Inlines.h 7.22 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
/***********************************************************************
Copyright (c) 2006-2011, Skype Limited. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, (subject to the limitations in the disclaimer below)
are permitted provided that the following conditions are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Skype Limited, nor the names of specific
contributors, may be used to endorse or promote products derived from
this software without specific prior written permission.
NO EXPRESS OR IMPLIED LICENSES TO ANY PARTY'S PATENT RIGHTS ARE GRANTED
BY THIS LICENSE. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
CONTRIBUTORS ''AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING,
BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/

/*! \file silk_Inlines.h
 *  \brief silk_Inlines.h defines inline signal processing functions.
 */

#ifndef _SILK_FIX_INLINES_H_
#define _SILK_FIX_INLINES_H_

#ifdef  __cplusplus
extern "C"
{
#endif

/* count leading zeros of opus_int64 */
41
static inline opus_int32 silk_CLZ64(opus_int64 in)
42
43
44
{
    opus_int32 in_upper;

45
    in_upper = (opus_int32)silk_RSHIFT64(in, 32);
46
47
48
49
50
51
52
53
54
55
    if (in_upper == 0) {
        /* Search in the lower 32 bits */
        return 32 + silk_CLZ32( (opus_int32) in );
    } else {
        /* Search in the upper 32 bits */
        return silk_CLZ32( in_upper );
    }
}

/* get number of leading zeros and fractional part (the bits right after the leading one */
56
static inline void silk_CLZ_FRAC(opus_int32 in,            /* I: input */
57
58
59
60
61
62
63
64
65
66
67
68
                                    opus_int32 *lz,           /* O: number of leading zeros */
                                    opus_int32 *frac_Q7)      /* O: the 7 bits right after the leading one */
{
    opus_int32 lzeros = silk_CLZ32(in);

    * lz = lzeros;
    * frac_Q7 = silk_ROR32(in, 24 - lzeros) & 0x7f;
}

/* Approximation of square root                                          */
/* Accuracy: < +/- 10%  for output values > 15                           */
/*           < +/- 2.5% for output values > 120                          */
69
static inline opus_int32 silk_SQRT_APPROX(opus_int32 x)
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
{
    opus_int32 y, lz, frac_Q7;

    if( x <= 0 ) {
        return 0;
    }

    silk_CLZ_FRAC(x, &lz, &frac_Q7);

    if( lz & 1 ) {
        y = 32768;
    } else {
        y = 46214;        /* 46214 = sqrt(2) * 32768 */
    }

    /* get scaling right */
86
    y >>= silk_RSHIFT(lz, 1);
87
88

    /* increment using fractional part of input */
89
    y = silk_SMLAWB(y, y, silk_SMULBB(213, frac_Q7));
90
91
92
93
94

    return y;
}

/* Divide two int32 values and return result as int32 in a given Q-domain */
95
static inline opus_int32 silk_DIV32_varQ(    /* O    returns a good approximation of "(a32 << Qres) / b32" */
96
97
98
99
100
101
102
103
    const opus_int32     a32,            /* I    numerator (Q0)                  */
    const opus_int32     b32,            /* I    denominator (Q0)                */
    const opus_int       Qres            /* I    Q-domain of result (>= 0)       */
)
{
    opus_int   a_headrm, b_headrm, lshift;
    opus_int32 b32_inv, a32_nrm, b32_nrm, result;

104
105
    silk_assert( b32 != 0 );
    silk_assert( Qres >= 0 );
106
107

    /* Compute number of bits head room and normalize inputs */
108
109
110
111
    a_headrm = silk_CLZ32( silk_abs(a32) ) - 1;
    a32_nrm = silk_LSHIFT(a32, a_headrm);                                    /* Q: a_headrm                    */
    b_headrm = silk_CLZ32( silk_abs(b32) ) - 1;
    b32_nrm = silk_LSHIFT(b32, b_headrm);                                    /* Q: b_headrm                    */
112
113

    /* Inverse of b32, with 14 bits of precision */
114
    b32_inv = silk_DIV32_16( silk_int32_MAX >> 2, silk_RSHIFT(b32_nrm, 16) );  /* Q: 29 + 16 - b_headrm        */
115
116

    /* First approximation */
117
    result = silk_SMULWB(a32_nrm, b32_inv);                                  /* Q: 29 + a_headrm - b_headrm    */
118
119

    /* Compute residual by subtracting product of denominator and first approximation */
120
121
    /* It's OK to overflow because the final value of a32_nrm should always be small */
    a32_nrm = silk_SUB32_ovflw(a32_nrm, silk_LSHIFT_ovflw( silk_SMMUL(b32_nrm, result), 3 ));  /* Q: a_headrm                    */
122
123

    /* Refinement */
124
    result = silk_SMLAWB(result, a32_nrm, b32_inv);                          /* Q: 29 + a_headrm - b_headrm    */
125
126
127
128

    /* Convert to Qres domain */
    lshift = 29 + a_headrm - b_headrm - Qres;
    if( lshift < 0 ) {
129
        return silk_LSHIFT_SAT32(result, -lshift);
130
131
    } else {
        if( lshift < 32){
132
            return silk_RSHIFT(result, lshift);
133
134
135
136
137
138
139
140
        } else {
            /* Avoid undefined result */
            return 0;
        }
    }
}

/* Invert int32 value and return result as int32 in a given Q-domain */
141
static inline opus_int32 silk_INVERSE32_varQ(    /* O    returns a good approximation of "(1 << Qres) / b32" */
142
143
144
145
146
147
148
    const opus_int32     b32,                /* I    denominator (Q0)                */
    const opus_int       Qres                /* I    Q-domain of result (> 0)        */
)
{
    opus_int   b_headrm, lshift;
    opus_int32 b32_inv, b32_nrm, err_Q32, result;

149
150
    silk_assert( b32 != 0 );
    silk_assert( Qres > 0 );
151
152

    /* Compute number of bits head room and normalize input */
153
154
    b_headrm = silk_CLZ32( silk_abs(b32) ) - 1;
    b32_nrm = silk_LSHIFT(b32, b_headrm);                                    /* Q: b_headrm                */
155
156

    /* Inverse of b32, with 14 bits of precision */
157
    b32_inv = silk_DIV32_16( silk_int32_MAX >> 2, silk_RSHIFT(b32_nrm, 16) );  /* Q: 29 + 16 - b_headrm    */
158
159

    /* First approximation */
160
    result = silk_LSHIFT(b32_inv, 16);                                       /* Q: 61 - b_headrm            */
161
162

    /* Compute residual by subtracting product of denominator and first approximation from one */
163
    err_Q32 = silk_LSHIFT( (1<<29) - silk_SMULWB(b32_nrm, b32_inv), 3 );         /* Q32                        */
164
165

    /* Refinement */
166
    result = silk_SMLAWW(result, err_Q32, b32_inv);                          /* Q: 61 - b_headrm            */
167
168
169
170

    /* Convert to Qres domain */
    lshift = 61 - b_headrm - Qres;
    if( lshift <= 0 ) {
171
        return silk_LSHIFT_SAT32(result, -lshift);
172
173
    } else {
        if( lshift < 32){
174
            return silk_RSHIFT(result, lshift);
175
176
177
178
179
180
181
182
183
184
185
        }else{
            /* Avoid undefined result */
            return 0;
        }
    }
}

#ifdef  __cplusplus
}
#endif

186
#endif /*_SILK_FIX_INLINES_H_*/