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 06:40:52 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 7 -name t1fv_7 -include t1f.h */
31
* This function contains 36 FP additions, 36 FP multiplications,
32
* (or, 15 additions, 15 multiplications, 21 fused multiply/add),
33
* 42 stack variables, 6 constants, and 14 memory accesses
37
static void t1fv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
39
DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
40
DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
41
DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
42
DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
43
DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
44
DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
48
for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(rs)) {
49
V T1, T2, T4, Te, Tc, T9, T7;
50
T1 = LD(&(x[0]), ms, &(x[0]));
51
T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
52
T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
53
Te = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
54
Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
55
T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
56
T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
58
V T3, T5, Tf, Td, Ta, T8;
59
T3 = BYTWJ(&(W[0]), T2);
60
T5 = BYTWJ(&(W[TWVL * 10]), T4);
61
Tf = BYTWJ(&(W[TWVL * 6]), Te);
62
Td = BYTWJ(&(W[TWVL * 4]), Tc);
63
Ta = BYTWJ(&(W[TWVL * 8]), T9);
64
T8 = BYTWJ(&(W[TWVL * 2]), T7);
66
V T6, Tk, Tg, Tl, Tb, Tm;
74
V Th, Ts, Tp, Tu, Tn, Tx, Ti, Tt;
75
Th = VFNMS(LDK(KP356895867), T6, Tg);
76
Ts = VFMA(LDK(KP554958132), Tl, Tk);
77
ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0]));
78
Tp = VFNMS(LDK(KP356895867), Tb, T6);
79
Tu = VFNMS(LDK(KP356895867), Tg, Tb);
80
Tn = VFMA(LDK(KP554958132), Tm, Tl);
81
Tx = VFNMS(LDK(KP554958132), Tk, Tm);
82
Ti = VFNMS(LDK(KP692021471), Th, Tb);
83
Tt = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Ts, Tm));
85
V Tq, Tv, To, Ty, Tj, Tr, Tw;
86
Tq = VFNMS(LDK(KP692021471), Tp, Tg);
87
Tv = VFNMS(LDK(KP692021471), Tu, T6);
88
To = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tn, Tk));
89
Ty = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tx, Tl));
90
Tj = VFNMS(LDK(KP900968867), Ti, T1);
91
Tr = VFNMS(LDK(KP900968867), Tq, T1);
92
Tw = VFNMS(LDK(KP900968867), Tv, T1);
93
ST(&(x[WS(rs, 2)]), VFMAI(To, Tj), ms, &(x[0]));
94
ST(&(x[WS(rs, 5)]), VFNMSI(To, Tj), ms, &(x[WS(rs, 1)]));
95
ST(&(x[WS(rs, 1)]), VFMAI(Tt, Tr), ms, &(x[WS(rs, 1)]));
96
ST(&(x[WS(rs, 6)]), VFNMSI(Tt, Tr), ms, &(x[0]));
97
ST(&(x[WS(rs, 3)]), VFMAI(Ty, Tw), ms, &(x[WS(rs, 1)]));
98
ST(&(x[WS(rs, 4)]), VFNMSI(Ty, Tw), ms, &(x[0]));
106
static const tw_instr twinstr[] = {
116
static const ct_desc desc = { 7, "t1fv_7", twinstr, &GENUS, {15, 15, 21, 0}, 0, 0, 0 };
118
void X(codelet_t1fv_7) (planner *p) {
119
X(kdft_dit_register) (p, t1fv_7, &desc);
123
/* Generated by: ../../../genfft/gen_twiddle_c -simd -compact -variables 4 -pipeline-latency 8 -n 7 -name t1fv_7 -include t1f.h */
126
* This function contains 36 FP additions, 30 FP multiplications,
127
* (or, 24 additions, 18 multiplications, 12 fused multiply/add),
128
* 21 stack variables, 6 constants, and 14 memory accesses
132
static void t1fv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
134
DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
135
DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
136
DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
137
DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
138
DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
139
DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
143
for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(rs)) {
144
V T1, Tg, Tj, T6, Ti, Tb, Tk, Tp, To;
145
T1 = LD(&(x[0]), ms, &(x[0]));
148
Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
149
Td = BYTWJ(&(W[TWVL * 4]), Tc);
150
Te = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
151
Tf = BYTWJ(&(W[TWVL * 6]), Te);
157
T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
158
T3 = BYTWJ(&(W[0]), T2);
159
T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
160
T5 = BYTWJ(&(W[TWVL * 10]), T4);
166
T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
167
T8 = BYTWJ(&(W[TWVL * 2]), T7);
168
T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
169
Ta = BYTWJ(&(W[TWVL * 8]), T9);
173
ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0]));
174
Tp = VBYI(VFMA(LDK(KP433883739), Ti, VFNMS(LDK(KP781831482), Tk, VMUL(LDK(KP974927912), Tj))));
175
To = VFMA(LDK(KP623489801), Tb, VFNMS(LDK(KP222520933), Tg, VFNMS(LDK(KP900968867), T6, T1)));
176
ST(&(x[WS(rs, 4)]), VSUB(To, Tp), ms, &(x[0]));
177
ST(&(x[WS(rs, 3)]), VADD(To, Tp), ms, &(x[WS(rs, 1)]));
180
Tl = VBYI(VFNMS(LDK(KP781831482), Tj, VFNMS(LDK(KP433883739), Tk, VMUL(LDK(KP974927912), Ti))));
181
Th = VFMA(LDK(KP623489801), Tg, VFNMS(LDK(KP900968867), Tb, VFNMS(LDK(KP222520933), T6, T1)));
182
ST(&(x[WS(rs, 5)]), VSUB(Th, Tl), ms, &(x[WS(rs, 1)]));
183
ST(&(x[WS(rs, 2)]), VADD(Th, Tl), ms, &(x[0]));
184
Tn = VBYI(VFMA(LDK(KP781831482), Ti, VFMA(LDK(KP974927912), Tk, VMUL(LDK(KP433883739), Tj))));
185
Tm = VFMA(LDK(KP623489801), T6, VFNMS(LDK(KP900968867), Tg, VFNMS(LDK(KP222520933), Tb, T1)));
186
ST(&(x[WS(rs, 6)]), VSUB(Tm, Tn), ms, &(x[0]));
187
ST(&(x[WS(rs, 1)]), VADD(Tm, Tn), ms, &(x[WS(rs, 1)]));
192
static const tw_instr twinstr[] = {
202
static const ct_desc desc = { 7, "t1fv_7", twinstr, &GENUS, {24, 18, 12, 0}, 0, 0, 0 };
204
void X(codelet_t1fv_7) (planner *p) {
205
X(kdft_dit_register) (p, t1fv_7, &desc);
207
#endif /* HAVE_FMA */