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_notw_c -standalone -fma -reorder-insns -simd -compact -variables 100000 -with-ostride 2 -include fftw-spu.h -n 15 -name X(spu_n2fv_15) */
|
d@0
|
21
|
d@0
|
22 /*
|
d@0
|
23 * This function contains 78 FP additions, 49 FP multiplications,
|
d@0
|
24 * (or, 36 additions, 7 multiplications, 42 fused multiply/add),
|
d@0
|
25 * 89 stack variables, 8 constants, and 30 memory accesses
|
d@0
|
26 */
|
d@0
|
27 #include "fftw-spu.h"
|
d@0
|
28
|
d@0
|
29 void X(spu_n2fv_15) (const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) {
|
d@0
|
30 DVK(KP910592997, +0.910592997310029334643087372129977886038870291);
|
d@0
|
31 DVK(KP823639103, +0.823639103546331925877420039278190003029660514);
|
d@0
|
32 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
|
d@0
|
33 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
|
d@0
|
34 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
|
d@0
|
35 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
|
d@0
|
36 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
|
d@0
|
37 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
|
d@0
|
38 INT i;
|
d@0
|
39 const R *xi;
|
d@0
|
40 R *xo;
|
d@0
|
41 xi = ri;
|
d@0
|
42 xo = ro;
|
d@0
|
43 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(os)) {
|
d@0
|
44 V TX, TB, T5, TO, TU, TV, TR, Tg, Tx, Tw, Tr, TI, TK, T12, T10;
|
d@0
|
45 V T1, T4, T2, T3, TM, TC, Ta, TQ, TG, Tq, TN, TD, Tf, TP, TF;
|
d@0
|
46 V Tl, T6, T9, T7, T8, Tm, Tp, Tn, To, Tb, Te, Tc, Td, Th, Tk;
|
d@0
|
47 V Ti, Tj, TY, TE, TH, TZ, TJ, T11, T1f, T1g, Ts, Tu, Tt, Ty, TA;
|
d@0
|
48 V Tv, Tz, T18, TS, TW, T1a, T13, T1b, TL, T17, T15, T16, T19, T1c, TT;
|
d@0
|
49 V T14, T1d, T1e;
|
d@0
|
50 T1 = LD(&(xi[0]), ivs, &(xi[0]));
|
d@0
|
51 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
|
d@0
|
52 T3 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
|
d@0
|
53 T4 = VADD(T2, T3);
|
d@0
|
54 TX = VSUB(T3, T2);
|
d@0
|
55 TB = VFNMS(LDK(KP500000000), T4, T1);
|
d@0
|
56 T5 = VADD(T1, T4);
|
d@0
|
57 T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
|
d@0
|
58 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
|
d@0
|
59 T8 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
|
d@0
|
60 TM = VSUB(T8, T7);
|
d@0
|
61 T9 = VADD(T7, T8);
|
d@0
|
62 TC = VFNMS(LDK(KP500000000), T9, T6);
|
d@0
|
63 Ta = VADD(T6, T9);
|
d@0
|
64 Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
|
d@0
|
65 Tn = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
|
d@0
|
66 To = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
|
d@0
|
67 TQ = VSUB(To, Tn);
|
d@0
|
68 Tp = VADD(Tn, To);
|
d@0
|
69 TG = VFNMS(LDK(KP500000000), Tp, Tm);
|
d@0
|
70 Tq = VADD(Tm, Tp);
|
d@0
|
71 Tb = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
|
d@0
|
72 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
|
d@0
|
73 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
|
d@0
|
74 TN = VSUB(Td, Tc);
|
d@0
|
75 Te = VADD(Tc, Td);
|
d@0
|
76 TD = VFNMS(LDK(KP500000000), Te, Tb);
|
d@0
|
77 Tf = VADD(Tb, Te);
|
d@0
|
78 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
|
d@0
|
