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 14:46:08 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 9 -name t1fv_9 -include t1f.h */
31
* This function contains 54 FP additions, 54 FP multiplications,
32
* (or, 20 additions, 20 multiplications, 34 fused multiply/add),
33
* 67 stack variables, and 18 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 *t1fv_9(R *ri, R *ii, const R *W, stride ios, INT m, INT dist)
46
DVK(KP939692620, +0.939692620785908384054109277324731469936208134);
47
DVK(KP826351822, +0.826351822333069651148283373230685203999624323);
48
DVK(KP879385241, +0.879385241571816768108218554649462939872416269);
49
DVK(KP984807753, +0.984807753012208059366743024589523013670643252);
50
DVK(KP666666666, +0.666666666666666666666666666666666666666666667);
51
DVK(KP852868531, +0.852868531952443209628250963940074071936020296);
52
DVK(KP907603734, +0.907603734547952313649323976213898122064543220);
53
DVK(KP420276625, +0.420276625461206169731530603237061658838781920);
54
DVK(KP673648177, +0.673648177666930348851716626769314796000375677);
55
DVK(KP898197570, +0.898197570222573798468955502359086394667167570);
56
DVK(KP347296355, +0.347296355333860697703433253538629592000751354);
57
DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
58
DVK(KP439692620, +0.439692620785908384054109277324731469936208134);
59
DVK(KP203604859, +0.203604859554852403062088995281827210665664861);
60
DVK(KP152703644, +0.152703644666139302296566746461370407999248646);
61
DVK(KP586256827, +0.586256827714544512072145703099641959914944179);
62
DVK(KP968908795, +0.968908795874236621082202410917456709164223497);
63
DVK(KP726681596, +0.726681596905677465811651808188092531873167623);
64
DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
68
for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 16), MAKE_VOLATILE_STRIDE(ios)) {
69
V T1, T3, T5, T9, Th, Tb, Td, Tj, Tl, TD, T6;
70
T1 = LD(&(x[0]), dist, &(x[0]));
73
T2 = LD(&(x[WS(ios, 3)]), dist, &(x[WS(ios, 1)]));
74
T4 = LD(&(x[WS(ios, 6)]), dist, &(x[0]));
75
T8 = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)]));
76
Tg = LD(&(x[WS(ios, 2)]), dist, &(x[0]));
79
Ta = LD(&(x[WS(ios, 4)]), dist, &(x[0]));
80
Tc = LD(&(x[WS(ios, 7)]), dist, &(x[WS(ios, 1)]));
81
Ti = LD(&(x[WS(ios, 5)]), dist, &(x[WS(ios, 1)]));
82
Tk = LD(&(x[WS(ios, 8)]), dist, &(x[0]));
83
T3 = BYTWJ(&(W[TWVL * 4]), T2);
84
T5 = BYTWJ(&(W[TWVL * 10]), T4);
85
T9 = BYTWJ(&(W[0]), T8);
86
Th = BYTWJ(&(W[TWVL * 2]), Tg);
87
Tb = BYTWJ(&(W[TWVL * 6]), Ta);
88
Td = BYTWJ(&(W[TWVL * 12]), Tc);
89
Tj = BYTWJ(&(W[TWVL * 8]), Ti);
90
Tl = BYTWJ(&(W[TWVL * 14]), Tk);
96
V Tt, Te, Tu, Tm, Tr, T7;
101
Tr = VFNMS(LDK(KP500000000), T6, T1);
105
Tv = VFNMS(LDK(KP500000000), Te, T9);
107
Ts = VFNMS(LDK(KP500000000), Tm, Th);
110
V TG, TK, Tw, TJ, TF, TA, To, Tq;
111
TG = VFNMS(LDK(KP726681596), Tt, Tv);
112
TK = VFMA(LDK(KP968908795), Tv, Tt);
113
Tw = VFNMS(LDK(KP586256827), Tv, Tu);
114
TJ = VFNMS(LDK(KP152703644), Tu, Ts);
115
TF = VFMA(LDK(KP203604859), Ts, Tu);
