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 07:04:56 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 -sign 1 -n 4 -name n2bv_4 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 8 FP additions, 2 FP multiplications,
32 * (or, 6 additions, 0 multiplications, 2 fused multiply/add),
33 * 11 stack variables, 0 constants, and 10 memory accesses
34 */
35 #include "dft/simd/n2b.h"
36
n2bv_4(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n2bv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 {
40 INT i;
41 const R *xi;
42 R *xo;
43 xi = ii;
44 xo = io;
45 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(8, is), MAKE_VOLATILE_STRIDE(8, os)) {
46 V T3, T7, T6, T8;
47 {
48 V T1, T2, T4, T5;
49 T1 = LD(&(xi[0]), ivs, &(xi[0]));
50 T2 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
51 T3 = VSUB(T1, T2);
52 T7 = VADD(T1, T2);
53 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
54 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
55 T6 = VSUB(T4, T5);
56 T8 = VADD(T4, T5);
57 }
58 {
59 V T9, Ta, Tb, Tc;
60 T9 = VFNMSI(T6, T3);
61 STM2(&(xo[6]), T9, ovs, &(xo[2]));
62 Ta = VADD(T7, T8);
63 STM2(&(xo[0]), Ta, ovs, &(xo[0]));
64 Tb = VFMAI(T6, T3);
65 STM2(&(xo[2]), Tb, ovs, &(xo[2]));
66 STN2(&(xo[0]), Ta, Tb, ovs);
67 Tc = VSUB(T7, T8);
68 STM2(&(xo[4]), Tc, ovs, &(xo[0]));
69 STN2(&(xo[4]), Tc, T9, ovs);
70 }
71 }
72 }
73 VLEAVE();
74 }
75
76 static const kdft_desc desc = { 4, XSIMD_STRING("n2bv_4"), { 6, 0, 2, 0 }, &GENUS, 0, 2, 0, 0 };
77
XSIMD(codelet_n2bv_4)78 void XSIMD(codelet_n2bv_4) (planner *p) { X(kdft_register) (p, n2bv_4, &desc);
79 }
80
81 #else
82
83 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 4 -name n2bv_4 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
84
85 /*
86 * This function contains 8 FP additions, 0 FP multiplications,
87 * (or, 8 additions, 0 multiplications, 0 fused multiply/add),
88 * 11 stack variables, 0 constants, and 10 memory accesses
89 */
90 #include "dft/simd/n2b.h"
91
n2bv_4(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)92 static void n2bv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
93 {
94 {
95 INT i;
96 const R *xi;
97 R *xo;
98 xi = ii;
99 xo = io;
100 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(8, is), MAKE_VOLATILE_STRIDE(8, os)) {
101 V T3, T7, T6, T8;
102 {
103 V T1, T2, T4, T5;
104 T1 = LD(&(xi[0]), ivs, &(xi[0]));
105 T2 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
106 T3 = VSUB(T1, T2);
107 T7 = VADD(T1, T2);
108 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
109 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
110 T6 = VBYI(VSUB(T4, T5));
111 T8 = VADD(T4, T5);
112 }
113 {
114 V T9, Ta, Tb, Tc;
115 T9 = VSUB(T3, T6);
116 STM2(&(xo[6]), T9, ovs, &(xo[2]));
117 Ta = VADD(T7, T8);
118 STM2(&(xo[0]), Ta, ovs, &(xo[0]));
119 Tb = VADD(T3, T6);
120 STM2(&(xo[2]), Tb, ovs, &(xo[2]));
121 STN2(&(xo[0]), Ta, Tb, ovs);
122 Tc = VSUB(T7, T8);
123 STM2(&(xo[4]), Tc, ovs, &(xo[0]));
124 STN2(&(xo[4]), Tc, T9, ovs);
125 }
126 }
127 }
128 VLEAVE();
129 }
130
131 static const kdft_desc desc = { 4, XSIMD_STRING("n2bv_4"), { 8, 0, 0, 0 }, &GENUS, 0, 2, 0, 0 };
132
XSIMD(codelet_n2bv_4)133 void XSIMD(codelet_n2bv_4) (planner *p) { X(kdft_register) (p, n2bv_4, &desc);
134 }
135
136 #endif
137