1 |
|
|
/* |
2 |
|
|
* WMA compatible decoder |
3 |
|
|
* Copyright (c) 2002 The FFmpeg Project |
4 |
|
|
* |
5 |
|
|
* This file is part of FFmpeg. |
6 |
|
|
* |
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. |
11 |
|
|
* |
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. |
16 |
|
|
* |
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 |
20 |
|
|
*/ |
21 |
|
|
|
22 |
|
|
/** |
23 |
|
|
* @file |
24 |
|
|
* WMA compatible decoder. |
25 |
|
|
* This decoder handles Microsoft Windows Media Audio data, versions 1 & 2. |
26 |
|
|
* WMA v1 is identified by audio format 0x160 in Microsoft media files |
27 |
|
|
* (ASF/AVI/WAV). WMA v2 is identified by audio format 0x161. |
28 |
|
|
* |
29 |
|
|
* To use this decoder, a calling application must supply the extra data |
30 |
|
|
* bytes provided with the WMA data. These are the extra, codec-specific |
31 |
|
|
* bytes at the end of a WAVEFORMATEX data structure. Transmit these bytes |
32 |
|
|
* to the decoder using the extradata[_size] fields in AVCodecContext. There |
33 |
|
|
* should be 4 extra bytes for v1 data and 6 extra bytes for v2 data. |
34 |
|
|
*/ |
35 |
|
|
|
36 |
|
|
#include "libavutil/attributes.h" |
37 |
|
|
#include "libavutil/ffmath.h" |
38 |
|
|
|
39 |
|
|
#include "avcodec.h" |
40 |
|
|
#include "internal.h" |
41 |
|
|
#include "wma.h" |
42 |
|
|
|
43 |
|
|
#define EXPVLCBITS 8 |
44 |
|
|
#define EXPMAX ((19 + EXPVLCBITS - 1) / EXPVLCBITS) |
45 |
|
|
|
46 |
|
|
#define HGAINVLCBITS 9 |
47 |
|
|
#define HGAINMAX ((13 + HGAINVLCBITS - 1) / HGAINVLCBITS) |
48 |
|
|
|
49 |
|
|
static void wma_lsp_to_curve_init(WMACodecContext *s, int frame_len); |
50 |
|
|
|
51 |
|
|
#ifdef TRACE |
52 |
|
|
static void dump_floats(WMACodecContext *s, const char *name, |
53 |
|
|
int prec, const float *tab, int n) |
54 |
|
|
{ |
55 |
|
|
int i; |
56 |
|
|
|
57 |
|
|
ff_tlog(s->avctx, "%s[%d]:\n", name, n); |
58 |
|
|
for (i = 0; i < n; i++) { |
59 |
|
|
if ((i & 7) == 0) |
60 |
|
|
ff_tlog(s->avctx, "%4d: ", i); |
61 |
|
|
ff_tlog(s->avctx, " %8.*f", prec, tab[i]); |
62 |
|
|
if ((i & 7) == 7) |
63 |
|
|
ff_tlog(s->avctx, "\n"); |
64 |
|
|
} |
65 |
|
|
if ((i & 7) != 0) |
66 |
|
|
ff_tlog(s->avctx, "\n"); |
67 |
|
|
} |
68 |
|
|
#endif /* TRACE */ |
69 |
|
|
|
70 |
|
14 |
static av_cold int wma_decode_init(AVCodecContext *avctx) |
71 |
|
|
{ |
72 |
|
14 |
WMACodecContext *s = avctx->priv_data; |
73 |
|
|
int i, flags2; |
74 |
|
|
uint8_t *extradata; |
75 |
|
|
|
76 |
✗✓ |
14 |
if (!avctx->block_align) { |
77 |
|
|
av_log(avctx, AV_LOG_ERROR, "block_align is not set\n"); |
78 |
|
|
return AVERROR(EINVAL); |
79 |
|
|
} |
80 |
|
|
|
81 |
|
14 |
s->avctx = avctx; |
82 |
|
|
|
83 |
|
|
/* extract flag info */ |
84 |
|
