2 * copyright (c) 2008 Michael Niedermayer <michaelni@gmx.at>
3 * Copyright (C) 2016 foo86
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 #include "synth_filter.h"
27 static void synth_filter_float(FFTContext *imdct,
28 float *synth_buf_ptr, int *synth_buf_offset,
29 float synth_buf2[32], const float window[512],
30 float out[32], const float in[32], float scale)
32 float *synth_buf = synth_buf_ptr + *synth_buf_offset;
35 imdct->imdct_half(imdct, synth_buf, in);
37 for (i = 0; i < 16; i++) {
38 float a = synth_buf2[i ];
39 float b = synth_buf2[i + 16];
42 for (j = 0; j < 512 - *synth_buf_offset; j += 64) {
43 a += window[i + j ] * (-synth_buf[15 - i + j ]);
44 b += window[i + j + 16] * ( synth_buf[ i + j ]);
45 c += window[i + j + 32] * ( synth_buf[16 + i + j ]);
46 d += window[i + j + 48] * ( synth_buf[31 - i + j ]);
48 for ( ; j < 512; j += 64) {
49 a += window[i + j ] * (-synth_buf[15 - i + j - 512]);
50 b += window[i + j + 16] * ( synth_buf[ i + j - 512]);
51 c += window[i + j + 32] * ( synth_buf[16 + i + j - 512]);
52 d += window[i + j + 48] * ( synth_buf[31 - i + j - 512]);
55 out[i + 16] = b * scale;
57 synth_buf2[i + 16] = d;
60 *synth_buf_offset = (*synth_buf_offset - 32) & 511;
63 static void synth_filter_float_64(FFTContext *imdct,
64 float *synth_buf_ptr, int *synth_buf_offset,
65 float synth_buf2[64], const float window[1024],
66 float out[64], const float in[64], float scale)
68 float *synth_buf = synth_buf_ptr + *synth_buf_offset;
71 imdct->imdct_half(imdct, synth_buf, in);
73 for (i = 0; i < 32; i++) {
74 float a = synth_buf2[i ];
75 float b = synth_buf2[i + 32];
78 for (j = 0; j < 1024 - *synth_buf_offset; j += 128) {
79 a += window[i + j ] * (-synth_buf[31 - i + j ]);
80 b += window[i + j + 32] * ( synth_buf[ i + j ]);
81 c += window[i + j + 64] * ( synth_buf[32 + i + j ]);
82 d += window[i + j + 96] * ( synth_buf[63 - i + j ]);
84 for ( ; j < 1024; j += 128) {
85 a += window[i + j ] * (-synth_buf[31 - i + j - 1024]);
86 b += window[i + j + 32] * ( synth_buf[ i + j - 1024]);
87 c += window[i + j + 64] * ( synth_buf[32 + i + j - 1024]);
88 d += window[i + j + 96] * ( synth_buf[63 - i + j - 1024]);
91 out[i + 32] = b * scale;
93 synth_buf2[i + 32] = d;
96 *synth_buf_offset = (*synth_buf_offset - 64) & 1023;
99 static void synth_filter_fixed(DCADCTContext *imdct,
100 int32_t *synth_buf_ptr, int *synth_buf_offset,
101 int32_t synth_buf2[32], const int32_t window[512],
102 int32_t out[32], const int32_t in[32])
104 int32_t *synth_buf = synth_buf_ptr + *synth_buf_offset;
107 imdct->imdct_half[0](synth_buf, in);
109 for (i = 0; i < 16; i++) {
110 int64_t a = synth_buf2[i ] * (INT64_C(1) << 21);
111 int64_t b = synth_buf2[i + 16] * (INT64_C(1) << 21);
114 for (j = 0; j < 512 - *synth_buf_offset; j += 64) {
115 a += (int64_t)window[i + j ] * synth_buf[ i + j ];
116 b += (int64_t)window[i + j + 16] * synth_buf[15 - i + j ];
117 c += (int64_t)window[i + j + 32] * synth_buf[16 + i + j ];
118 d += (int64_t)window[i + j + 48] * synth_buf[31 - i + j ];
120 for ( ; j < 512; j += 64) {
121 a += (int64_t)window[i + j ] * synth_buf[ i + j - 512];
122 b += (int64_t)window[i + j + 16] * synth_buf[15 - i + j - 512];
123 c += (int64_t)window[i + j + 32] * synth_buf[16 + i + j - 512];
124 d += (int64_t)window[i + j + 48] * synth_buf[31 - i + j - 512];
126 out[i ] = clip23(norm21(a));
127 out[i + 16] = clip23(norm21(b));
128 synth_buf2[i ] = norm21(c);
129 synth_buf2[i + 16] = norm21(d);
132 *synth_buf_offset = (*synth_buf_offset - 32) & 511;
135 static void synth_filter_fixed_64(DCADCTContext *imdct,
136 int32_t *synth_buf_ptr, int *synth_buf_offset,
137 int32_t synth_buf2[64], const int32_t window[1024],
138 int32_t out[64], const int32_t in[64])
140 int32_t *synth_buf = synth_buf_ptr + *synth_buf_offset;
143 imdct->imdct_half[1](synth_buf, in);
145 for (i = 0; i < 32; i++) {
146 int64_t a = synth_buf2[i ] * (INT64_C(1) << 20);
147 int64_t b = synth_buf2[i + 32] * (INT64_C(1) << 20);
150 for (j = 0; j < 1024 - *synth_buf_offset; j += 128) {
151 a += (int64_t)window[i + j ] * synth_buf[ i + j ];
152 b += (int64_t)window[i + j + 32] * synth_buf[31 - i + j ];
153 c += (int64_t)window[i + j + 64] * synth_buf[32 + i + j ];
154 d += (int64_t)window[i + j + 96] * synth_buf[63 - i + j ];
156 for ( ; j < 1024; j += 128) {
157 a += (int64_t)window[i + j ] * synth_buf[ i + j - 1024];
158 b += (int64_t)window[i + j + 32] * synth_buf[31 - i + j - 1024];
159 c += (int64_t)window[i + j + 64] * synth_buf[32 + i + j - 1024];
160 d += (int64_t)window[i + j + 96] * synth_buf[63 - i + j - 1024];
162 out[i ] = clip23(norm20(a));
163 out[i + 32] = clip23(norm20(b));
164 synth_buf2[i ] = norm20(c);
165 synth_buf2[i + 32] = norm20(d);
168 *synth_buf_offset = (*synth_buf_offset - 64) & 1023;
171 av_cold void ff_synth_filter_init(SynthFilterContext *c)
173 c->synth_filter_float = synth_filter_float;
174 c->synth_filter_float_64 = synth_filter_float_64;
175 c->synth_filter_fixed = synth_filter_fixed;
176 c->synth_filter_fixed_64 = synth_filter_fixed_64;
179 ff_synth_filter_init_aarch64(c);
181 ff_synth_filter_init_arm(c);
183 ff_synth_filter_init_x86(c);