1 /*Copyright (c) 2013, Xiph.Org Foundation and contributors.
2
3 All rights reserved.
4
5 Redistribution and use in source and binary forms, with or without
6 modification, are permitted provided that the following conditions are met:
7
8 * Redistributions of source code must retain the above copyright notice,
9 this list of conditions and the following disclaimer.
10 * Redistributions in binary form must reproduce the above copyright notice,
11 this list of conditions and the following disclaimer in the
12 documentation and/or other materials provided with the distribution.
13
14 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
15 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
18 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
19 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
20 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
21 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
22 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
23 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
24 POSSIBILITY OF SUCH DAMAGE.*/
25
26 #ifndef KISS_FFT_MIPSR1_H
27 #define KISS_FFT_MIPSR1_H
28
29 #if !defined(KISS_FFT_GUTS_H)
30 #error "This file should only be included from _kiss_fft_guts.h"
31 #endif
32
33 #ifdef FIXED_POINT
34
35 #define S_MUL_ADD(a, b, c, d) (S_MUL(a,b)+S_MUL(c,d))
36 #define S_MUL_SUB(a, b, c, d) (S_MUL(a,b)-S_MUL(c,d))
37
38 #undef S_MUL_ADD
S_MUL_ADD(int a,int b,int c,int d)39 static inline int S_MUL_ADD(int a, int b, int c, int d) {
40 int m;
41 asm volatile("MULT $ac1, %0, %1" : : "r" ((int)a), "r" ((int)b));
42 asm volatile("madd $ac1, %0, %1" : : "r" ((int)c), "r" ((int)d));
43 asm volatile("EXTR.W %0,$ac1, %1" : "=r" (m): "i" (15));
44 return m;
45 }
46
47 #undef S_MUL_SUB
S_MUL_SUB(int a,int b,int c,int d)48 static inline int S_MUL_SUB(int a, int b, int c, int d) {
49 int m;
50 asm volatile("MULT $ac1, %0, %1" : : "r" ((int)a), "r" ((int)b));
51 asm volatile("msub $ac1, %0, %1" : : "r" ((int)c), "r" ((int)d));
52 asm volatile("EXTR.W %0,$ac1, %1" : "=r" (m): "i" (15));
53 return m;
54 }
55
56 #undef C_MUL
57 # define C_MUL(m,a,b) (m=C_MUL_fun(a,b))
C_MUL_fun(kiss_fft_cpx a,kiss_twiddle_cpx b)58 static inline kiss_fft_cpx C_MUL_fun(kiss_fft_cpx a, kiss_twiddle_cpx b) {
59 kiss_fft_cpx m;
60
61 asm volatile("MULT $ac1, %0, %1" : : "r" ((int)a.r), "r" ((int)b.r));
62 asm volatile("msub $ac1, %0, %1" : : "r" ((int)a.i), "r" ((int)b.i));
63 asm volatile("EXTR.W %0,$ac1, %1" : "=r" (m.r): "i" (15));
64 asm volatile("MULT $ac1, %0, %1" : : "r" ((int)a.r), "r" ((int)b.i));
65 asm volatile("madd $ac1, %0, %1" : : "r" ((int)a.i), "r" ((int)b.r));
66 asm volatile("EXTR.W %0,$ac1, %1" : "=r" (m.i): "i" (15));
67
68 return m;
69 }
70 #undef C_MULC
71 # define C_MULC(m,a,b) (m=C_MULC_fun(a,b))
C_MULC_fun(kiss_fft_cpx a,kiss_twiddle_cpx b)72 static inline kiss_fft_cpx C_MULC_fun(kiss_fft_cpx a, kiss_twiddle_cpx b) {