14 |
flags2 = 0; |
85 |
|
14 |
extradata = avctx->extradata; |
86 |
✓✓✓✗
|
14 |
if (avctx->codec->id == AV_CODEC_ID_WMAV1 && avctx->extradata_size >= 4) |
87 |
|
2 |
flags2 = AV_RL16(extradata + 2); |
88 |
✓✗✓✗
|
12 |
else if (avctx->codec->id == AV_CODEC_ID_WMAV2 && avctx->extradata_size >= 6) |
89 |
|
12 |
flags2 = AV_RL16(extradata + 4); |
90 |
|
|
|
91 |
|
14 |
s->use_exp_vlc = flags2 & 0x0001; |
92 |
|
14 |
s->use_bit_reservoir = flags2 & 0x0002; |
93 |
|
14 |
s->use_variable_block_len = flags2 & 0x0004; |
94 |
|
|
|
95 |
✓✓✓✓
|
14 |
if (avctx->codec->id == AV_CODEC_ID_WMAV2 && avctx->extradata_size >= 8){ |
96 |
✗✓✗✗
|
11 |
if (AV_RL16(extradata+4)==0xd && s->use_variable_block_len){ |
97 |
|
|
av_log(avctx, AV_LOG_WARNING, "Disabling use_variable_block_len, if this fails contact the ffmpeg developers and send us the file\n"); |
98 |
|
|
s->use_variable_block_len= 0; // this fixes issue1503 |
99 |
|
|
} |
100 |
|
|
} |
101 |
|
|
|
102 |
✓✓ |
42 |
for (i=0; i<MAX_CHANNELS; i++) |
103 |
|
28 |
s->max_exponent[i] = 1.0; |
104 |
|
|
|
105 |
✗✓ |
14 |
if (ff_wma_init(avctx, flags2) < 0) |
106 |
|
|
return -1; |
107 |
|
|
|
108 |
|
|
/* init MDCT */ |
109 |
✓✓ |
51 |
for (i = 0; i < s->nb_block_sizes; i++) |
110 |
|
37 |
ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 1, 1.0 / 32768.0); |
111 |
|
|
|
112 |
✓✓ |
14 |
if (s->use_noise_coding) { |
113 |
|
2 |
ff_init_vlc_from_lengths(&s->hgain_vlc, HGAINVLCBITS, FF_ARRAY_ELEMS(ff_wma_hgain_hufftab), |
114 |
|
|
&ff_wma_hgain_hufftab[0][1], 2, |
115 |
|
|
&ff_wma_hgain_hufftab[0][0], 2, 1, -18, 0, avctx); |
116 |
|
|
} |
117 |
|
|
|
118 |
✓✓ |
14 |
if (s->use_exp_vlc) |
119 |
|
11 |
init_vlc(&s->exp_vlc, EXPVLCBITS, sizeof(ff_aac_scalefactor_bits), // FIXME move out of context |
120 |
|
|
ff_aac_scalefactor_bits, 1, 1, |
121 |
|
|
ff_aac_scalefactor_code, 4, 4, 0); |
122 |
|
|
else |
123 |
|
3 |
wma_lsp_to_curve_init(s, s->frame_len); |
124 |
|
|
|
125 |
|
14 |
avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; |
126 |
|
|
|
127 |
|
14 |
return 0; |
128 |
|
|
} |
129 |
|
|
|
130 |
|
|
/** |
131 |
|
|
* compute x^-0.25 with an exponent and mantissa table. We use linear |
132 |
|
|
* interpolation to reduce the mantissa table size at a small speed |
133 |
|
|
* expense (linear interpolation approximately doubles the number of |
134 |
|
|
* bits of precision). |
135 |
|
|
*/ |
136 |
|
|
static inline float pow_m1_4(WMACodecContext *s, float x) |
137 |
|
|
{ |
138 |
|
|
union { |
139 |
|
|
float f; |
140 |
|
|
unsigned int v; |
141 |
|
|
} u, t; |
142 |
|
|
unsigned int e, m; |
143 |
|
|
float a, b; |
144 |
|
|
|
145 |
|
|
u.f = x; |
146 |
|
|
e = u.v >> 23; |
147 |
|
|
m = (u.v >> (23 - LSP_POW_BITS)) & ((1 << LSP_POW_BITS) - 1); |
