|
d@0
|
1 /*
|
|
d@0
|
2 * Copyright (c) 2003, 2007-8 Matteo Frigo
|
|
d@0
|
3 * Copyright (c) 2003, 2007-8 Massachusetts Institute of Technology
|
|
d@0
|
4 *
|
|
d@0
|
5 * This program is free software; you can redistribute it and/or modify
|
|
d@0
|
6 * it under the terms of the GNU General Public License as published by
|
|
d@0
|
7 * the Free Software Foundation; either version 2 of the License, or
|
|
d@0
|
8 * (at your option) any later version.
|
|
d@0
|
9 *
|
|
d@0
|
10 * This program is distributed in the hope that it will be useful,
|
|
d@0
|
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
d@0
|
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
d@0
|
13 * GNU General Public License for more details.
|
|
d@0
|
14 *
|
|
d@0
|
15 * You should have received a copy of the GNU General Public License
|
|
d@0
|
16 * along with this program; if not, write to the Free Software
|
|
d@0
|
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
d@0
|
18 *
|
|
d@0
|
19 */
|
|
d@0
|
20 /* Generated by: ../../genfft/gen_twiddle_c -standalone -fma -reorder-insns -simd -compact -variables 100000 -include fftw-spu.h -trivial-stores -n 5 -name X(spu_t1fv_5) */
|
|
d@0
|
21
|
|
d@0
|
22 /*
|
|
d@0
|
23 * This function contains 20 FP additions, 19 FP multiplications,
|
|
d@0
|
24 * (or, 11 additions, 10 multiplications, 9 fused multiply/add),
|
|
d@0
|
25 * 31 stack variables, 4 constants, and 10 memory accesses
|
|
d@0
|
26 */
|
|
d@0
|
27 #include "fftw-spu.h"
|
|
d@0
|
28
|
|
d@0
|
29 void X(spu_t1fv_5) (R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) {
|
|
d@0
|
30 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
|
|
d@0
|
31 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
|
|
d@0
|
32 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
|
|
d@0
|
33 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
|
|
d@0
|
34 INT m;
|
|
d@0
|
35 R *x;
|
|
d@0
|
36 x = ri;
|
|
d@0
|
37 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(rs)) {
|
|
d@0
|
38 V T1, Tg, Th, Tc, Te, T3, Ta, T5, T8, T2, T9, T4, T7, T6, Tb;
|
|
d@0
|
39 V Td, Tp, Ti, Tm, Tl, Tf, Tk, Tn, Tj, To;
|
|
d@0
|
40 T1 = LD(&(x[0]), ms, &(x[0]));
|
|
d@0
|
41 T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
|
|
d@0
|
42 T3 = BYTWJ(&(W[0]), T2);
|
|
d@0
|
43 T9 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
|
|
d@0
|
44 Ta = BYTWJ(&(W[TWVL * 4]), T9);
|
|
d@0
|
45 T4 = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
|
|
d@0
|
46 T5 = BYTWJ(&(W[TWVL * 6]), T4);
|
|
d@0
|
47 T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
|
|
d@0
|
48 T8 = BYTWJ(&(W[TWVL * 2]), T7);
|
|
d@0
|
49 Tg = VSUB(T3, T5);
|
|
d@0
|
50 T6 = VADD(T3, T5);
|
|
d@0
|
51 Tb = VADD(T8, Ta);
|
|
d@0
|
52 Th = VSUB(T8, Ta);
|
|
d@0
|
53 Tc = VADD(T6, Tb);
|
|
d@0
|
54 Te = VSUB(T6, Tb);
|
|
d@0
|
55 Tp = VADD(T1, Tc);
|
|
d@0
|
56 Td = VFNMS(LDK(KP250000000), Tc, T1);
|
|
d@0
|
57 ST(&(x[0]), Tp, ms, &(x[0]));
|
|
d@0
|
58 Ti = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Th, Tg));
|
|
d@0
|
59 Tm = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tg, Th));
|
|
d@0
|
60 Tl = VFNMS(LDK(KP559016994), Te, Td);
|
|
d@0
|
61 Tf = VFMA(LDK(KP559016994), Te, Td);
|
|
d@0
|
62 Tk = VFMAI(Ti, Tf);
|
|
d@0
|
63 Tj = VFNMSI(Ti, Tf);
|
|
d@0
|
64 ST(&(x[WS(rs, 1)]), Tj, ms, &(x[WS(rs, 1)]));
|
|
d@0
|
65 Tn = VFMAI(Tm, Tl);
|
|
d@0
|
66 To = VFNMSI(Tm, Tl);
|
|
d@0
|
67 ST(&(x[WS(rs, 3)]), To, ms, &(x[WS(rs, 1)]));
|
|
d@0
|
68 ST(&(x[WS(rs, 4)]), Tk, ms, &(x[0]));
|
|
d@0
|
69 ST(&(x[WS(rs, 2)]), Tn, ms, &(x[0]));
|
|
d@0
|
70 }
|
|
d@0
|
71 }
|
