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 Sun Jul 2 15:59:17 EDT 2006 */
24
#include "codelet-rdft.h"
28
/* Generated by: ../../../genfft/gen_hc2hc -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -dit -name hf2_4 -include hf.h */
31
* This function contains 24 FP additions, 16 FP multiplications,
32
* (or, 16 additions, 8 multiplications, 8 fused multiply/add),
33
* 29 stack variables, and 16 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_hc2hc.ml,v 1.16 2006-02-12 23:34:12 athena Exp $
44
static const R *hf2_4(R *rio, R *iio, const R *W, stride ios, INT m, INT dist)
47
for (i = m - 2; i > 0; i = i - 2, rio = rio + dist, iio = iio - dist, W = W + 4, MAKE_VOLATILE_STRIDE(ios)) {
56
E T1, Tx, Tm, Ts, Tf, Th, Td, Tg, Tp, Tw;
58
E Tj, Tl, Tb, T7, Tk, Tr, Ta, T4;
64
Tb = FNMS(T5, T3, Ta);
68
Tx = iio[-WS(ios, 3)];
73
Ts = FNMS(T6, Tj, Tr);
74
Tc = iio[-WS(ios, 1)];
77
Th = iio[-WS(ios, 2)];
82
Tw = FNMS(Tb, T8, Tv);
86
E To, Te, Ti, Tq, Ty, Tn, Tt, Tu;
90
Tq = FNMS(T5, Tf, Tp);
98
iio[-WS(ios, 2)] = Te - Tn;
100
rio[WS(ios, 2)] = Tu - Ty;
101
rio[WS(ios, 1)] = To + Tt;
102
iio[-WS(ios, 3)] = To - Tt;
106
iio[-WS(ios, 1)] = Tz + TA;
107
rio[WS(ios, 3)] = Tz - TA;
112
static const tw_instr twinstr[] = {
118
static const hc2hc_desc desc = { 4, "hf2_4", twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
120
void X(codelet_hf2_4) (planner *p) {
121
X(khc2hc_register) (p, hf2_4, &desc);
125
/* Generated by: ../../../genfft/gen_hc2hc -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -dit -name hf2_4 -include hf.h */
128
* This function contains 24 FP additions, 16 FP multiplications,
129
* (or, 16 additions, 8 multiplications, 8 fused multiply/add),
130
* 21 stack variables, and 16 memory accesses
134
* $Id: algsimp.ml,v 1.9 2006-02-12 23:34:12 athena Exp $
135
* $Id: fft.ml,v 1.4 2006-01-05 03:04:27 stevenj Exp $
136
* $Id: gen_hc2hc.ml,v 1.16 2006-02-12 23:34:12 athena Exp $
141
static const R *hf2_4(R *rio, R *iio, const R *W, stride ios, INT m, INT dist)
144
for (i = m - 2; i > 0; i = i - 2, rio = rio + dist, iio = iio - dist, W = W + 4, MAKE_VOLATILE_STRIDE(ios)) {
145
E T2, T4, T3, T5, T6, T8;
150
T6 = FMA(T2, T3, T4 * T5);
151
T8 = FNMS(T4, T3, T2 * T5);
153
E T1, Th, Tl, Tp, Ta, To, Te, Tk, Tf, Tg;
155
Tf = rio[WS(ios, 3)];
157
Th = FMA(T3, Tf, T5 * Tg);
158
Tl = FNMS(T5, Tf, T3 * Tg);
159
Tp = iio[-WS(ios, 3)];
162
T7 = rio[WS(ios, 2)];
163
T9 = iio[-WS(ios, 1)];
164
Ta = FMA(T6, T7, T8 * T9);
165
To = FNMS(T8, T7, T6 * T9);
166
Tc = rio[WS(ios, 1)];
167
Td = iio[-WS(ios, 2)];
168
Te = FMA(T2, Tc, T4 * Td);
169
Tk = FNMS(T4, Tc, T2 * Td);
175
iio[-WS(ios, 2)] = Tb - Ti;
179
rio[WS(ios, 2)] = Tn - Tq;
186
iio[-WS(ios, 3)] = Tj - Tm;
187
rio[WS(ios, 1)] = Tj + Tm;
190
rio[WS(ios, 3)] = Tr - Ts;
191
iio[-WS(ios, 1)] = Tr + Ts;
198
static const tw_instr twinstr[] = {
204
static const hc2hc_desc desc = { 4, "hf2_4", twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
206
void X(codelet_hf2_4) (planner *p) {
207
X(khc2hc_register) (p, hf2_4, &desc);
209
#endif /* HAVE_FMA */