148 |
|
|
/* build interpolation scale: 1 <= t < 2. */ |
149 |
|
|
t.v = ((u.v << LSP_POW_BITS) & ((1 << 23) - 1)) | (127 << 23); |
150 |
|
|
a = s->lsp_pow_m_table1[m]; |
151 |
|
|
b = s->lsp_pow_m_table2[m]; |
152 |
|
|
return s->lsp_pow_e_table[e] * (a + b * t.f); |
153 |
|
|
} |
154 |
|
|
|
155 |
|
3 |
static av_cold void wma_lsp_to_curve_init(WMACodecContext *s, int frame_len) |
156 |
|
|
{ |
157 |
|
|
float wdel, a, b; |
158 |
|
|
int i, e, m; |
159 |
|
|
|
160 |
|
3 |
wdel = M_PI / frame_len; |
161 |
✓✓ |
1539 |
for (i = 0; i < frame_len; i++) |
162 |
|
1536 |
s->lsp_cos_table[i] = 2.0f * cos(wdel * i); |
163 |
|
|
|
164 |
|
|
/* tables for x^-0.25 computation */ |
165 |
✓✓ |
771 |
for (i = 0; i < 256; i++) { |
166 |
|
768 |
e = i - 126; |
167 |
|
768 |
s->lsp_pow_e_table[i] = exp2f(e * -0.25); |
168 |
|
|
} |
169 |
|
|
|
170 |
|
|
/* NOTE: these two tables are needed to avoid two operations in |
171 |
|
|
* pow_m1_4 */ |
172 |
|
3 |
b = 1.0; |
173 |
✓✓ |
387 |
for (i = (1 << LSP_POW_BITS) - 1; i >= 0; i--) { |
174 |
|
384 |
m = (1 << LSP_POW_BITS) + i; |
175 |
|
384 |
a = (float) m * (0.5 / (1 << LSP_POW_BITS)); |
176 |
|
384 |
a = 1/sqrt(sqrt(a)); |
177 |
|
384 |
s->lsp_pow_m_table1[i] = 2 * a - b; |
178 |
|
384 |
s->lsp_pow_m_table2[i] = b - a; |
179 |
|
384 |
b = a; |
180 |
|
|
} |
181 |
|
3 |
} |
182 |
|
|
|
183 |
|
|
/** |
184 |
|
|
* NOTE: We use the same code as Vorbis here |
185 |
|
|
* @todo optimize it further with SSE/3Dnow |
186 |
|
|
*/ |
187 |
|
|
static void wma_lsp_to_curve(WMACodecContext *s, float *out, float *val_max_ptr, |
188 |
|
|
int n, float *lsp) |
189 |
|
|
{ |
190 |
|
|
int i, j; |
191 |
|
|
float p, q, w, v, val_max; |
192 |
|
|
|
193 |
|
|
val_max = 0; |
194 |
|
|
for (i = 0; i < n; i++) { |
195 |
|
|
p = 0.5f; |
196 |
|
|
q = 0.5f; |
197 |
|
|
w = s->lsp_cos_table[i]; |
198 |
|
|
for (j = 1; j < NB_LSP_COEFS; j += 2) { |
199 |
|
|
q *= w - lsp[j - 1]; |
200 |
|
|
p *= w - lsp[j]; |
201 |
|
|
} |
202 |
|
|
p *= p * (2.0f - w); |
203 |
|
|
q *= q * (2.0f + w); |
204 |
|
|
v = p + q; |
205 |
|
|
v = pow_m1_4(s, v); |
206 |
|
|
if (v > val_max) |
207 |
|
|
val_max = v; |
208 |
|
|
out[i] = v; |
209 |
|
|
} |
210 |
|
|
*val_max_ptr = val_max; |
211 |
|
|
} |
212 |
|
|
|
213 |
|
|
/** |
214 |
|
|
* decode exponents coded with LSP coefficients (same idea as Vorbis) |
215 |
|
|
*/ |
216 |
|
|
static void decode_exp_lsp(WMACodecContext *s, int ch) |
217 |
|
|
{ |
218 |
|
|
float lsp_coefs[NB_LSP_COEFS]; |
219 |
|
|
int val, i; |
220 |
|
|
|
221 |
|
|
for (i = 0; i < NB_LSP_COEFS; i++) { |
222 |
|
|
if (i == 0 || i >= 8) |
223 |
|
|
val = get_bits(&s->gb, 3); |
224 |
|
|
else |
225 |
|
|
val = get_bits(&s->gb, 4); |
226 |
|
|
