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 Dec 10 07:06:56 EST 2020 */
23
24 #include "rdft/codelet-rdft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_r2r.native -fma -compact -variables 4 -pipeline-latency 4 -redft10 -n 8 -name e10_8 -include rdft/scalar/r2r.h */
29
30 /*
31 * This function contains 26 FP additions, 18 FP multiplications,
32 * (or, 16 additions, 8 multiplications, 10 fused multiply/add),
33 * 28 stack variables, 9 constants, and 16 memory accesses
34 */
35 #include "rdft/scalar/r2r.h"
36
e10_8(const R * I,R * O,stride is,stride os,INT v,INT ivs,INT ovs)37 static void e10_8(const R *I, R *O, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
40 DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
41 DK(KP198912367, +0.198912367379658006911597622644676228597850501);
42 DK(KP1_961570560, +1.961570560806460898252364472268478073947867462);
43 DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
44 DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
45 DK(KP668178637, +0.668178637919298919997757686523080761552472251);
46 DK(KP1_662939224, +1.662939224605090474157576755235811513477121624);
47 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
48 {
49 INT i;
50 for (i = v; i > 0; i = i - 1, I = I + ivs, O = O + ovs, MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
51 E T3, Tj, Te, Tk, Ta, Tn, Tf, Tm;
52 {
53 E T1, T2, Tc, Td;
54 T1 = I[0];
55 T2 = I[WS(is, 7)];
56 T3 = T1 - T2;
57 Tj = T1 + T2;
58 Tc = I[WS(is, 4)];
59 Td = I[WS(is, 3)];
60 Te = Tc - Td;
61 Tk = Tc + Td;
62 {
63 E T4, T5, T6, T7, T8, T9;
64 T4 = I[WS(is, 2)];
65 T5 = I[WS(is, 5)];
66 T6 = T4 - T5;
67 T7 = I[WS(is, 1)];
68 T8 = I[WS(is, 6)];
69 T9 = T7 - T8;
70 Ta = T6 + T9;
71 Tn = T7 + T8;
72 Tf = T6 - T9;
73 Tm = T4 + T5;
74 }
75 }
76 {
77 E Tb, Tg, Tp, Tq;
78 Tb = FNMS(KP707106781, Ta, T3);
79 Tg = FNMS(KP707106781, Tf, Te);
80 O[WS(os, 3)] = KP1_662939224 * (FMA(KP668178637, Tg, Tb));
81 O[WS(os, 5)] = -(KP1_662939224 * (FNMS(KP668178637, Tb, Tg)));
82 Tp = Tj + Tk;
83 Tq = Tm + Tn;
84 O[WS(os, 4)] = KP1_414213562 * (Tp - Tq);
85 O[0] = KP2_000000000 * (Tp + Tq);
86 }
87 {
88 E Th, Ti, Tl, To;
89 Th = FMA(KP707106781, Ta, T3);
90 Ti = FMA(KP707106781, Tf, Te);
91 O[WS(os, 1)] = KP1_961570560 * (FNMS(KP198912367, Ti, Th));
92 O[WS(os, 7)] = KP1_961570560 * (FMA(KP198912367, Th, Ti));
93 Tl = Tj - Tk;
94 To = Tm - Tn;
95 O[WS(os, 2)] = KP1_847759065 * (FNMS(KP414213562, To, Tl));
96 O[WS(os, 6)] = KP1_847759065 * (FMA(KP414213562, Tl, To));
97 }
98 }
99 }
100 }
101
102 static const kr2r_desc desc = { 8, "e10_8", { 16, 8, 10, 0 }, &GENUS, REDFT10 };
103
X(codelet_e10_8)104 void X(codelet_e10_8) (planner *p) { X(kr2r_register) (p, e10_8, &desc);
105 }
106
107 #else
108
109 /* Generated by: ../../../genfft/gen_r2r.native -compact -variables 4 -pipeline-latency 4 -redft10 -n 8 -name e10_8 -include rdft/scalar/r2r.h */
110
111 /*
112 * This function contains 26 FP additions, 16 FP multiplications,
113 * (or, 20 additions, 10 multiplications, 6 fused multiply/add),
114 * 28 stack variables, 9 constants, and 16 memory accesses
115 */
116 #include "rdft/scalar/r2r.h"
117
e10_8(const R * I,R * O,stride is,stride os,INT v,INT ivs,INT ovs)118 static void e10_8(const R *I, R *O, stride is, stride os, INT v, INT ivs, INT ovs)
119 {
120 DK(KP765366864, +0.765366864730179543456919968060797733522689125);
121 DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
122 DK(KP390180644, +0.390180644032256535696569736954044481855383236);
123 DK(KP1_961570560, +1.961570560806460898252364472268478073947867462);
124 DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
125 DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
126 DK(KP1_111140466, +1.111140466039204449485661627897065748749874382);
127 DK(KP1_662939224, +1.662939224605090474157576755235811513477121624);
128 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
129 {
130 INT i;
131 for (i = v; i > 0; i = i - 1, I = I + ivs, O = O + ovs, MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
132 E T3, Tj, Tf, Tk, Ta, Tn, Tc, Tm;
133 {
134 E T1, T2, Td, Te;
135 T1 = I[0];
136 T2 = I[WS(is, 7)];
137 T3 = T1 - T2;
138 Tj = T1 + T2;
139 Td = I[WS(is, 4)];
140 Te = I[WS(is, 3)];
141 Tf = Td - Te;
142 Tk = Td + Te;
143 {
144 E T4, T5, T6, T7, T8, T9;
145 T4 = I[WS(is, 2)];
146 T5 = I[WS(is, 5)];
147 T6 = T4 - T5;
148 T7 = I[WS(is, 1)];
149 T8 = I[WS(is, 6)];
150 T9 = T7 - T8;
151 Ta = KP707106781 * (T6 + T9);
152 Tn = T7 + T8;
153 Tc = KP707106781 * (T6 - T9);
154 Tm = T4 + T5;
155 }
156 }
157 {
158 E Tb, Tg, Tp, Tq;
159 Tb = T3 - Ta;
160 Tg = Tc - Tf;
161 O[WS(os, 3)] = FNMS(KP1_111140466, Tg, KP1_662939224 * Tb);
162 O[WS(os, 5)] = FMA(KP1_662939224, Tg, KP1_111140466 * Tb);
163 Tp = Tj + Tk;
164 Tq = Tm + Tn;
165 O[WS(os, 4)] = KP1_414213562 * (Tp - Tq);
166 O[0] = KP2_000000000 * (Tp + Tq);
167 }
168 {
169 E Th, Ti, Tl, To;
170 Th = T3 + Ta;
171 Ti = Tf + Tc;
172 O[WS(os, 1)] = FNMS(KP390180644, Ti, KP1_961570560 * Th);
173 O[WS(os, 7)] = FMA(KP1_961570560, Ti, KP390180644 * Th);
174 Tl = Tj - Tk;
175 To = Tm - Tn;
176 O[WS(os, 2)] = FNMS(KP765366864, To, KP1_847759065 * Tl);
177 O[WS(os, 6)] = FMA(KP765366864, Tl, KP1_847759065 * To);
178 }
179 }
180 }
181 }
182
183 static const kr2r_desc desc = { 8, "e10_8", { 20, 10, 6, 0 }, &GENUS, REDFT10 };
184
X(codelet_e10_8)185 void X(codelet_e10_8) (planner *p) { X(kr2r_register) (p, e10_8, &desc);
186 }
187
188 #endif
189