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 07:06:27 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_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 3 -dif -name hb_3 -include rdft/scalar/hb.h */
29
30 /*
31 * This function contains 16 FP additions, 14 FP multiplications,
32 * (or, 6 additions, 4 multiplications, 10 fused multiply/add),
33 * 17 stack variables, 2 constants, and 12 memory accesses
34 */
35 #include "rdft/scalar/hb.h"
36
hb_3(R * cr,R * ci,const R * W,stride rs,INT mb,INT me,INT ms)37 static void hb_3(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT m;
43 for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs)) {
44 E T1, T4, T6, Tg, Td, Te, T9, Tf;
45 {
46 E T2, T3, T7, T8;
47 T1 = cr[0];
48 T2 = cr[WS(rs, 1)];
49 T3 = ci[0];
50 T4 = T2 + T3;
51 T6 = FNMS(KP500000000, T4, T1);
52 Tg = T2 - T3;
53 Td = ci[WS(rs, 2)];
54 T7 = ci[WS(rs, 1)];
55 T8 = cr[WS(rs, 2)];
56 Te = T7 - T8;
57 T9 = T7 + T8;
58 Tf = FNMS(KP500000000, Te, Td);
59 }
60 cr[0] = T1 + T4;
61 ci[0] = Td + Te;
62 {
63 E Th, T5, Tb, Tc, Ti, Ta;
64 Th = FMA(KP866025403, Tg, Tf);
65 Ta = FNMS(KP866025403, T9, T6);
66 T5 = W[0];
67 Tb = T5 * Ta;
68 Tc = W[1];
69 Ti = Tc * Ta;
70 cr[WS(rs, 1)] = FNMS(Tc, Th, Tb);
71 ci[WS(rs, 1)] = FMA(T5, Th, Ti);
72 }
73 {
74 E Tn, Tj, Tl, Tm, To, Tk;
75 Tn = FNMS(KP866025403, Tg, Tf);
76 Tk = FMA(KP866025403, T9, T6);
77 Tj = W[2];
78 Tl = Tj * Tk;
79 Tm = W[3];
80 To = Tm * Tk;
81 cr[WS(rs, 2)] = FNMS(Tm, Tn, Tl);
82 ci[WS(rs, 2)] = FMA(Tj, Tn, To);
83 }
84 }
85 }
86 }
87
88 static const tw_instr twinstr[] = {
89 { TW_FULL, 1, 3 },
90 { TW_NEXT, 1, 0 }
91 };
92
93 static const hc2hc_desc desc = { 3, "hb_3", twinstr, &GENUS, { 6, 4, 10, 0 } };
94
X(codelet_hb_3)95 void X(codelet_hb_3) (planner *p) {
96 X(khc2hc_register) (p, hb_3, &desc);
97 }
98 #else
99
100 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 3 -dif -name hb_3 -include rdft/scalar/hb.h */
101
102 /*
103 * This function contains 16 FP additions, 12 FP multiplications,
104 * (or, 10 additions, 6 multiplications, 6 fused multiply/add),
105 * 15 stack variables, 2 constants, and 12 memory accesses
106 */
107 #include "rdft/scalar/hb.h"
108
hb_3(R * cr,R * ci,const R * W,stride rs,INT mb,INT me,INT ms)109 static void hb_3(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
110 {
111 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
112 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
113 {
114 INT m;
115 for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs)) {
116 E T1, T4, Ta, Te, T5, T8, Tb, Tf;
117 {
118 E T2, T3, T6, T7;
119 T1 = cr[0];
120 T2 = cr[WS(rs, 1)];
121 T3 = ci[0];
122 T4 = T2 + T3;
123 Ta = FNMS(KP500000000, T4, T1);
124 Te = KP866025403 * (T2 - T3);
125 T5 = ci[WS(rs, 2)];
126 T6 = ci[WS(rs, 1)];
127 T7 = cr[WS(rs, 2)];
128 T8 = T6 - T7;
129 Tb = KP866025403 * (T6 + T7);
130 Tf = FNMS(KP500000000, T8, T5);
131 }
132 cr[0] = T1 + T4;
133 ci[0] = T5 + T8;
134 {
135 E Tc, Tg, T9, Td;
136 Tc = Ta - Tb;
137 Tg = Te + Tf;
138 T9 = W[0];
139 Td = W[1];
140 cr[WS(rs, 1)] = FNMS(Td, Tg, T9 * Tc);
141 ci[WS(rs, 1)] = FMA(T9, Tg, Td * Tc);
142 }
143 {
144 E Ti, Tk, Th, Tj;
145 Ti = Ta + Tb;
146 Tk = Tf - Te;
147 Th = W[2];
148 Tj = W[3];
149 cr[WS(rs, 2)] = FNMS(Tj, Tk, Th * Ti);
150 ci[WS(rs, 2)] = FMA(Th, Tk, Tj * Ti);
151 }
152 }
153 }
154 }
155
156 static const tw_instr twinstr[] = {
157 { TW_FULL, 1, 3 },
158 { TW_NEXT, 1, 0 }
159 };
160
161 static const hc2hc_desc desc = { 3, "hb_3", twinstr, &GENUS, { 10, 6, 6, 0 } };
162
X(codelet_hb_3)163 void X(codelet_hb_3) (planner *p) {
164 X(khc2hc_register) (p, hb_3, &desc);
165 }
166 #endif
167