2
* Copyright (c) 2003, 2007-8 Matteo Frigo
3
* Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology
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.
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.
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21
/* This file was automatically generated --- DO NOT EDIT */
22
/* Generated on Sun Jul 12 06:42:29 EDT 2009 */
24
#include "codelet-dft.h"
28
/* Generated by: ../../../genfft/gen_twiddle_c -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 6 -name t1bv_6 -include t1b.h -sign 1 */
31
* This function contains 23 FP additions, 18 FP multiplications,
32
* (or, 17 additions, 12 multiplications, 6 fused multiply/add),
33
* 27 stack variables, 2 constants, and 12 memory accesses
37
static void t1bv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
39
DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
40
DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
44
for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(rs)) {
45
V T1, T2, Ta, Tc, T5, T7;
46
T1 = LD(&(x[0]), ms, &(x[0]));
47
T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
48
Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
49
Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
50
T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
51
T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
54
T3 = BYTW(&(W[TWVL * 4]), T2);
55
Tb = BYTW(&(W[TWVL * 6]), Ta);
56
Td = BYTW(&(W[0]), Tc);
57
T6 = BYTW(&(W[TWVL * 2]), T5);
58
T8 = BYTW(&(W[TWVL * 8]), T7);
60
V Ti, T4, Tk, Te, Tj, T9;
68
V Tl, Tn, Tf, Th, Tm, Tg;
70
Tn = VMUL(LDK(KP866025403), VSUB(Tj, Tk));
72
Th = VMUL(LDK(KP866025403), VSUB(T9, Te));
73
ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0]));
74
Tm = VFNMS(LDK(KP500000000), Tl, Ti);
75
ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)]));
76
Tg = VFNMS(LDK(KP500000000), Tf, T4);
77
ST(&(x[WS(rs, 4)]), VFMAI(Tn, Tm), ms, &(x[0]));
78
ST(&(x[WS(rs, 2)]), VFNMSI(Tn, Tm), ms, &(x[0]));
79
ST(&(x[WS(rs, 5)]), VFNMSI(Th, Tg), ms, &(x[WS(rs, 1)]));
80
ST(&(x[WS(rs, 1)]), VFMAI(Th, Tg), ms, &(x[WS(rs, 1)]));
87
static const tw_instr twinstr[] = {
96
static const ct_desc desc = { 6, "t1bv_6", twinstr, &GENUS, {17, 12, 6, 0}, 0, 0, 0 };
98
void X(codelet_t1bv_6) (planner *p) {
99
X(kdft_dit_register) (p, t1bv_6, &desc);
103
/* Generated by: ../../../genfft/gen_twiddle_c -simd -compact -variables 4 -pipeline-latency 8 -n 6 -name t1bv_6 -include t1b.h -sign 1 */
106
* This function contains 23 FP additions, 14 FP multiplications,
107
* (or, 21 additions, 12 multiplications, 2 fused multiply/add),
108
* 19 stack variables, 2 constants, and 12 memory accesses
112
static void t1bv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
114
DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
115
DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
119
for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(rs)) {
120
V Tf, Ti, Ta, Tk, T5, Tj, Tc, Te, Td;
121
Tc = LD(&(x[0]), ms, &(x[0]));
122
Td = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
123
Te = BYTW(&(W[TWVL * 4]), Td);
128
T6 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
129
T7 = BYTW(&(W[TWVL * 6]), T6);
130
T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
131
T9 = BYTW(&(W[0]), T8);
137
T1 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
138
T2 = BYTW(&(W[TWVL * 2]), T1);
139
T3 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
140
T4 = BYTW(&(W[TWVL * 8]), T3);
145
V Tb, Tg, Th, Tn, Tl, Tm;
146
Tb = VBYI(VMUL(LDK(KP866025403), VSUB(T5, Ta)));
148
Th = VFNMS(LDK(KP500000000), Tg, Tf);
149
ST(&(x[WS(rs, 1)]), VADD(Tb, Th), ms, &(x[WS(rs, 1)]));
150
ST(&(x[WS(rs, 3)]), VADD(Tf, Tg), ms, &(x[WS(rs, 1)]));
151
ST(&(x[WS(rs, 5)]), VSUB(Th, Tb), ms, &(x[WS(rs, 1)]));
152
Tn = VBYI(VMUL(LDK(KP866025403), VSUB(Tj, Tk)));
154
Tm = VFNMS(LDK(KP500000000), Tl, Ti);
155
ST(&(x[WS(rs, 2)]), VSUB(Tm, Tn), ms, &(x[0]));
156
ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0]));
157
ST(&(x[WS(rs, 4)]), VADD(Tn, Tm), ms, &(x[0]));
162
static const tw_instr twinstr[] = {
171
static const ct_desc desc = { 6, "t1bv_6", twinstr, &GENUS, {21, 12, 2, 0}, 0, 0, 0 };
173
void X(codelet_t1bv_6) (planner *p) {
174
X(kdft_dit_register) (p, t1bv_6, &desc);
176
#endif /* HAVE_FMA */