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 07:04:40 EST 2020 */
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 5 -name n1fv_5 -include dft/simd/n1f.h */
29
30 /*
31 * This function contains 16 FP additions, 11 FP multiplications,
32 * (or, 7 additions, 2 multiplications, 9 fused multiply/add),
33 * 18 stack variables, 4 constants, and 10 memory accesses
34 */
35 #include "dft/simd/n1f.h"
36
n1fv_5(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1fv_5(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(10, is), MAKE_VOLATILE_STRIDE(10, os)) {
50 V T1, T8, Td, Ta, Tc;
51 T1 = LD(&(xi[0]), ivs, &(xi[0]));
52 {
53 V T2, T3, T4, T5, T6, T7;
54 T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
55 T3 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
56 T4 = VADD(T2, T3);
57 T5 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
58 T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
59 T7 = VADD(T5, T6);
60 T8 = VADD(T4, T7);
61 Td = VSUB(T5, T6);
62 Ta = VSUB(T4, T7);
63 Tc = VSUB(T2, T3);
64 }
65 ST(&(xo[0]), VADD(T1, T8), ovs, &(xo[0]));
66 {
67 V Te, Tg, Tb, Tf, T9;
68 Te = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Td, Tc));
69 Tg = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tc, Td));
70 T9 = VFNMS(LDK(KP250000000), T8, T1);
71 Tb = VFMA(LDK(KP559016994), Ta, T9);
72 Tf = VFNMS(LDK(KP559016994), Ta, T9);
73 ST(&(xo[WS(os, 1)]), VFNMSI(Te, Tb), ovs, &(xo[WS(os, 1)]));
74 ST(&(xo[WS(os, 3)]), VFNMSI(Tg, Tf), ovs, &(xo[WS(os, 1)]));
75 ST(&(xo[WS(os, 4)]), VFMAI(Te, Tb), ovs, &(xo[0]));
76 ST(&(xo[WS(os, 2)]), VFMAI(Tg, Tf), ovs, &(xo[0]));
77 }
78 }
79 }
80 VLEAVE();
81 }
82
83 static const kdft_desc desc = { 5, XSIMD_STRING("n1fv_5"), { 7, 2, 9, 0 }, &GENUS, 0, 0, 0, 0 };
84
XSIMD(codelet_n1fv_5)85 void XSIMD(codelet_n1fv_5) (planner *p) { X(kdft_register) (p, n1fv_5, &desc);
86 }
87
88 #else
89
90 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 5 -name n1fv_5 -include dft/simd/n1f.h */
91
92 /*
93 * This function contains 16 FP additions, 6 FP multiplications,
94 * (or, 13 additions, 3 multiplications, 3 fused multiply/add),
95 * 18 stack variables, 4 constants, and 10 memory accesses
96 */
97 #include "dft/simd/n1f.h"
98
n1fv_5(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)99 static void n1fv_5(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
100 {
101 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
102 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
103 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
104 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
105 {
106 INT i;
107 const R *xi;
108 R *xo;
109 xi = ri;
110 xo = ro;
111 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(10, is), MAKE_VOLATILE_STRIDE(10, os)) {
112 V T8, T7, Td, T9, Tc;
113 T8 = LD(&(xi[0]), ivs, &(xi[0]));
114 {
115 V T1, T2, T3, T4, T5, T6;
116 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
117 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
118 T3 = VADD(T1, T2);
119 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
120 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
121 T6 = VADD(T4, T5);
122 T7 = VMUL(LDK(KP559016994), VSUB(T3, T6));
123 Td = VSUB(T4, T5);
124 T9 = VADD(T3, T6);
125 Tc = VSUB(T1, T2);
126 }
127 ST(&(xo[0]), VADD(T8, T9), ovs, &(xo[0]));
128 {
129 V Te, Tf, Tb, Tg, Ta;
130 Te = VBYI(VFMA(LDK(KP951056516), Tc, VMUL(LDK(KP587785252), Td)));
131 Tf = VBYI(VFNMS(LDK(KP587785252), Tc, VMUL(LDK(KP951056516), Td)));
132 Ta = VFNMS(LDK(KP250000000), T9, T8);
133 Tb = VADD(T7, Ta);
134 Tg = VSUB(Ta, T7);
135 ST(&(xo[WS(os, 1)]), VSUB(Tb, Te), ovs, &(xo[WS(os, 1)]));
136 ST(&(xo[WS(os, 3)]), VSUB(Tg, Tf), ovs, &(xo[WS(os, 1)]));
137 ST(&(xo[WS(os, 4)]), VADD(Te, Tb), ovs, &(xo[0]));
138 ST(&(xo[WS(os, 2)]), VADD(Tf, Tg), ovs, &(xo[0]));
139 }
140 }
141 }
142 VLEAVE();
143 }
144
145 static const kdft_desc desc = { 5, XSIMD_STRING("n1fv_5"), { 13, 3, 3, 0 }, &GENUS, 0, 0, 0, 0 };
146
XSIMD(codelet_n1fv_5)147 void XSIMD(codelet_n1fv_5) (planner *p) { X(kdft_register) (p, n1fv_5, &desc);
148 }
149
150 #endif
151