Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n2bv_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 08:05:11 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 -sign 1 -n 12 -name n2bv_12 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 48 FP additions, 20 FP multiplications,
32 * (or, 30 additions, 2 multiplications, 18 fused multiply/add),
33 * 33 stack variables, 2 constants, and 30 memory accesses
34 */
35 #include "dft/simd/n2b.h"
36
37 static void n2bv_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 = ii;
46 xo = io;
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, TJ, TB, Tq, Tp, Tg, Tl, TG, Ty, Tt, Ts;
49 {
50 V T1, T6, T4, Tz, T9, TA;
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 Tz = VSUB(T2, T3);
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 TA = VSUB(T7, T8);
63 }
64 T5 = VADD(T1, T4);
65 Ta = VADD(T6, T9);
66 TJ = VSUB(Tz, TA);
67 TB = VADD(Tz, TA);
68 Tq = VFNMS(LDK(KP500000000), T9, T6);
69 Tp = VFNMS(LDK(KP500000000), T4, T1);
70 }
71 {
72 V Tc, Th, Tf, Tw, Tk, Tx;
73 Tc = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
74 Th = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
75 {
76 V Td, Te, Ti, Tj;
77 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
78 Te = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
79 Tf = VADD(Td, Te);
80 Tw = VSUB(Td, Te);
81 Ti = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
82 Tj = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
83 Tk = VADD(Ti, Tj);
84 Tx = VSUB(Tj, Ti);
85 }
86 Tg = VADD(Tc, Tf);
87 Tl = VADD(Th, Tk);
88 TG = VADD(Tw, Tx);
89 Ty = VSUB(Tw, Tx);
90 Tt = VFNMS(LDK(KP500000000), Tk, Th);
91 Ts = VFNMS(LDK(KP500000000), Tf, Tc);
92 }
93 {
94 V TN, TO, TP, TQ, TR, TS;
95 {
96 V Tb, Tm, Tn, To;
97 Tb = VSUB(T5, Ta);
98 Tm = VSUB(Tg, Tl);
99 TN = VFNMSI(Tm, Tb);
100 STM2(&(xo[6]), TN, ovs, &(xo[2]));
101 TO = VFMAI(Tm, Tb);
102 STM2(&(xo[18]), TO, ovs, &(xo[2]));
103 Tn = VADD(T5, Ta);
104 To = VADD(Tg, Tl);
105 TP = VSUB(Tn, To);
106 STM2(&(xo[12]), TP, ovs, &(xo[0]));
107 TQ = VADD(Tn, To);
108 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
109 }
110 {
111 V TC, TE, Tv, TD, Tr, Tu, TT, TU;
112 TC = VMUL(LDK(KP866025403), VSUB(Ty, TB));
113 TE = VMUL(LDK(KP866025403), VADD(TB, Ty));
114 Tr = VADD(Tp, Tq);
115 Tu = VADD(Ts, Tt);
116 Tv = VSUB(Tr, Tu);
117 TD = VADD(Tr, Tu);
118 TR = VFNMSI(TC, Tv);
119 STM2(&(xo[20]), TR, ovs, &(xo[0]));
120 TS = VFMAI(TE, TD);
121 STM2(&(xo[8]), TS, ovs, &(xo[0]));
122 TT = VFMAI(TC, Tv);
123 STM2(&(xo[4]), TT, ovs, &(xo[0]));
124 STN2(&(xo[4]), TT, TN, ovs);
125 TU = VFNMSI(TE, TD);
126 STM2(&(xo[16]), TU, ovs, &(xo[0]));
127 STN2(&(xo[16]), TU, TO, ovs);
128 }
129 {
130 V TH, TL, TK, TM, TF, TI;
131 TF = VSUB(Tp, Tq);
132 TH = VFNMS(LDK(KP866025403), TG, TF);
133 TL = VFMA(LDK(KP866025403), TG, TF);
134 TI = VSUB(Ts, Tt);
135 TK = VFMA(LDK(KP866025403), TJ, TI);
136 TM = VFNMS(LDK(KP866025403), TJ, TI);
137 {
138 V TV, TW, TX, TY;
139 TV = VFMAI(TK, TH);
140 STM2(&(xo[2]), TV, ovs, &(xo[2]));
141 STN2(&(xo[0]), TQ, TV, ovs);
142 TW = VFNMSI(TM, TL);
143 STM2(&(xo[14]), TW, ovs, &(xo[2]));
144 STN2(&(xo[12]), TP, TW, ovs);
145 TX = VFNMSI(TK, TH);
146 STM2(&(xo[22]), TX, ovs, &(xo[2]));
147 STN2(&(xo[20]), TR, TX, ovs);
148 TY = VFMAI(TM, TL);
149 STM2(&(xo[10]), TY, ovs, &(xo[2]));
150 STN2(&(xo[8]), TS, TY, ovs);
151 }
152 }
153 }
154 }
155 }
156 VLEAVE();
157 }
158
159 static const kdft_desc desc = { 12, XSIMD_STRING("n2bv_12"), {30, 2, 18, 0}, &GENUS, 0, 2, 0, 0 };
160
161 void XSIMD(codelet_n2bv_12) (planner *p) {
162 X(kdft_register) (p, n2bv_12, &desc);
163 }
164
165 #else
166
167 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 12 -name n2bv_12 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
168
169 /*
170 * This function contains 48 FP additions, 8 FP multiplications,
