GCC Code Coverage Report
Directory: ../../../ffmpeg/ Exec Total Coverage
File: src/libavcodec/celp_filters.c Lines: 94 103 91.3 %
Date: 2019-11-18 18:00:01 Branches: 23 32 71.9 %

Line Branch Exec Source
1
/*
2
 * various filters for ACELP-based codecs
3
 *
4
 * Copyright (c) 2008 Vladimir Voroshilov
5
 *
6
 * This file is part of FFmpeg.
7
 *
8
 * FFmpeg is free software; you can redistribute it and/or
9
 * modify it under the terms of the GNU Lesser General Public
10
 * License as published by the Free Software Foundation; either
11
 * version 2.1 of the License, or (at your option) any later version.
12
 *
13
 * FFmpeg is distributed in the hope that it will be useful,
14
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
16
 * Lesser General Public License for more details.
17
 *
18
 * You should have received a copy of the GNU Lesser General Public
19
 * License along with FFmpeg; if not, write to the Free Software
20
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21
 */
22
23
#include <inttypes.h>
24
25
#include "avcodec.h"
26
#include "celp_filters.h"
27
#include "libavutil/avassert.h"
28
#include "libavutil/common.h"
29
30
void ff_celp_convolve_circ(int16_t* fc_out, const int16_t* fc_in,
31
                           const int16_t* filter, int len)
32
{
33
    int i, k;
34
35
    memset(fc_out, 0, len * sizeof(int16_t));
36
37
    /* Since there are few pulses over an entire subframe (i.e. almost
38
       all fc_in[i] are zero) it is faster to loop over fc_in first. */
39
    for (i = 0; i < len; i++) {
40
        if (fc_in[i]) {
41
            for (k = 0; k < i; k++)
42
                fc_out[k] += (fc_in[i] * filter[len + k - i]) >> 15;
43
44
            for (k = i; k < len; k++)
45
                fc_out[k] += (fc_in[i] * filter[      k - i]) >> 15;
46
        }
47
    }
48
}
49
50
34942
void ff_celp_circ_addf(float *out, const float *in,
51
                       const float *lagged, int lag, float fac, int n)
52
{
53
    int k;
54
1003092
    for (k = 0; k < lag; k++)
55
968150
        out[k] = in[k] + fac * lagged[n + k - lag];
56
978768
    for (; k < n; k++)
57
943826
        out[k] = in[k] + fac * lagged[    k - lag];
58
34942
}
59
60
26880
int ff_celp_lp_synthesis_filter(int16_t *out, const int16_t *filter_coeffs,
61
                                const int16_t *in, int buffer_length,
62
                                int filter_length, int stop_on_overflow,
63
                                int shift, int rounder)
64
{
65
    int i,n;
66
67
1152160
    for (n = 0; n < buffer_length; n++) {
68
1125280
        int sum = -rounder, sum1;
69
12378080
        for (i = 1; i <= filter_length; i++)
70
11252800
            sum += (unsigned)(filter_coeffs[i-1] * out[n-i]);
71
72
1125280
        sum1 = ((-sum >> 12) + in[n]) >> shift;
73
1125280
        sum  = av_clip_int16(sum1);
74
75

1125280
        if (stop_on_overflow && sum != sum1)
76
            return 1;
77
78
1125280
        out[n] = sum;
79
    }
80
81
26880
    return 0;
82
}
83
84
699422
void ff_celp_lp_synthesis_filterf(float *out, const float *filter_coeffs,
85
                                  const float* in, int buffer_length,
86
                                  int filter_length)
87
{
88
    int i,n;
89
90
#if 0 // Unoptimized code path for improved readability
91
    for (n = 0; n < buffer_length; n++) {
92
        out[n] = in[n];
93
        for (i = 1; i <= filter_length; i++)
94
            out[n] -= filter_coeffs[i-1] * out[n-i];
95
    }
96
#else
97
    float out0, out1, out2, out3;
98
    float old_out0, old_out1, old_out2, old_out3;
99
    float a,b,c;
100
101
699422
    a = filter_coeffs[0];
102
699422
    b = filter_coeffs[1];
103
699422
    c = filter_coeffs[2];
104
699422
    b -= filter_coeffs[0] * filter_coeffs[0];
