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 07:05:28 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 3 -name t1buv_3 -include dft/simd/t1bu.h -sign 1 */
29
30 /*
31 * This function contains 8 FP additions, 8 FP multiplications,
32 * (or, 5 additions, 5 multiplications, 3 fused multiply/add),
33 * 12 stack variables, 2 constants, and 6 memory accesses
34 */
35 #include "dft/simd/t1bu.h"
36
t1buv_3(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)37 static void t1buv_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT m;
43 R *x;
44 x = ii;
45 for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(3, rs)) {
46 V T1, T3, T5, T6, T2, T4, T7, T8;
47 T1 = LD(&(x[0]), ms, &(x[0]));
48 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
49 T3 = BYTW(&(W[0]), T2);
50 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
51 T5 = BYTW(&(W[TWVL * 2]), T4);
52 T6 = VADD(T3, T5);
53 ST(&(x[0]), VADD(T1, T6), ms, &(x[0]));
54 T7 = VFNMS(LDK(KP500000000), T6, T1);
55 T8 = VMUL(LDK(KP866025403), VSUB(T3, T5));
56 ST(&(x[WS(rs, 1)]), VFMAI(T8, T7), ms, &(x[WS(rs, 1)]));
57 ST(&(x[WS(rs, 2)]), VFNMSI(T8, T7), ms, &(x[0]));
58 }
59 }
60 VLEAVE();
61 }
62
63 static const tw_instr twinstr[] = {
64 VTW(0, 1),
65 VTW(0, 2),
66 { TW_NEXT, VL, 0 }
67 };
68
69 static const ct_desc desc = { 3, XSIMD_STRING("t1buv_3"), twinstr, &GENUS, { 5, 5, 3, 0 }, 0, 0, 0 };
70
XSIMD(codelet_t1buv_3)71 void XSIMD(codelet_t1buv_3) (planner *p) {
72 X(kdft_dit_register) (p, t1buv_3, &desc);
73 }
74 #else
75
76 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 3 -name t1buv_3 -include dft/simd/t1bu.h -sign 1 */
77
78 /*
79 * This function contains 8 FP additions, 6 FP multiplications,
80 * (or, 7 additions, 5 multiplications, 1 fused multiply/add),
81 * 12 stack variables, 2 constants, and 6 memory accesses
82 */
83 #include "dft/simd/t1bu.h"
84
t1buv_3(R * ri,R * ii,const R * W,stride rs,INT mb,INT me,INT ms)85 static void t1buv_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
86 {
87 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
88 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
89 {
90 INT m;
91 R *x;
92 x = ii;
93 for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(3, rs)) {
94 V T6, T2, T4, T7, T1, T3, T5, T8;
95 T6 = LD(&(x[0]), ms, &(x[0]));
96 T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
97 T2 = BYTW(&(W[0]), T1);
98 T3 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
99 T4 = BYTW(&(W[TWVL * 2]), T3);
100 T7 = VADD(T2, T4);
101 ST(&(x[0]), VADD(T6, T7), ms, &(x[0]));
102 T5 = VBYI(VMUL(LDK(KP866025403), VSUB(T2, T4)));
103 T8 = VFNMS(LDK(KP500000000), T7, T6);
104 ST(&(x[WS(rs, 1)]), VADD(T5, T8), ms, &(x[WS(rs, 1)]));
105 ST(&(x[WS(rs, 2)]), VSUB(T8, T5), ms, &(x[0]));
106 }
107 }
108 VLEAVE();
109 }
110
111 static const tw_instr twinstr[] = {
112 VTW(0, 1),
113 VTW(0, 2),
114 { TW_NEXT, VL, 0 }
115 };
116
117 static const ct_desc desc = { 3, XSIMD_STRING("t1buv_3"), twinstr, &GENUS, { 7, 5, 1, 0 }, 0, 0, 0 };
118
XSIMD(codelet_t1buv_3)119 void XSIMD(codelet_t1buv_3) (planner *p) {
120 X(kdft_dit_register) (p, t1buv_3, &desc);
121 }
122 #endif
123