lsp_coefs[i] = ff_wma_lsp_codebook[i][val]; |
227 |
|
|
} |
228 |
|
|
|
229 |
|
|
wma_lsp_to_curve(s, s->exponents[ch], &s->max_exponent[ch], |
230 |
|
|
s->block_len, lsp_coefs); |
231 |
|
|
} |
232 |
|
|
|
233 |
|
|
/** pow(10, i / 16.0) for i in -60..95 */ |
234 |
|
|
static const float pow_tab[] = { |
235 |
|
|
1.7782794100389e-04, 2.0535250264571e-04, |
236 |
|
|
2.3713737056617e-04, 2.7384196342644e-04, |
237 |
|
|
3.1622776601684e-04, 3.6517412725484e-04, |
238 |
|
|
4.2169650342858e-04, 4.8696752516586e-04, |
239 |
|
|
5.6234132519035e-04, 6.4938163157621e-04, |
240 |
|
|
7.4989420933246e-04, 8.6596432336006e-04, |
241 |
|
|
1.0000000000000e-03, 1.1547819846895e-03, |
242 |
|
|
1.3335214321633e-03, 1.5399265260595e-03, |
243 |
|
|
1.7782794100389e-03, 2.0535250264571e-03, |
244 |
|
|
2.3713737056617e-03, 2.7384196342644e-03, |
245 |
|
|
3.1622776601684e-03, 3.6517412725484e-03, |
246 |
|
|
4.2169650342858e-03, 4.8696752516586e-03, |
247 |
|
|
5.6234132519035e-03, 6.4938163157621e-03, |
248 |
|
|
7.4989420933246e-03, 8.6596432336006e-03, |
249 |
|
|
1.0000000000000e-02, 1.1547819846895e-02, |
250 |
|
|
1.3335214321633e-02, 1.5399265260595e-02, |
251 |
|
|
1.7782794100389e-02, 2.0535250264571e-02, |
252 |
|
|
2.3713737056617e-02, 2.7384196342644e-02, |
253 |
|
|
3.1622776601684e-02, 3.6517412725484e-02, |
254 |
|
|
4.2169650342858e-02, 4.8696752516586e-02, |
255 |
|
|
5.6234132519035e-02, 6.4938163157621e-02, |
256 |
|
|
7.4989420933246e-02, 8.6596432336007e-02, |
257 |
|
|
1.0000000000000e-01, 1.1547819846895e-01, |
258 |
|
|
1.3335214321633e-01, 1.5399265260595e-01, |
259 |
|
|
1.7782794100389e-01, 2.0535250264571e-01, |
260 |
|
|
2.3713737056617e-01, 2.7384196342644e-01, |
261 |
|
|
3.1622776601684e-01, 3.6517412725484e-01, |
262 |
|
|
4.2169650342858e-01, 4.8696752516586e-01, |
263 |
|
|
5.6234132519035e-01, 6.4938163157621e-01, |
264 |
|
|
7.4989420933246e-01, 8.6596432336007e-01, |
265 |
|
|
1.0000000000000e+00, 1.1547819846895e+00, |
266 |
|
|
1.3335214321633e+00, 1.5399265260595e+00, |
267 |
|
|
1.7782794100389e+00, 2.0535250264571e+00, |
268 |
|
|
2.3713737056617e+00, 2.7384196342644e+00, |
269 |
|
|
3.1622776601684e+00, 3.6517412725484e+00, |
270 |
|
|
4.2169650342858e+00, 4.8696752516586e+00, |
271 |
|
|
5.6234132519035e+00, 6.4938163157621e+00, |
272 |
|
|
7.4989420933246e+00, 8.6596432336007e+00, |
273 |
|
|
1.0000000000000e+01, 1.1547819846895e+01, |
274 |
|
|
1.3335214321633e+01, 1.5399265260595e+01, |
275 |
|
|
1.7782794100389e+01, 2.0535250264571e+01, |
276 |
|
|
2.3713737056617e+01, 2.7384196342644e+01, |
277 |
|
|
3.1622776601684e+01, 3.6517412725484e+01, |
278 |
|
|
4.2169650342858e+01, 4.8696752516586e+01, |
279 |
|
|
5.6234132519035e+01, 6.4938163157621e+01, |