105
699422
    c -= filter_coeffs[1] * filter_coeffs[0];
106
699422
    c -= filter_coeffs[0] * b;
107
108
    av_assert2((filter_length&1)==0 && filter_length>=4);
109
110
699422
    old_out0 = out[-4];
111
699422
    old_out1 = out[-3];
112
699422
    old_out2 = out[-2];
113
699422
    old_out3 = out[-1];
114
7646966
    for (n = 0; n <= buffer_length - 4; n+=4) {
115
        float tmp0,tmp1,tmp2;
116
        float val;
117
118
6947544
        out0 = in[0];
119
6947544
        out1 = in[1];
120
6947544
        out2 = in[2];
121
6947544
        out3 = in[3];
122
123
6947544
        out0 -= filter_coeffs[2] * old_out1;
124
6947544
        out1 -= filter_coeffs[2] * old_out2;
125
6947544
        out2 -= filter_coeffs[2] * old_out3;
126
127
6947544
        out0 -= filter_coeffs[1] * old_out2;
128
6947544
        out1 -= filter_coeffs[1] * old_out3;
129
130
6947544
        out0 -= filter_coeffs[0] * old_out3;
131
132
6947544
        val = filter_coeffs[3];
133
134
6947544
        out0 -= val * old_out0;
135
6947544
        out1 -= val * old_out1;
136
6947544
        out2 -= val * old_out2;
137
6947544
        out3 -= val * old_out3;
138
139
32827556
        for (i = 5; i < filter_length; i += 2) {
140
25880012
            old_out3 = out[-i];
141
25880012
            val = filter_coeffs[i-1];
142
143
25880012
            out0 -= val * old_out3;
144
25880012
            out1 -= val * old_out0;
145
25880012
            out2 -= val * old_out1;
146
25880012
            out3 -= val * old_out2;
147
148
25880012
            old_out2 = out[-i-1];
149
150
25880012
            val = filter_coeffs[i];
151
152
25880012
            out0 -= val * old_out2;
153
25880012
            out1 -= val * old_out3;
154
25880012
            out2 -= val * old_out0;
155
25880012
            out3 -= val * old_out1;
156
157
25880012
            FFSWAP(float, old_out0, old_out2);
158
25880012
            old_out1 = old_out3;
159
        }
160
161
6947544
        tmp0 = out0;
162
6947544
        tmp1 = out1;
163
6947544
        tmp2 = out2;
164
165
6947544
        out3 -= a * tmp2;
166
6947544
        out2 -= a * tmp1;
167
6947544
        out1 -= a * tmp0;
168
169
6947544
        out3 -= b * tmp1;
170
6947544
        out2 -= b * tmp0;
171
172
6947544
        out3 -= c * tmp0;
173
174
175
6947544
        out[0] = out0;
176
6947544
        out[1] = out1;
177
6947544
        out[2] = out2;
178
6947544
        out[3] = out3;
179
180
6947544
        old_out0 = out0;
181
6947544
        old_out1 = out1;
182
6947544
        old_out2 = out2;
183
6947544
        old_out3 = out3;
184
185
6947544
        out += 4;
186
6947544
        in  += 4;
187
    }
188
189
699422
    out -= n;
190
699422
    in -= n;
191
815434
    for (; n < buffer_length; n++) {
192
116012
        out[n] = in[n];
193
3429868
        for (i = 1; i <= filter_length; i++)
194
3313856
            out[n] -= filter_coeffs[i-1] * out[n-i];
195
    }
196
#endif
197
699422
}
198
199
19555
void ff_celp_lp_zero_synthesis_filterf(float *out, const float *filter_coeffs,
200
                                       const float *in, int buffer_length,
201
                                       int filter_length)
202
{
203
    int i,n;
204
205
1152515
    for (n = 0; n < buffer_length; n++) {
206
1132960
        out[n] = in[n];
207
14046560
        for (i = 1; i <= filter_length; i++)
208
12913600
            out[n] += filter_coeffs[i-1] * in[n-i];
209
    }
210
19555
}
211
212
36
void ff_celp_filter_init(CELPFContext *c)
213
{
214
36
    c->celp_lp_synthesis_filterf        = ff_celp_lp_synthesis_filterf;
215
36
    c->celp_lp_zero_synthesis_filterf   = ff_celp_lp_zero_synthesis_filterf;
216
217
    if(HAVE_MIPSFPU)
218
        ff_celp_filter_init_mips(c);
219
36
}