171 * (or, 44 additions, 4 multiplications, 4 fused multiply/add),
172 * 33 stack variables, 2 constants, and 30 memory accesses
173 */
174 #include "dft/simd/n2b.h"
175
176 static void n2bv_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
177 {
178 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
179 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
180 {
181 INT i;
182 const R *xi;
183 R *xo;
184 xi = ii;
185 xo = io;
186 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)) {
187 V T5, Ta, TG, TF, Ty, Tm, Ti, Tp, TJ, TI, Tx, Ts;
188 {
189 V T1, T6, T4, Tk, T9, Tl;
190 T1 = LD(&(xi[0]), ivs, &(xi[0]));
191 T6 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
192 {
193 V T2, T3, T7, T8;
194 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
195 T3 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
196 T4 = VADD(T2, T3);
197 Tk = VSUB(T2, T3);
198 T7 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
199 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
200 T9 = VADD(T7, T8);
201 Tl = VSUB(T7, T8);
202 }
203 T5 = VFNMS(LDK(KP500000000), T4, T1);
204 Ta = VFNMS(LDK(KP500000000), T9, T6);
205 TG = VADD(T6, T9);
206 TF = VADD(T1, T4);
207 Ty = VADD(Tk, Tl);
208 Tm = VMUL(LDK(KP866025403), VSUB(Tk, Tl));
209 }
210 {
211 V Tn, Tq, Te, To, Th, Tr;
212 Tn = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
213 Tq = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
214 {
215 V Tc, Td, Tf, Tg;
216 Tc = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
217 Td = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
218 Te = VSUB(Tc, Td);
219 To = VADD(Tc, Td);
220 Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
221 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
222 Th = VSUB(Tf, Tg);
223 Tr = VADD(Tf, Tg);
224 }
225 Ti = VMUL(LDK(KP866025403), VSUB(Te, Th));
226 Tp = VFNMS(LDK(KP500000000), To, Tn);
227 TJ = VADD(Tq, Tr);
228 TI = VADD(Tn, To);
229 Tx = VADD(Te, Th);
230 Ts = VFNMS(LDK(KP500000000), Tr, Tq);
231 }
232 {
233 V TN, TO, TP, TQ, TR, TS;
234 {
235 V TH, TK, TL, TM;
236 TH = VSUB(TF, TG);
237 TK = VBYI(VSUB(TI, TJ));
238 TN = VSUB(TH, TK);
239 STM2(&(xo[6]), TN, ovs, &(xo[2]));
240 TO = VADD(TH, TK);
241 STM2(&(xo[18]), TO, ovs, &(xo[2]));
242 TL = VADD(TF, TG);
243 TM = VADD(TI, TJ);
244 TP = VSUB(TL, TM);
245 STM2(&(xo[12]), TP, ovs, &(xo[0]));
246 TQ = VADD(TL, TM);
247 STM2(&(xo[0]), TQ, ovs, &(xo[0]));
248 }
249 {
250 V Tj, Tv, Tu, Tw, Tb, Tt, TT, TU;
251 Tb = VSUB(T5, Ta);
252 Tj = VSUB(Tb, Ti);
253 Tv = VADD(Tb, Ti);
254 Tt = VSUB(Tp, Ts);
255 Tu = VBYI(VADD(Tm, Tt));
256 Tw = VBYI(VSUB(Tt, Tm));
257 TR = VSUB(Tj, Tu);
258 STM2(&(xo[22]), TR, ovs, &(xo[2]));
259 TS = VADD(Tv, Tw);
260 STM2(&(xo[10]), TS, ovs, &(xo[2]));
261 TT = VADD(Tj, Tu);
262 STM2(&(xo[2]), TT, ovs, &(xo[2]));
263 STN2(&(xo[0]), TQ, TT, ovs);
264 TU = VSUB(Tv, Tw);
265 STM2(&(xo[14]), TU, ovs, &(xo[2]));
266 STN2(&(xo[12]), TP, TU, ovs);
267 }
268 {
269 V Tz, TD, TC, TE, TA, TB;
270 Tz = VBYI(VMUL(LDK(KP866025403), VSUB(Tx, Ty)));
271 TD = VBYI(VMUL(LDK(KP866025403), VADD(Ty, Tx)));
272 TA = VADD(T5, Ta);
273 TB = VADD(Tp, Ts);
274 TC = VSUB(TA, TB);
275 TE = VADD(TA, TB);
276 {
277 V TV, TW, TX, TY;
278 TV = VADD(Tz, TC);
279 STM2(&(xo[4]), TV, ovs, &(xo[0]));
280 STN2(&(xo[4]), TV, TN, ovs);
281 TW = VSUB(TE, TD);
282 STM2(&(xo[16]), TW, ovs, &(xo[0]));
283 STN2(&(xo[16]), TW, TO, ovs);
284 TX = VSUB(TC, Tz);
285 STM2(&(xo[20]), TX, ovs, &(xo[0]));
286 STN2(&(xo[20]), TX, TR, ovs);
287 TY = VADD(TD, TE);
288 STM2(&(xo[8]), TY, ovs, &(xo[0]));
289 STN2(&(xo[8]), TY, TS, ovs);
290 }
291 }
292 }
293 }
294 }
295 VLEAVE();
296 }
297
298 static const kdft_desc desc = { 12, XSIMD_STRING("n2bv_12"), {44, 4, 4, 0}, &GENUS, 0, 2, 0, 0 };
299
300 void XSIMD(codelet_n2bv_12) (planner *p) {
301 X(kdft_register) (p, n2bv_12, &desc);
302 }
303
304 #endif