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 06:42:02 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 10 -name t2fv_10 -include t2f.h */
31
* This function contains 51 FP additions, 40 FP multiplications,
32
* (or, 33 additions, 22 multiplications, 18 fused multiply/add),
33
* 43 stack variables, 4 constants, and 20 memory accesses
37
static void t2fv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
39
DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40
DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
41
DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
42
DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
46
for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(rs)) {
47
V Td, TA, T4, Ta, Tk, TE, Tp, TF, TB, T9, T1, T2, Tb;
48
T1 = LD(&(x[0]), ms, &(x[0]));
49
T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
52
Tg = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
53
Tn = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
54
Ti = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
55
Tl = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
58
T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
59
Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
61
V T3, Th, To, Tj, Tm, T7;
62
T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
63
T3 = BYTWJ(&(W[TWVL * 8]), T2);
64
Th = BYTWJ(&(W[TWVL * 6]), Tg);
65
To = BYTWJ(&(W[0]), Tn);
66
Tj = BYTWJ(&(W[TWVL * 16]), Ti);
67
Tm = BYTWJ(&(W[TWVL * 10]), Tl);
68
T6 = BYTWJ(&(W[TWVL * 2]), T5);
69
Td = BYTWJ(&(W[TWVL * 4]), Tc);
70
T8 = BYTWJ(&(W[TWVL * 12]), T7);
73
Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
83
Tb = BYTWJ(&(W[TWVL * 14]), Ta);
85
V TL, TG, Tw, Tq, TC, Te;
99
V TP, TN, TH, TJ, Tz, Tx, Tr, Tt, TI, Ts;
100
TP = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TL, TM));
101
TN = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TM, TL));
104
Tz = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tv, Tw));
105
Tx = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tw, Tv));
108
ST(&(x[0]), VADD(TA, TH), ms, &(x[0]));
109
TI = VFNMS(LDK(KP250000000), TH, TA);
110
ST(&(x[WS(rs, 5)]), VADD(T4, Tr), ms, &(x[WS(rs, 1)]));
111
Ts = VFNMS(LDK(KP250000000), Tr, T4);
114
TK = VFNMS(LDK(KP559016994), TJ, TI);
115
TO = VFMA(LDK(KP559016994), TJ, TI);
116
Tu = VFMA(LDK(KP559016994), Tt, Ts);
117
Ty = VFNMS(LDK(KP559016994), Tt, Ts);
118
ST(&(x[WS(rs, 8)]), VFNMSI(TN, TK), ms, &(x[0]));
119
ST(&(x[WS(rs, 2)]), VFMAI(TN, TK), ms, &(x[0]));
120
ST(&(x[WS(rs, 6)]), VFNMSI(TP, TO), ms, &(x[0]));
121
ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0]));
122
ST(&(x[WS(rs, 9)]), VFMAI(Tx, Tu), ms, &(x[WS(rs, 1)]));
123
ST(&(x[WS(rs, 1)]), VFNMSI(Tx, Tu), ms, &(x[WS(rs, 1)]));
124
ST(&(x[WS(rs, 7)]), VFMAI(Tz, Ty), ms, &(x[WS(rs, 1)]));
125
ST(&(x[WS(rs, 3)]), VFNMSI(Tz, Ty), ms, &(x[WS(rs, 1)]));
133
static const tw_instr twinstr[] = {
146
static const ct_desc desc = { 10, "t2fv_10", twinstr, &GENUS, {33, 22, 18, 0}, 0, 0, 0 };
148
void X(codelet_t2fv_10) (planner *p) {
149
X(kdft_dit_register) (p, t2fv_10, &desc);
153
/* Generated by: ../../../genfft/gen_twiddle_c -simd -compact -variables 4 -pipeline-latency 8 -n 10 -name t2fv_10 -include t2f.h */
156
* This function contains 51 FP additions, 30 FP multiplications,
157
* (or, 45 additions, 24 multiplications, 6 fused multiply/add),
158
* 32 stack variables, 4 constants, and 20 memory accesses
162
static void t2fv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
164
DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
165
DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
166
DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
167
DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
171
for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(rs)) {
172
V Tr, TH, Tg, Tl, Tm, TA, TB, TJ, T5, Ta, Tb, TD, TE, TI, To;
174
To = LD(&(x[0]), ms, &(x[0]));
175
Tp = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
176
Tq = BYTWJ(&(W[TWVL * 8]), Tp);
183
Tc = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
184
Td = BYTWJ(&(W[TWVL * 6]), Tc);
185
Tj = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
186
Tk = BYTWJ(&(W[0]), Tj);
187
Te = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
188
Tf = BYTWJ(&(W[TWVL * 16]), Te);
189
Th = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
190
Ti = BYTWJ(&(W[TWVL * 10]), Th);
203
T1 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
204
T2 = BYTWJ(&(W[TWVL * 2]), T1);
205
T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
206
T9 = BYTWJ(&(W[TWVL * 4]), T8);
207
T3 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
208
T4 = BYTWJ(&(W[TWVL * 12]), T3);
209
T6 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
210
T7 = BYTWJ(&(W[TWVL * 14]), T6);
220
V Tn, Ts, Tt, Tx, Tz, Tv, Tw, Ty, Tu;
221
Tn = VMUL(LDK(KP559016994), VSUB(Tb, Tm));
223
Tt = VFNMS(LDK(KP250000000), Ts, Tr);
226
Tx = VBYI(VFMA(LDK(KP951056516), Tv, VMUL(LDK(KP587785252), Tw)));
227
Tz = VBYI(VFNMS(LDK(KP587785252), Tv, VMUL(LDK(KP951056516), Tw)));
228
ST(&(x[WS(rs, 5)]), VADD(Tr, Ts), ms, &(x[WS(rs, 1)]));
230
ST(&(x[WS(rs, 3)]), VSUB(Ty, Tz), ms, &(x[WS(rs, 1)]));
231
ST(&(x[WS(rs, 7)]), VADD(Tz, Ty), ms, &(x[WS(rs, 1)]));
233
ST(&(x[WS(rs, 1)]), VSUB(Tu, Tx), ms, &(x[WS(rs, 1)]));
234
ST(&(x[WS(rs, 9)]), VADD(Tx, Tu), ms, &(x[WS(rs, 1)]));
237
V TM, TK, TL, TG, TO, TC, TF, TP, TN;
238
TM = VMUL(LDK(KP559016994), VSUB(TI, TJ));
240
TL = VFNMS(LDK(KP250000000), TK, TH);
243
TG = VBYI(VFNMS(LDK(KP587785252), TF, VMUL(LDK(KP951056516), TC)));
244
TO = VBYI(VFMA(LDK(KP951056516), TF, VMUL(LDK(KP587785252), TC)));
245
ST(&(x[0]), VADD(TH, TK), ms, &(x[0]));
247
ST(&(x[WS(rs, 4)]), VADD(TO, TP), ms, &(x[0]));
248
ST(&(x[WS(rs, 6)]), VSUB(TP, TO), ms, &(x[0]));
250
ST(&(x[WS(rs, 2)]), VADD(TG, TN), ms, &(x[0]));
251
ST(&(x[WS(rs, 8)]), VSUB(TN, TG), ms, &(x[0]));
256
static const tw_instr twinstr[] = {
269
static const ct_desc desc = { 10, "t2fv_10", twinstr, &GENUS, {45, 24, 6, 0}, 0, 0, 0 };
271
void X(codelet_t2fv_10) (planner *p) {
272
X(kdft_dit_register) (p, t2fv_10, &desc);
274
#endif /* HAVE_FMA */