|
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 4 -name X(spu_t1fv_4) */
|
|
d@0
|
21
|
|
d@0
|
22 /*
|
|
d@0
|
23 * This function contains 11 FP additions, 8 FP multiplications,
|
|
d@0
|
24 * (or, 9 additions, 6 multiplications, 2 fused multiply/add),
|
|
d@0
|
25 * 17 stack variables, 0 constants, and 8 memory accesses
|
|
d@0
|
26 */
|
|
d@0
|
27 #include "fftw-spu.h"
|
|
d@0
|
28
|
|
d@0
|
29 void X(spu_t1fv_4) (R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) {
|
|
d@0
|
30 INT m;
|
|
d@0
|
31 R *x;
|
|
d@0
|
32 x = ri;
|
|
d@0
|
33 for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(rs)) {
|
|
d@0
|
34 V T4, Tc, T9, Td, T1, T3, T2, T6, T8, T5, T7, Tb, Te, Ta, Tf;
|
|
d@0
|
35 T1 = LD(&(x[0]), ms, &(x[0]));
|
|
d@0
|
36 T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
|
|
d@0
|
37 T3 = BYTWJ(&(W[TWVL * 2]), T2);
|
|
d@0
|
38 T4 = VSUB(T1, T3);
|
|
d@0
|
39 Tc = VADD(T1, T3);
|
|
d@0
|
40 T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
|
|
d@0
|
41 T6 = BYTWJ(&(W[0]), T5);
|
|
d@0
|
42 T7 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
|
|
d@0
|
43 T8 = BYTWJ(&(W[TWVL * 4]), T7);
|
|
d@0
|
44 T9 = VSUB(T6, T8);
|
|
d@0
|
45 Td = VADD(T6, T8);
|
|
d@0
|
46 Ta = VFNMSI(T9, T4);
|
|
d@0
|
47 Tb = VFMAI(T9, T4);
|
|
d@0
|
48 ST(&(x[WS(rs, 1)]), Ta, ms, &(x[WS(rs, 1)]));
|
|
d@0
|
49 Tf = VADD(Tc, Td);
|
|
d@0
|
50 Te = VSUB(Tc, Td);
|
|
d@0
|
51 ST(&(x[0]), Tf, ms, &(x[0]));
|
|
d@0
|
52 ST(&(x[WS(rs, 3)]), Tb, ms, &(x[WS(rs, 1)]));
|
|
d@0
|
53 ST(&(x[WS(rs, 2)]), Te, ms, &(x[0]));
|
|
d@0
|
54 }
|
|
d@0
|
55 }
|