280 |
|
|
7.4989420933246e+01, 8.6596432336007e+01, |
281 |
|
|
1.0000000000000e+02, 1.1547819846895e+02, |
282 |
|
|
1.3335214321633e+02, 1.5399265260595e+02, |
283 |
|
|
1.7782794100389e+02, 2.0535250264571e+02, |
284 |
|
|
2.3713737056617e+02, 2.7384196342644e+02, |
285 |
|
|
3.1622776601684e+02, 3.6517412725484e+02, |
286 |
|
|
4.2169650342858e+02, 4.8696752516586e+02, |
287 |
|
|
5.6234132519035e+02, 6.4938163157621e+02, |
288 |
|
|
7.4989420933246e+02, 8.6596432336007e+02, |
289 |
|
|
1.0000000000000e+03, 1.1547819846895e+03, |
290 |
|
|
1.3335214321633e+03, 1.5399265260595e+03, |
291 |
|
|
1.7782794100389e+03, 2.0535250264571e+03, |
292 |
|
|
2.3713737056617e+03, 2.7384196342644e+03, |
293 |
|
|
3.1622776601684e+03, 3.6517412725484e+03, |
294 |
|
|
4.2169650342858e+03, 4.8696752516586e+03, |
295 |
|
|
5.6234132519035e+03, 6.4938163157621e+03, |
296 |
|
|
7.4989420933246e+03, 8.6596432336007e+03, |
297 |
|
|
1.0000000000000e+04, 1.1547819846895e+04, |
298 |
|
|
1.3335214321633e+04, 1.5399265260595e+04, |
299 |
|
|
1.7782794100389e+04, 2.0535250264571e+04, |
300 |
|
|
2.3713737056617e+04, 2.7384196342644e+04, |
301 |
|
|
3.1622776601684e+04, 3.6517412725484e+04, |
302 |
|
|
4.2169650342858e+04, 4.8696752516586e+04, |
303 |
|
|
5.6234132519035e+04, 6.4938163157621e+04, |
304 |
|
|
7.4989420933246e+04, 8.6596432336007e+04, |
305 |
|
|
1.0000000000000e+05, 1.1547819846895e+05, |
306 |
|
|
1.3335214321633e+05, 1.5399265260595e+05, |
307 |
|
|
1.7782794100389e+05, 2.0535250264571e+05, |
308 |
|
|
2.3713737056617e+05, 2.7384196342644e+05, |
309 |
|
|
3.1622776601684e+05, 3.6517412725484e+05, |
310 |
|
|
4.2169650342858e+05, 4.8696752516586e+05, |
311 |
|
|
5.6234132519035e+05, 6.4938163157621e+05, |
312 |
|
|
7.4989420933246e+05, 8.6596432336007e+05, |
313 |
|
|
}; |
314 |
|
|
|
315 |
|
|
/** |
316 |
|
|
* decode exponents coded with VLC codes |
317 |
|
|
*/ |
318 |
|
1136 |
static int decode_exp_vlc(WMACodecContext *s, int ch) |
319 |
|
|
{ |
320 |
|
|
int last_exp, n, code; |
321 |
|
|
const uint16_t *ptr; |
322 |
|
|
float v, max_scale; |
323 |
|
|
uint32_t *q, *q_end, iv; |
324 |
|
1136 |
const float *ptab = pow_tab + 60; |
325 |
|
1136 |
const uint32_t *iptab = (const uint32_t *) ptab; |
326 |
|
|
|
327 |
|
1136 |
ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits]; |
328 |
|
1136 |
q = (uint32_t *) s->exponents[ch]; |
329 |
|
1136 |
q_end = q + s->block_len; |
330 |
|
1136 |
max_scale = 0; |
331 |
✓✓ |
1136 |
if (s->version == 1) { |
332 |
|
410 |
last_exp = get_bits(&s->gb, 5) + 10; |
333 |
|
410 |
v = ptab[last_exp]; |
334 |
|
410 |
iv = iptab[last_exp]; |
335 |
|
410 |
max_scale = v; |
336 |
|
410 |
n = *ptr++; |
337 |
✗✗✗✓
✗ |
410 |
switch (n & 3) do { |
