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 07:05:22 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_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name t2fv_4 -include dft/simd/t2f.h */
29
30 /*
31 * This function contains 11 FP additions, 8 FP multiplications,
32 * (or, 9 additions, 6 multiplications, 2 fused multiply/add),
33 * 13 stack variables, 0 constants, and 8 memory accesses
34 */
35 #include "dft/simd/t2f.h"
36
t2fv_4(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)37 static void t2fv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 {
40 INT m;
41 R *x;
42 x = ri;
43 for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(4, rs)) {
44 V T1, T8, T3, T6, T7, T2, T5;
45 T1 = LD(&(x[0]), ms, &(x[0]));
46 T7 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
47 T8 = BYTWJ(&(W[TWVL * 4]), T7);
48 T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
49 T3 = BYTWJ(&(W[TWVL * 2]), T2);
50 T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
51 T6 = BYTWJ(&(W[0]), T5);
52 {
53 V T4, T9, Ta, Tb;
54 T4 = VSUB(T1, T3);
55 T9 = VSUB(T6, T8);
56 ST(&(x[WS(rs, 1)]), VFNMSI(T9, T4), ms, &(x[WS(rs, 1)]));
57 ST(&(x[WS(rs, 3)]), VFMAI(T9, T4), ms, &(x[WS(rs, 1)]));
58 Ta = VADD(T1, T3);
59 Tb = VADD(T6, T8);
60 ST(&(x[WS(rs, 2)]), VSUB(Ta, Tb), ms, &(x[0]));
61 ST(&(x[0]), VADD(Ta, Tb), ms, &(x[0]));
62 }
63 }
64 }
65 VLEAVE();
66 }
67
68 static const tw_instr twinstr[] = {
69 VTW(0, 1),
70 VTW(0, 2),
71 VTW(0, 3),
72 { TW_NEXT, VL, 0 }
73 };
74
75 static const ct_desc desc = { 4, XSIMD_STRING("t2fv_4"), twinstr, &GENUS, { 9, 6, 2, 0 }, 0, 0, 0 };
76
XSIMD(codelet_t2fv_4)77 void XSIMD(codelet_t2fv_4) (planner *p) {
78 X(kdft_dit_register) (p, t2fv_4, &desc);
79 }
80 #else
81
82 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name t2fv_4 -include dft/simd/t2f.h */
83
84 /*
85 * This function contains 11 FP additions, 6 FP multiplications,
86 * (or, 11 additions, 6 multiplications, 0 fused multiply/add),
87 * 13 stack variables, 0 constants, and 8 memory accesses
88 */
89 #include "dft/simd/t2f.h"
90
t2fv_4(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)91 static void t2fv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
92 {
93 {
94 INT m;
95 R *x;
96 x = ri;
97 for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(4, rs)) {
98 V T1, T8, T3, T6, T7, T2, T5;
99 T1 = LD(&(x[0]), ms, &(x[0]));
100 T7 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
101 T8 = BYTWJ(&(W[TWVL * 4]), T7);
102 T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
103 T3 = BYTWJ(&(W[TWVL * 2]), T2);
104 T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
105 T6 = BYTWJ(&(W[0]), T5);
106 {
107 V T4, T9, Ta, Tb;
108 T4 = VSUB(T1, T3);
109 T9 = VBYI(VSUB(T6, T8));
110 ST(&(x[WS(rs, 1)]), VSUB(T4, T9), ms, &(x[WS(rs, 1)]));
111 ST(&(x[WS(rs, 3)]), VADD(T4, T9), ms, &(x[WS(rs, 1)]));
112 Ta = VADD(T1, T3);
113 Tb = VADD(T6, T8);
114 ST(&(x[WS(rs, 2)]), VSUB(Ta, Tb), ms, &(x[0]));
115 ST(&(x[0]), VADD(Ta, Tb), ms, &(x[0]));
116 }
117 }
118 }
119 VLEAVE();
120 }
121
122 static const tw_instr twinstr[] = {
123 VTW(0, 1),
124 VTW(0, 2),
125 VTW(0, 3),
126 { TW_NEXT, VL, 0 }
127 };
128
129 static const ct_desc desc = { 4, XSIMD_STRING("t2fv_4"), twinstr, &GENUS, { 11, 6, 0, 0 }, 0, 0, 0 };
130
XSIMD(codelet_t2fv_4)131 void XSIMD(codelet_t2fv_4) (planner *p) {
132 X(kdft_dit_register) (p, t2fv_4, &desc);
133 }
134 #endif
135