Chris@82
|
1 /*
|
Chris@82
|
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
|
Chris@82
|
3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
|
Chris@82
|
4 *
|
Chris@82
|
5 * This program is free software; you can redistribute it and/or modify
|
Chris@82
|
6 * it under the terms of the GNU General Public License as published by
|
Chris@82
|
7 * the Free Software Foundation; either version 2 of the License, or
|
Chris@82
|
8 * (at your option) any later version.
|
Chris@82
|
9 *
|
Chris@82
|
10 * This program is distributed in the hope that it will be useful,
|
Chris@82
|
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
Chris@82
|
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
Chris@82
|
13 * GNU General Public License for more details.
|
Chris@82
|
14 *
|
Chris@82
|
15 * You should have received a copy of the GNU General Public License
|
Chris@82
|
16 * along with this program; if not, write to the Free Software
|
Chris@82
|
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
Chris@82
|
18 *
|
Chris@82
|
19 */
|
Chris@82
|
20
|
Chris@82
|
21 /* This file was automatically generated --- DO NOT EDIT */
|
Chris@82
|
22 /* Generated on Thu May 24 08:05:26 EDT 2018 */
|
Chris@82
|
23
|
Chris@82
|
24 #include "dft/codelet-dft.h"
|
Chris@82
|
25
|
Chris@82
|
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
|
Chris@82
|
27
|
Chris@82
|
28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 3 -name t1fv_3 -include dft/simd/t1f.h */
|
Chris@82
|
29
|
Chris@82
|
30 /*
|
Chris@82
|
31 * This function contains 8 FP additions, 8 FP multiplications,
|
Chris@82
|
32 * (or, 5 additions, 5 multiplications, 3 fused multiply/add),
|
Chris@82
|
33 * 12 stack variables, 2 constants, and 6 memory accesses
|
Chris@82
|
34 */
|
Chris@82
|
35 #include "dft/simd/t1f.h"
|
Chris@82
|
36
|
Chris@82
|
37 static void t1fv_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
|
Chris@82
|
38 {
|
Chris@82
|
39 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
|
Chris@82
|
40 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
|
Chris@82
|
41 {
|
Chris@82
|
42 INT m;
|
Chris@82
|
43 R *x;
|
Chris@82
|
44 x = ri;
|
Chris@82
|
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)) {
|
Chris@82
|
46 V T1, T3, T5, T6, T2, T4, T7, T8;
|
Chris@82
|
47 T1 = LD(&(x[0]), ms, &(x[0]));
|
Chris@82
|
48 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
|
Chris@82
|
49 T3 = BYTWJ(&(W[0]), T2);
|
Chris@82
|
50 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
|
Chris@82
|
51 T5 = BYTWJ(&(W[TWVL * 2]), T4);
|
Chris@82
|
52 T6 = VADD(T3, T5);
|
Chris@82
|
53 ST(&(x[0]), VADD(T1, T6), ms, &(x[0]));
|
Chris@82
|
54 T7 = VFNMS(LDK(KP500000000), T6, T1);
|
Chris@82
|
55 T8 = VMUL(LDK(KP866025403), VSUB(T5, T3));
|
Chris@82
|
56 ST(&(x[WS(rs, 2)]), VFNMSI(T8, T7), ms, &(x[0]));
|
Chris@82
|
57 ST(&(x[WS(rs, 1)]), VFMAI(T8, T7), ms, &(x[WS(rs, 1)]));
|
Chris@82
|
58 }
|
Chris@82
|
59 }
|
Chris@82
|
60 VLEAVE();
|
Chris@82
|
61 }
|
Chris@82
|
62
|
Chris@82
|
63 static const tw_instr twinstr[] = {
|
Chris@82
|
64 VTW(0, 1),
|
Chris@82
|
65 VTW(0, 2),
|
Chris@82
|
66 {TW_NEXT, VL, 0}
|
Chris@82
|
67 };
|
Chris@82
|
68
|
Chris@82
|
69 static const ct_desc desc = { 3, XSIMD_STRING("t1fv_3"), twinstr, &GENUS, {5, 5, 3, 0}, 0, 0, 0 };
|
Chris@82
|
70
|
Chris@82
|
71 void XSIMD(codelet_t1fv_3) (planner *p) {
|
Chris@82
|
72 X(kdft_dit_register) (p, t1fv_3, &desc);
|
Chris@82
|
73 }
|
Chris@82
|
74 #else
|
Chris@82
|
75
|
Chris@82
|
76 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 3 -name t1fv_3 -include dft/simd/t1f.h */
|
Chris@82
|
77
|
Chris@82
|
78 /*
|
Chris@82
|
79 * This function contains 8 FP additions, 6 FP multiplications,
|
Chris@82
|
80 * (or, 7 additions, 5 multiplications, 1 fused multiply/add),
|
Chris@82
|
81 * 12 stack variables, 2 constants, and 6 memory accesses
|
Chris@82
|
82 */
|
Chris@82
|
83 #include "dft/simd/t1f.h"
|
Chris@82
|
84
|
Chris@82
|
85 static void t1fv_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
|
Chris@82
|
86 {
|
Chris@82
|
87 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
|
Chris@82
|
88 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
|
Chris@82
|
89 {
|
Chris@82
|
90 INT m;
|
Chris@82
|
91 R *x;
|
Chris@82
|
92 x = ri;
|
Chris@82
|
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)) {
|
Chris@82
|
94 V T1, T3, T5, T6, T2, T4, T7, T8;
|
Chris@82
|
95 T1 = LD(&(x[0]), ms, &(x[0]));
|
Chris@82
|
96 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
|
Chris@82
|
97 T3 = BYTWJ(&(W[0]), T2);
|
Chris@82
|
98 T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
|
Chris@82
|
99 T5 = BYTWJ(&(W[TWVL * 2]), T4);
|
Chris@82
|
100 T6 = VADD(T3, T5);
|
Chris@82
|
101 ST(&(x[0]), VADD(T1, T6), ms, &(x[0]));
|
Chris@82
|
102 T7 = VFNMS(LDK(KP500000000), T6, T1);
|
Chris@82
|
103 T8 = VBYI(VMUL(LDK(KP866025403), VSUB(T5, T3)));
|
Chris@82
|
104 ST(&(x[WS(rs, 2)]), VSUB(T7, T8), ms, &(x[0]));
|
Chris@82
|
105 ST(&(x[WS(rs, 1)]), VADD(T7, T8), ms, &(x[WS(rs, 1)]));
|
Chris@82
|
106 }
|
Chris@82
|
107 }
|
Chris@82
|
108 VLEAVE();
|
Chris@82
|
109 }
|
Chris@82
|
110
|
Chris@82
|
111 static const tw_instr twinstr[] = {
|
Chris@82
|
112 VTW(0, 1),
|
Chris@82
|
113 VTW(0, 2),
|
Chris@82
|
114 {TW_NEXT, VL, 0}
|
Chris@82
|
115 };
|
Chris@82
|
116
|
Chris@82
|
117 static const ct_desc desc = { 3, XSIMD_STRING("t1fv_3"), twinstr, &GENUS, {7, 5, 1, 0}, 0, 0, 0 };
|
Chris@82
|
118
|
Chris@82
|
119 void XSIMD(codelet_t1fv_3) (planner *p) {
|
Chris@82
|
120 X(kdft_dit_register) (p, t1fv_3, &desc);
|
Chris@82
|
121 }
|
Chris@82
|
122 #endif
|