Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/t1fv_7.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 08:05:27 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_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 7 -name t1fv_7 -include dft/simd/t1f.h */
29
30 /*
31 * This function contains 36 FP additions, 36 FP multiplications,
32 * (or, 15 additions, 15 multiplications, 21 fused multiply/add),
33 * 30 stack variables, 6 constants, and 14 memory accesses
34 */
35 #include "dft/simd/t1f.h"
36
37 static void t1fv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
40 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
41 DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
42 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
43 DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
44 DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
45 {
46 INT m;
47 R *x;
48 x = ri;
49 for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(7, rs)) {
50 V T1, Tk, Tm, Tl, T6, Tg, Tb, Th, Tu, Tp;
51 T1 = LD(&(x[0]), ms, &(x[0]));
52 {
53 V T3, T5, Tf, Td, Ta, T8;
54 {
55 V T2, T4, Te, Tc, T9, T7;
56 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
57 T3 = BYTWJ(&(W[0]), T2);
58 T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
59 T5 = BYTWJ(&(W[TWVL * 10]), T4);
60 Te = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
61 Tf = BYTWJ(&(W[TWVL * 6]), Te);
62 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
63 Td = BYTWJ(&(W[TWVL * 4]), Tc);
64 T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
65 Ta = BYTWJ(&(W[TWVL * 8]), T9);
66 T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
67 T8 = BYTWJ(&(W[TWVL * 2]), T7);
68 }
69 Tk = VSUB(T5, T3);
70 Tm = VSUB(Ta, T8);
71 Tl = VSUB(Tf, Td);
72 T6 = VADD(T3, T5);
73 Tg = VADD(Td, Tf);
74 Tb = VADD(T8, Ta);
75 Th = VFNMS(LDK(KP356895867), T6, Tg);
76 Tu = VFNMS(LDK(KP356895867), Tg, Tb);
77 Tp = VFNMS(LDK(KP356895867), Tb, T6);
78 }
79 ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0]));
80 {
81 V Tw, Ty, Tv, Tx;
82 Tv = VFNMS(LDK(KP692021471), Tu, T6);
83 Tw = VFNMS(LDK(KP900968867), Tv, T1);
84 Tx = VFNMS(LDK(KP554958132), Tk, Tm);
85 Ty = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tx, Tl));
86 ST(&(x[WS(rs, 4)]), VFNMSI(Ty, Tw), ms, &(x[0]));
87 ST(&(x[WS(rs, 3)]), VFMAI(Ty, Tw), ms, &(x[WS(rs, 1)]));
88 }
89 {
90 V Tj, To, Ti, Tn;
91 Ti = VFNMS(LDK(KP692021471), Th, Tb);
92 Tj = VFNMS(LDK(KP900968867), Ti, T1);
93 Tn = VFMA(LDK(KP554958132), Tm, Tl);
94 To = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tn, Tk));
95 ST(&(x[WS(rs, 5)]), VFNMSI(To, Tj), ms, &(x[WS(rs, 1)]));
96 ST(&(x[WS(rs, 2)]), VFMAI(To, Tj), ms, &(x[0]));
97 }
98 {
99 V Tr, Tt, Tq, Ts;
100 Tq = VFNMS(LDK(KP692021471), Tp, Tg);
101 Tr = VFNMS(LDK(KP900968867), Tq, T1);
102 Ts = VFMA(LDK(KP554958132), Tl, Tk);
103 Tt = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Ts, Tm));
104 ST(&(x[WS(rs, 6)]), VFNMSI(Tt, Tr), ms, &(x[0]));
105 ST(&(x[WS(rs, 1)]), VFMAI(Tt, Tr), ms, &(x[WS(rs, 1)]));
106 }
107 }
108 }
109 VLEAVE();
110 }
111
112 static const tw_instr twinstr[] = {
113 VTW(0, 1),
114 VTW(0, 2),
115 VTW(0, 3),
116 VTW(0, 4),
117 VTW(0, 5),
118 VTW(0, 6),
119 {TW_NEXT, VL, 0}
120 };
121
122 static const ct_desc desc = { 7, XSIMD_STRING("t1fv_7"), twinstr, &GENUS, {15, 15, 21, 0}, 0, 0, 0 };
