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 06:40:51 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 4 -name t1fv_4 -include t1f.h */
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
37
static void t1fv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
42
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(rs)) {
43
V T1, T7, T2, T5, T8, T3, T6;
44
T1 = LD(&(x[0]), ms, &(x[0]));
45
T7 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
46
T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
47
T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
48
T8 = BYTWJ(&(W[TWVL * 4]), T7);
49
T3 = BYTWJ(&(W[TWVL * 2]), T2);
50
T6 = BYTWJ(&(W[0]), T5);
57
ST(&(x[0]), VADD(Ta, Tb), ms, &(x[0]));
58
ST(&(x[WS(rs, 2)]), VSUB(Ta, Tb), ms, &(x[0]));
59
ST(&(x[WS(rs, 3)]), VFMAI(T9, T4), ms, &(x[WS(rs, 1)]));
60
ST(&(x[WS(rs, 1)]), VFNMSI(T9, T4), ms, &(x[WS(rs, 1)]));
65
static const tw_instr twinstr[] = {
72
static const ct_desc desc = { 4, "t1fv_4", twinstr, &GENUS, {9, 6, 2, 0}, 0, 0, 0 };
74
void X(codelet_t1fv_4) (planner *p) {
75
X(kdft_dit_register) (p, t1fv_4, &desc);
79
/* Generated by: ../../../genfft/gen_twiddle_c -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name t1fv_4 -include t1f.h */
82
* This function contains 11 FP additions, 6 FP multiplications,
83
* (or, 11 additions, 6 multiplications, 0 fused multiply/add),
84
* 13 stack variables, 0 constants, and 8 memory accesses
88
static void t1fv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
93
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(rs)) {
94
V T1, T8, T3, T6, T7, T2, T5;
95
T1 = LD(&(x[0]), ms, &(x[0]));
96
T7 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
97
T8 = BYTWJ(&(W[TWVL * 4]), T7);
98
T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
99
T3 = BYTWJ(&(W[TWVL * 2]), T2);
100
T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
101
T6 = BYTWJ(&(W[0]), T5);
105
T9 = VBYI(VSUB(T6, T8));
106
ST(&(x[WS(rs, 1)]), VSUB(T4, T9), ms, &(x[WS(rs, 1)]));
107
ST(&(x[WS(rs, 3)]), VADD(T4, T9), ms, &(x[WS(rs, 1)]));
110
ST(&(x[WS(rs, 2)]), VSUB(Ta, Tb), ms, &(x[0]));
111
ST(&(x[0]), VADD(Ta, Tb), ms, &(x[0]));
116
static const tw_instr twinstr[] = {
123
static const ct_desc desc = { 4, "t1fv_4", twinstr, &GENUS, {11, 6, 0, 0}, 0, 0, 0 };
125
void X(codelet_t1fv_4) (planner *p) {
126
X(kdft_dit_register) (p, t1fv_4, &desc);
128
#endif /* HAVE_FMA */