116
TA = VFNMS(LDK(KP439692620), Tt, Ts);
118
Tq = VMUL(LDK(KP866025403), VSUB(Tn, Tf));
120
V TQ, TH, TL, TN, TB, Tp, Ty, TI, Tx;
121
Tx = VFNMS(LDK(KP347296355), Tw, Tt);
122
TQ = VFNMS(LDK(KP898197570), TG, TF);
123
TH = VFMA(LDK(KP898197570), TG, TF);
124
TL = VFMA(LDK(KP673648177), TK, TJ);
125
TN = VFNMS(LDK(KP673648177), TK, TJ);
126
TB = VFNMS(LDK(KP420276625), TA, Tu);
127
ST(&(x[0]), VADD(T7, To), dist, &(x[0]));
128
Tp = VFNMS(LDK(KP500000000), To, T7);
129
Ty = VFNMS(LDK(KP907603734), Tx, Ts);
130
TI = VFMA(LDK(KP852868531), TH, Tr);
132
V TO, TR, TM, TC, Tz, TP, TS, TE;
133
TO = VFNMS(LDK(KP500000000), TH, TN);
134
TR = VFMA(LDK(KP666666666), TL, TQ);
135
TM = VMUL(LDK(KP984807753), VFNMS(LDK(KP879385241), TD, TL));
136
TC = VFNMS(LDK(KP826351822), TB, Tv);
137
ST(&(x[WS(ios, 6)]), VFNMSI(Tq, Tp), dist, &(x[0]));
138
ST(&(x[WS(ios, 3)]), VFMAI(Tq, Tp), dist, &(x[WS(ios, 1)]));
139
Tz = VFNMS(LDK(KP939692620), Ty, Tr);
140
TP = VFMA(LDK(KP852868531), TO, Tr);
141
TS = VMUL(LDK(KP866025403), VFMA(LDK(KP852868531), TR, TD));
142
ST(&(x[WS(ios, 8)]), VFMAI(TM, TI), dist, &(x[0]));
143
ST(&(x[WS(ios, 1)]), VFNMSI(TM, TI), dist, &(x[WS(ios, 1)]));
144
TE = VMUL(LDK(KP984807753), VFMA(LDK(KP879385241), TD, TC));
145
ST(&(x[WS(ios, 4)]), VFMAI(TS, TP), dist, &(x[0]));
146
ST(&(x[WS(ios, 5)]), VFNMSI(TS, TP), dist, &(x[WS(ios, 1)]));
147
ST(&(x[WS(ios, 7)]), VFMAI(TE, Tz), dist, &(x[WS(ios, 1)]));
148
ST(&(x[WS(ios, 2)]), VFNMSI(TE, Tz), dist, &(x[0]));
158
static const tw_instr twinstr[] = {
170
static const ct_desc desc = { 9, "t1fv_9", twinstr, &GENUS, {20, 20, 34, 0}, 0, 0, 0 };
172
void X(codelet_t1fv_9) (planner *p) {
173
X(kdft_dit_register) (p, t1fv_9, &desc);
177
/* Generated by: ../../../genfft/gen_twiddle_c -simd -compact -variables 4 -pipeline-latency 8 -n 9 -name t1fv_9 -include t1f.h */
180
* This function contains 54 FP additions, 42 FP multiplications,
181
* (or, 38 additions, 26 multiplications, 16 fused multiply/add),
182
* 38 stack variables, and 18 memory accesses
186
* $Id: algsimp.ml,v 1.9 2006-02-12 23:34:12 athena Exp $
187
* $Id: fft.ml,v 1.4 2006-01-05 03:04:27 stevenj Exp $
188
* $Id: gen_twiddle_c.ml,v 1.14 2006-02-12 23:34:12 athena Exp $
193
static const R *t1fv_9(R *ri, R *ii, const R *W, stride ios, INT m, INT dist)
195
DVK(KP939692620, +0.939692620785908384054109277324731469936208134);
196
DVK(KP296198132, +0.296198132726023843175338011893050938967728390);
197
DVK(KP852868531, +0.852868531952443209628250963940074071936020296);
198
DVK(KP173648177, +0.173648177666930348851716626769314796000375677);
199
DVK(KP556670399, +0.556670399226419366452912952047023132968291906);
200
DVK(KP766044443, +0.766044443118978035202392650555416673935832457);
201
DVK(KP642787609, +0.642787609686539326322643409907263432907559884);
202
DVK(KP663413948, +0.663413948168938396205421319635891297216863310);
203
DVK(KP984807753, +0.984807753012208059366743024589523013670643252);
204
DVK(KP150383733, +0.150383733180435296639271897612501926072238258);