123
124 void XSIMD(codelet_t1fv_7) (planner *p) {
125 X(kdft_dit_register) (p, t1fv_7, &desc);
126 }
127 #else
128
129 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 7 -name t1fv_7 -include dft/simd/t1f.h */
130
131 /*
132 * This function contains 36 FP additions, 30 FP multiplications,
133 * (or, 24 additions, 18 multiplications, 12 fused multiply/add),
134 * 21 stack variables, 6 constants, and 14 memory accesses
135 */
136 #include "dft/simd/t1f.h"
137
138 static void t1fv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
139 {
140 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
141 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
142 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
143 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
144 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
145 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
146 {
147 INT m;
148 R *x;
149 x = ri;
150 for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(7, rs)) {
151 V T1, Tg, Tj, T6, Ti, Tb, Tk, Tp, To;
152 T1 = LD(&(x[0]), ms, &(x[0]));
153 {
154 V Td, Tf, Tc, Te;
155 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
156 Td = BYTWJ(&(W[TWVL * 4]), Tc);
157 Te = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
158 Tf = BYTWJ(&(W[TWVL * 6]), Te);
159 Tg = VADD(Td, Tf);
160 Tj = VSUB(Tf, Td);
161 }
162 {
163 V T3, T5, T2, T4;
164 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
165 T3 = BYTWJ(&(W[0]), T2);
166 T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
167 T5 = BYTWJ(&(W[TWVL * 10]), T4);
168 T6 = VADD(T3, T5);
169 Ti = VSUB(T5, T3);
170 }
171 {
172 V T8, Ta, T7, T9;
173 T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
174 T8 = BYTWJ(&(W[TWVL * 2]), T7);
175 T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
176 Ta = BYTWJ(&(W[TWVL * 8]), T9);
177 Tb = VADD(T8, Ta);
178 Tk = VSUB(Ta, T8);
179 }
180 ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0]));
181 Tp = VBYI(VFMA(LDK(KP433883739), Ti, VFNMS(LDK(KP781831482), Tk, VMUL(LDK(KP974927912), Tj))));
182 To = VFMA(LDK(KP623489801), Tb, VFNMS(LDK(KP222520933), Tg, VFNMS(LDK(KP900968867), T6, T1)));
183 ST(&(x[WS(rs, 4)]), VSUB(To, Tp), ms, &(x[0]));
184 ST(&(x[WS(rs, 3)]), VADD(To, Tp), ms, &(x[WS(rs, 1)]));
185 {
186 V Tl, Th, Tn, Tm;
187 Tl = VBYI(VFNMS(LDK(KP781831482), Tj, VFNMS(LDK(KP433883739), Tk, VMUL(LDK(KP974927912), Ti))));
188 Th = VFMA(LDK(KP623489801), Tg, VFNMS(LDK(KP900968867), Tb, VFNMS(LDK(KP222520933), T6, T1)));
189 ST(&(x[WS(rs, 5)]), VSUB(Th, Tl), ms, &(x[WS(rs, 1)]));
190 ST(&(x[WS(rs, 2)]), VADD(Th, Tl), ms, &(x[0]));
191 Tn = VBYI(VFMA(LDK(KP781831482), Ti, VFMA(LDK(KP974927912), Tk, VMUL(LDK(KP433883739), Tj))));
192 Tm = VFMA(LDK(KP623489801), T6, VFNMS(LDK(KP900968867), Tg, VFNMS(LDK(KP222520933), Tb, T1)));
193 ST(&(x[WS(rs, 6)]), VSUB(Tm, Tn), ms, &(x[0]));
194 ST(&(x[WS(rs, 1)]), VADD(Tm, Tn), ms, &(x[WS(rs, 1)]));
195 }
196 }
197 }
198 VLEAVE();
199 }
200
201 static const tw_instr twinstr[] = {
202 VTW(0, 1),
203 VTW(0, 2),
204 VTW(0, 3),
205 VTW(0, 4),
206 VTW(0, 5),
207 VTW(0, 6),
208 {TW_NEXT, VL, 0}
209 };
210
211 static const ct_desc desc = { 7, XSIMD_STRING("t1fv_7"), twinstr, &GENUS, {24, 18, 12, 0}, 0, 0, 0 };
212
213 void XSIMD(codelet_t1fv_7) (planner *p) {
214 X(kdft_dit_register) (p, t1fv_7, &desc);
215 }
216 #endif