2
* Copyright (c) 2003, 2006 Matteo Frigo
3
* Copyright (c) 2003, 2006 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 Sat Jul 1 14:59:07 EDT 2006 */
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 2 -name t1bv_2 -include t1b.h -sign 1 */
31
* This function contains 3 FP additions, 2 FP multiplications,
32
* (or, 3 additions, 2 multiplications, 0 fused multiply/add),
33
* 5 stack variables, and 4 memory accesses
37
* $Id: algsimp.ml,v 1.9 2006-02-12 23:34:12 athena Exp $
38
* $Id: fft.ml,v 1.4 2006-01-05 03:04:27 stevenj Exp $
39
* $Id: gen_twiddle_c.ml,v 1.14 2006-02-12 23:34:12 athena Exp $
44
static const R *t1bv_2(R *ri, R *ii, const R *W, stride ios, INT m, INT dist)
49
for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(ios)) {
51
T1 = LD(&(x[0]), dist, &(x[0]));
52
T2 = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)]));
53
T3 = BYTW(&(W[0]), T2);
54
ST(&(x[0]), VADD(T1, T3), dist, &(x[0]));
55
ST(&(x[WS(ios, 1)]), VSUB(T1, T3), dist, &(x[WS(ios, 1)]));
60
static const tw_instr twinstr[] = {
65
static const ct_desc desc = { 2, "t1bv_2", twinstr, &GENUS, {3, 2, 0, 0}, 0, 0, 0 };
67
void X(codelet_t1bv_2) (planner *p) {
68
X(kdft_dit_register) (p, t1bv_2, &desc);
72
/* Generated by: ../../../genfft/gen_twiddle_c -simd -compact -variables 4 -pipeline-latency 8 -n 2 -name t1bv_2 -include t1b.h -sign 1 */
75
* This function contains 3 FP additions, 2 FP multiplications,
76
* (or, 3 additions, 2 multiplications, 0 fused multiply/add),
77
* 5 stack variables, and 4 memory accesses
81
* $Id: algsimp.ml,v 1.9 2006-02-12 23:34:12 athena Exp $
82
* $Id: fft.ml,v 1.4 2006-01-05 03:04:27 stevenj Exp $
83
* $Id: gen_twiddle_c.ml,v 1.14 2006-02-12 23:34:12 athena Exp $
88
static const R *t1bv_2(R *ri, R *ii, const R *W, stride ios, INT m, INT dist)
93
for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(ios)) {
95
T1 = LD(&(x[0]), dist, &(x[0]));
96
T2 = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)]));
97
T3 = BYTW(&(W[0]), T2);
98
ST(&(x[WS(ios, 1)]), VSUB(T1, T3), dist, &(x[WS(ios, 1)]));
99
ST(&(x[0]), VADD(T1, T3), dist, &(x[0]));
104
static const tw_instr twinstr[] = {
109
static const ct_desc desc = { 2, "t1bv_2", twinstr, &GENUS, {3, 2, 0, 0}, 0, 0, 0 };
111
void X(codelet_t1bv_2) (planner *p) {
112
X(kdft_dit_register) (p, t1bv_2, &desc);
114
#endif /* HAVE_FMA */