = VFNMSI(TE, TB);
131 STM2(&(xo[16]), TP, ovs, &(xo[0]));
132 STN2(&(xo[16]), TP, TL, ovs);
133 TQ = VFMAI(TG, TF);
134 STM2(&(xo[8]), TQ, ovs, &(xo[0]));
135 STN2(&(xo[8]), TQ, TH, ovs);
136 }
137 }
138 }
139 }
140 }
141 VLEAVE();
142 }
143
144 static const kdft_desc desc = { 10, XSIMD_STRING("n2fv_10"), {24, 4, 18, 0}, &GENUS, 0, 2, 0, 0 };
145
146 void XSIMD(codelet_n2fv_10) (planner *p) {
147 X(kdft_register) (p, n2fv_10, &desc);
148 }
149
150 #else
151
152 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 10 -name n2fv_10 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */
153
154 /*
155 * This function contains 42 FP additions, 12 FP multiplications,
156 * (or, 36 additions, 6 multiplications, 6 fused multiply/add),
157 * 36 stack variables, 4 constants, and 25 memory accesses
158 */
159 #include "dft/simd/n2f.h"
160
161 static void n2fv_10(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
162 {
163 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
164 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
165 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
166 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
167 {
168 INT i;
169 const R *xi;
170 R *xo;
171 xi = ri;
172 xo = ro;
173 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)) {
174 V Ti, Ty, Tm, Tn, Tw, Tt, Tz, TA, TB, T7, Te, Tj, Tg, Th;
175 Tg = LD(&(xi[0]), ivs, &(xi[0]));
176 Th = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
177 Ti = VSUB(Tg, Th);
178 Ty = VADD(Tg, Th);
179 {
180 V T3, Tu, Td, Ts, T6, Tv, Ta, Tr;
181 {
182 V T1, T2, Tb, Tc;
183 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
184 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
185 T3 = VSUB(T1, T2);
186 Tu = VADD(T1, T2);
187 Tb = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
188 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
189 Td = VSUB(Tb, Tc);
190 Ts = VADD(Tb, Tc);
191 }
192 {
193 V T4, T5, T8, T9;
194 T4 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
195 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
196 T6 = VSUB(T4, T5);
197 Tv = VADD(T4, T5);
198 T8 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
199 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
200 Ta = VSUB(T8, T9);
201 Tr = VADD(T8, T9);
202 }
203 Tm = VSUB(T3, T6);
204 Tn = VSUB(Ta, Td);
205 Tw = VSUB(Tu, Tv);
206 Tt = VSUB(Tr, Ts);
207 Tz = VADD(Tu, Tv);
208 TA = VADD(Tr, Ts);
209 TB = VADD(Tz, TA);
210 T7 = VADD(T3, T6);
211 Te = VADD(Ta, Td);
212 Tj = VADD(T7, Te);
213 }
214 {
215 V TH, TI, TK, TL, TM;
216 TH = VADD(Ti, Tj);
217 STM2(&(xo[10]), TH, ovs, &(xo[2]));
218 TI = VADD(Ty, TB);
219 STM2(&(xo[0]), TI, ovs, &(xo[0]));
220 {
221 V To, Tq, Tl, Tp, Tf, Tk, TJ;
222 To = VBYI(VFMA(LDK(KP951056516), Tm, VMUL(LDK(KP587785252), Tn)));
223 Tq = VBYI(VFNMS(LDK(KP587785252), Tm, VMUL(LDK(KP951056516), Tn)));
224 Tf = VMUL(LDK(KP559016994), VSUB(T7, Te));
225 Tk = VFNMS(LDK(KP250000000), Tj, Ti);
226 Tl = VADD(Tf, Tk);
227 Tp = VSUB(Tk, Tf);
228 TJ = VSUB(Tl, To);
229 STM2(&(xo[2]), TJ, ovs, &(xo[2]));
230 STN2(&(xo[0]), TI, TJ, ovs);
231 TK = VADD(Tq, Tp);
232 STM2(&(xo[14]), TK, ovs, &(xo[2]));
233 TL = VADD(To, Tl);
234 STM2(&(xo[18]), TL, ovs, &(xo[2]));
235 TM = VSUB(Tp, Tq);
236 STM2(&(xo[6]), TM, ovs, &(xo[2]));
237 }
238 {
239 V Tx, TF, TE, TG, TC, TD;
240 Tx = VBYI(VFNMS(LDK(KP587785252), Tw, VMUL(LDK(KP951056516), Tt)));
241 TF = VBYI(VFMA(LDK(KP951056516), Tw, VMUL(LDK(KP587785252), Tt)));
242 TC = VFNMS(LDK(KP250000000), TB, Ty);
243 TD = VMUL(LDK(KP559016994), VSUB(Tz, TA));
244 TE = VSUB(TC, TD);
245 TG = VADD(TD, TC);
246 {
247 V TN, TO, TP, TQ;
248 TN = VADD(Tx, TE);
249 STM2(&(xo[4]), TN, ovs, &(xo[0]));
250 STN2(&(xo[4]), TN, TM, ovs);
251 TO = VSUB(TG, TF);
252 STM2(&(xo[12]), TO, ovs, &(xo[0]));
253 STN2(&(xo[12]), TO, TK, ovs);
254 TP = VSUB(TE, Tx);
255 STM2(&(xo[16]), TP, ovs, &(xo[0]));
256 STN2(&(xo[16]), TP, TL, ovs);
257 TQ = VADD(TF, TG);
258 STM2(&(xo[8]), TQ, ovs, &(xo[0]));
259 STN2(&(xo[8]), TQ, TH, ovs);
260 }
261 }
262 }
263 }
264 }
265 VLEAVE();
266 }
267
268 static const kdft_desc desc = { 10, XSIMD_STRING("n2fv_10"), {36, 6, 6, 0}, &GENUS, 0, 2, 0, 0 };
269
270 void XSIMD(codelet_n2fv_10) (planner *p) {
271 X(kdft_register) (p, n2fv_10, &desc);
272 }
273
274 #endif