79 Ti = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
|
d@0
|
80 Tj = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
|
d@0
|
81 TP = VSUB(Tj, Ti);
|
d@0
|
82 Tk = VADD(Ti, Tj);
|
d@0
|
83 TF = VFNMS(LDK(KP500000000), Tk, Th);
|
d@0
|
84 Tl = VADD(Th, Tk);
|
d@0
|
85 TO = VSUB(TM, TN);
|
d@0
|
86 TY = VADD(TM, TN);
|
d@0
|
87 TE = VADD(TC, TD);
|
d@0
|
88 TU = VSUB(TC, TD);
|
d@0
|
89 TV = VSUB(TF, TG);
|
d@0
|
90 TH = VADD(TF, TG);
|
d@0
|
91 TZ = VADD(TP, TQ);
|
d@0
|
92 TR = VSUB(TP, TQ);
|
d@0
|
93 Tg = VADD(Ta, Tf);
|
d@0
|
94 Tx = VSUB(Ta, Tf);
|
d@0
|
95 Tw = VSUB(Tl, Tq);
|
d@0
|
96 Tr = VADD(Tl, Tq);
|
d@0
|
97 TI = VADD(TE, TH);
|
d@0
|
98 TK = VSUB(TE, TH);
|
d@0
|
99 T12 = VSUB(TY, TZ);
|
d@0
|
100 T10 = VADD(TY, TZ);
|
d@0
|
101 TJ = VFNMS(LDK(KP250000000), TI, TB);
|
d@0
|
102 T1f = VADD(TB, TI);
|
d@0
|
103 T1g = VMUL(LDK(KP866025403), VADD(TX, T10));
|
d@0
|
104 T11 = VFNMS(LDK(KP250000000), T10, TX);
|
d@0
|
105 ST(&(xo[10]), VFNMSI(T1g, T1f), ovs, &(xo[2]));
|
d@0
|
106 ST(&(xo[20]), VFMAI(T1g, T1f), ovs, &(xo[0]));
|
d@0
|
107 Ts = VADD(Tg, Tr);
|
d@0
|
108 Tu = VSUB(Tg, Tr);
|
d@0
|
109 Tt = VFNMS(LDK(KP250000000), Ts, T5);
|
d@0
|
110 Ty = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tx, Tw));
|
d@0
|
111 TA = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tw, Tx));
|
d@0
|
112 ST(&(xo[0]), VADD(T5, Ts), ovs, &(xo[0]));
|
d@0
|
113 Tv = VFNMS(LDK(KP559016994), Tu, Tt);
|
d@0
|
114 Tz = VFMA(LDK(KP559016994), Tu, Tt);
|
d@0
|
115 ST(&(xo[12]), VFNMSI(TA, Tz), ovs, &(xo[0]));
|
d@0
|
116 ST(&(xo[18]), VFMAI(TA, Tz), ovs, &(xo[2]));
|
d@0
|
117 ST(&(xo[6]), VFNMSI(Ty, Tv), ovs, &(xo[2]));
|
d@0
|
118 ST(&(xo[24]), VFMAI(Ty, Tv), ovs, &(xo[0]));
|
d@0
|
119 T18 = VFNMS(LDK(KP618033988), TO, TR);
|
d@0
|
120 TS = VFMA(LDK(KP618033988), TR, TO);
|
d@0
|
121 TW = VFMA(LDK(KP618033988), TV, TU);
|
d@0
|
122 T1a = VFNMS(LDK(KP618033988), TU, TV);
|
d@0
|
123 T13 = VFMA(LDK(KP559016994), T12, T11);
|
d@0
|
124 T1b = VFNMS(LDK(KP559016994), T12, T11);
|
d@0
|
125 TL = VFMA(LDK(KP559016994), TK, TJ);
|
d@0
|
126 T17 = VFNMS(LDK(KP559016994), TK, TJ);
|
d@0
|
127 TT = VFMA(LDK(KP823639103), TS, TL);
|
d@0
|
128 T15 = VFNMS(LDK(KP823639103), TS, TL);
|
d@0
|
129 T16 = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T13, TW));
|
d@0
|
130 T14 = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T13, TW));
|
d@0
|
131 ST(&(xo[2]), VFNMSI(T14, TT), ovs, &(xo[2]));
|
d@0
|
132 ST(&(xo[28]), VFMAI(T14, TT), ovs, &(xo[0]));
|
d@0
|
133 T1d = VFNMS(LDK(KP823639103), T18, T17);
|
d@0
|
134 T19 = VFMA(LDK(KP823639103), T18, T17);
|
d@0
|
135 T1c = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1b, T1a));
|
d@0
|
136 T1e = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1b, T1a));
|
d@0
|
137 ST(&(xo[16]), VFNMSI(T1e, T1d), ovs, &(xo[0]));
|
d@0
|
138 ST(&(xo[14]), VFMAI(T1e, T1d), ovs, &(xo[2]));
|
d@0
|
139 ST(&(xo[22]), VFNMSI(T16, T15), ovs, &(xo[2]));
|
d@0
|
140 ST(&(xo[8]), VFMAI(T16, T15), ovs, &(xo[0]));
|
d@0
|
141 ST(&(xo[26]), VFNMSI(T1c, T19), ovs, &(xo[2]));
|
d@0
|
142 ST(&(xo[4]), VFMAI(T1c, T19), ovs, &(xo[0]));
|
d@0
|
143 }
|
d@0
|
144 }
|