205
DVK(KP342020143, +0.342020143325668733044099614682259580763083368);
206
DVK(KP813797681, +0.813797681349373692844693217248393223289101568);
207
DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
208
DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
212
for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 16), MAKE_VOLATILE_STRIDE(ios)) {
213
V T1, T6, TA, Tt, Tf, Ts, Tw, Tn, Tv;
214
T1 = LD(&(x[0]), dist, &(x[0]));
217
T2 = LD(&(x[WS(ios, 3)]), dist, &(x[WS(ios, 1)]));
218
T3 = BYTWJ(&(W[TWVL * 4]), T2);
219
T4 = LD(&(x[WS(ios, 6)]), dist, &(x[0]));
220
T5 = BYTWJ(&(W[TWVL * 10]), T4);
222
TA = VMUL(LDK(KP866025403), VSUB(T5, T3));
225
V T9, Td, Tb, T8, Tc, Ta, Te;
226
T8 = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)]));
227
T9 = BYTWJ(&(W[0]), T8);
228
Tc = LD(&(x[WS(ios, 7)]), dist, &(x[WS(ios, 1)]));
229
Td = BYTWJ(&(W[TWVL * 12]), Tc);
230
Ta = LD(&(x[WS(ios, 4)]), dist, &(x[0]));
231
Tb = BYTWJ(&(W[TWVL * 6]), Ta);
235
Ts = VFNMS(LDK(KP500000000), Te, T9);
238
V Th, Tl, Tj, Tg, Tk, Ti, Tm;
239
Tg = LD(&(x[WS(ios, 2)]), dist, &(x[0]));
240
Th = BYTWJ(&(W[TWVL * 2]), Tg);
241
Tk = LD(&(x[WS(ios, 8)]), dist, &(x[0]));
242
Tl = BYTWJ(&(W[TWVL * 14]), Tk);
243
Ti = LD(&(x[WS(ios, 5)]), dist, &(x[WS(ios, 1)]));
244
Tj = BYTWJ(&(W[TWVL * 8]), Ti);
248
Tv = VFNMS(LDK(KP500000000), Tm, Th);
252
Tq = VBYI(VMUL(LDK(KP866025403), VSUB(Tn, Tf)));
255
Tp = VFNMS(LDK(KP500000000), To, T7);
256
ST(&(x[0]), VADD(T7, To), dist, &(x[0]));
257
ST(&(x[WS(ios, 3)]), VADD(Tp, Tq), dist, &(x[WS(ios, 1)]));
258
ST(&(x[WS(ios, 6)]), VSUB(Tp, Tq), dist, &(x[0]));
261
V TI, TB, TC, TD, Tu, Tx, Ty, Tr, TH;
262
TI = VBYI(VSUB(VFNMS(LDK(KP342020143), Tv, VFNMS(LDK(KP150383733), Tt, VFNMS(LDK(KP984807753), Ts, VMUL(LDK(KP813797681), Tw)))), TA));
263
TB = VFNMS(LDK(KP642787609), Ts, VMUL(LDK(KP663413948), Tt));
264
TC = VFNMS(LDK(KP984807753), Tv, VMUL(LDK(KP150383733), Tw));
266
Tu = VFMA(LDK(KP766044443), Ts, VMUL(LDK(KP556670399), Tt));
267
Tx = VFMA(LDK(KP173648177), Tv, VMUL(LDK(KP852868531), Tw));
269
Tr = VFNMS(LDK(KP500000000), T6, T1);
270
TH = VFMA(LDK(KP173648177), Ts, VFNMS(LDK(KP296198132), Tw, VFNMS(LDK(KP939692620), Tv, VFNMS(LDK(KP852868531), Tt, Tr))));
271
ST(&(x[WS(ios, 7)]), VSUB(TH, TI), dist, &(x[WS(ios, 1)]));
272
ST(&(x[WS(ios, 2)]), VADD(TH, TI), dist, &(x[0]));
276
TE = VBYI(VADD(TA, TD));
277
ST(&(x[WS(ios, 8)]), VSUB(Tz, TE), dist, &(x[0]));
278
ST(&(x[WS(ios, 1)]), VADD(TE, Tz), dist, &(x[WS(ios, 1)]));
279
TF = VFMA(LDK(KP866025403), VSUB(TB, TC), VFNMS(LDK(KP500000000), Ty, Tr));
280
TG = VBYI(VADD(TA, VFNMS(LDK(KP500000000), TD, VMUL(LDK(KP866025403), VSUB(Tx, Tu)))));
281
ST(&(x[WS(ios, 5)]), VSUB(TF, TG), dist, &(x[WS(ios, 1)]));
282
ST(&(x[WS(ios, 4)]), VADD(TF, TG), dist, &(x[0]));
289
static const tw_instr twinstr[] = {
301
static const ct_desc desc = { 9, "t1fv_9", twinstr, &GENUS, {38, 26, 16, 0}, 0, 0, 0 };
303
void X(codelet_t1fv_9) (planner *p) {
304
X(kdft_dit_register) (p, t1fv_9, &desc);
306
#endif /* HAVE_FMA */