73 kiss_fft_cpx m;
74
75 asm volatile("MULT $ac1, %0, %1" : : "r" ((int)a.r), "r" ((int)b.r));
76 asm volatile("madd $ac1, %0, %1" : : "r" ((int)a.i), "r" ((int)b.i));
77 asm volatile("EXTR.W %0,$ac1, %1" : "=r" (m.r): "i" (15));
78 asm volatile("MULT $ac1, %0, %1" : : "r" ((int)a.i), "r" ((int)b.r));
79 asm volatile("msub $ac1, %0, %1" : : "r" ((int)a.r), "r" ((int)b.i));
80 asm volatile("EXTR.W %0,$ac1, %1" : "=r" (m.i): "i" (15));
81
82 return m;
83 }
84
85 #endif /* FIXED_POINT */
86
87 #define OVERRIDE_kf_bfly5
kf_bfly5(kiss_fft_cpx * Fout,const size_t fstride,const kiss_fft_state * st,int m,int N,int mm)88 static void kf_bfly5(
89 kiss_fft_cpx * Fout,
90 const size_t fstride,
91 const kiss_fft_state *st,
92 int m,
93 int N,
94 int mm
95 )
96 {
97 kiss_fft_cpx *Fout0,*Fout1,*Fout2,*Fout3,*Fout4;
98 int i, u;
99 kiss_fft_cpx scratch[13];
100
101 const kiss_twiddle_cpx *tw;
102 kiss_twiddle_cpx ya,yb;
103 kiss_fft_cpx * Fout_beg = Fout;
104
105 #ifdef FIXED_POINT
106 ya.r = 10126;
107 ya.i = -31164;
108 yb.r = -26510;
109 yb.i = -19261;
110 #else
111 ya = st->twiddles[fstride*m];
112 yb = st->twiddles[fstride*2*m];
113 #endif
114
115 tw=st->twiddles;
116
117 for (i=0;i<N;i++)
118 {
119 Fout = Fout_beg + i*mm;
120 Fout0=Fout;
121 Fout1=Fout0+m;
122 Fout2=Fout0+2*m;
123 Fout3=Fout0+3*m;
124 Fout4=Fout0+4*m;
125
126 /* For non-custom modes, m is guaranteed to be a multiple of 4. */
127 for ( u=0; u<m; ++u ) {
128 scratch[0] = *Fout0;
129
130
131 C_MUL(scratch[1] ,*Fout1, tw[u*fstride]);
132 C_MUL(scratch[2] ,*Fout2, tw[2*u*fstride]);
133 C_MUL(scratch[3] ,*Fout3, tw[3*u*fstride]);
134 C_MUL(scratch[4] ,*Fout4, tw[4*u*fstride]);
135
136 C_ADD( scratch[7],scratch[1],scratch[4]);
137 C_SUB( scratch[10],scratch[1],scratch[4]);
138 C_ADD( scratch[8],scratch[2],scratch[3]);
139 C_SUB( scratch[9],scratch[2],scratch[3]);
140
141 Fout0->r += scratch[7].r + scratch[8].r;
142 Fout0->i += scratch[7].i + scratch[8].i;
143 scratch[5].r = scratch[0].r + S_MUL_ADD(scratch[7].r,ya.r,scratch[8].r,yb.r);
144 scratch[5].i = scratch[0].i + S_MUL_ADD(scratch[7].i,ya.r,scratch[8].i,yb.r);
145
146 scratch[6].r = S_MUL_ADD(scratch[10].i,ya.i,scratch[9].i,yb.i);
147 scratch[6].i = -S_MUL_ADD(scratch[10].r,ya.i,scratch[9].r,yb.i);
148
149 C_SUB(*Fout1,scratch[5],scratch[6]);
150 C_ADD(*Fout4,scratch[5],scratch[6]);
151
152 scratch[11].r = scratch[0].r + S_MUL_ADD(scratch[7].r,yb.r,scratch[8].r,ya.r);
153 scratch[11].i = scratch[0].i + S_MUL_ADD(scratch[7].i,yb.r,scratch[8].i,ya.r);
154
155 scratch[12].r = S_MUL_SUB(scratch[9].i,ya.i,scratch[10].i,yb.i);
156 scratch[12].i = S_MUL_SUB(scratch[10].r,yb.i,scratch[9].r,ya.i);
157
158 C_ADD(*Fout2,scratch[11],scratch[12]);
159 C_SUB(*Fout3,scratch[11],scratch[12]);
160
161 ++Fout0;++Fout1;++Fout2;++Fout3;++Fout4;
162 }
163 }
164 }
165
166
167 #endif /* KISS_FFT_MIPSR1_H */
168