338 |
|
820 |
case 0: *q++ = iv; |
339 |
|
820 |
case 3: *q++ = iv; |
340 |
|
820 |
case 2: *q++ = iv; |
341 |
|
1230 |
case 1: *q++ = iv; |
342 |
✓✓ |
1230 |
} while ((n -= 4) > 0); |
343 |
|
|
} else |
344 |
|
726 |
last_exp = 36; |
345 |
|
|
|
346 |
✓✓ |
29126 |
while (q < q_end) { |
347 |
|
27990 |
code = get_vlc2(&s->gb, s->exp_vlc.table, EXPVLCBITS, EXPMAX); |
348 |
|
|
/* NOTE: this offset is the same as MPEG-4 AAC! */ |
349 |
|
27990 |
last_exp += code - 60; |
350 |
✗✓ |
27990 |
if ((unsigned) last_exp + 60 >= FF_ARRAY_ELEMS(pow_tab)) { |
351 |
|
|
av_log(s->avctx, AV_LOG_ERROR, "Exponent out of range: %d\n", |
352 |
|
|
last_exp); |
353 |
|
|
return -1; |
354 |
|
|
} |
355 |
|
27990 |
v = ptab[last_exp]; |
356 |
|
27990 |
iv = iptab[last_exp]; |
357 |
✓✓ |
27990 |
if (v > max_scale) |
358 |
|
1141 |
max_scale = v; |
359 |
|
27990 |
n = *ptr++; |
360 |
✓✓✓✓
✗ |
27990 |
switch (n & 3) do { |
361 |
|
576712 |
case 0: *q++ = iv; |
362 |
|
579172 |
case 3: *q++ = iv; |
363 |
|
581632 |
case 2: *q++ = iv; |
364 |
|
585322 |
case 1: *q++ = iv; |
365 |
✓✓ |
585322 |
} while ((n -= 4) > 0); |
366 |
|
|
} |
367 |
|
1136 |
s->max_exponent[ch] = max_scale; |
368 |
|
1136 |
return 0; |
369 |
|
|
} |
370 |
|
|
|
371 |
|
|
/** |
372 |
|
|
* Apply MDCT window and add into output. |
373 |
|
|
* |
374 |
|
|
* We ensure that when the windows overlap their squared sum |
375 |
|
|
* is always 1 (MDCT reconstruction rule). |
376 |
|
|
*/ |
377 |
|
1172 |
static void wma_window(WMACodecContext *s, float *out) |
378 |
|
|
{ |
379 |
|
1172 |
float *in = s->output; |
380 |
|
|
int block_len, bsize, n; |
381 |
|
|
|
382 |
|
|
/* left part */ |
383 |
✓✗ |
1172 |
if (s->block_len_bits <= s->prev_block_len_bits) { |
384 |
|
1172 |
block_len = s->block_len; |
385 |
|
1172 |
bsize = s->frame_len_bits - s->block_len_bits; |
386 |
|
|
|
387 |
|
1172 |
s->fdsp->vector_fmul_add(out, in, s->windows[bsize], |
388 |
|
|
out, block_len); |
389 |
|
|
} else { |
390 |
|
|
block_len = 1 << s->prev_block_len_bits; |
391 |
|
|
n = (s->block_len - block_len) / 2; |
392 |
|
|
bsize = s->frame_len_bits - s->prev_block_len_bits; |
393 |
|
|
|
394 |
|
|
s->fdsp->vector_fmul_add(out + n, in + n, s->windows[bsize], |
395 |
|
|
out + n, block_len); |
396 |
|
|
|
397 |
|
|
memcpy(out + n + block_len, in + n + block_len, n * sizeof(float)); |
398 |
|
|
} |
399 |
|
|
|
400 |
|
1172 |
out += s->block_len; |
401 |
|
1172 |
in += s->block_len; |
402 |
|
|
|
403 |
|
|
/* right part */ |
404 |
✓✗ |
1172 |
if (s->block_len_bits <= s->next_block_len_bits) { |
405 |
|
1172 |
block_len = s->block_len; |
406 |
|
1172 |
bsize = s->frame_len_bits - s->block_len_bits; |
407 |
|
|
|
408 |
|
1172 |
s->fdsp->vector_fmul_reverse(out, in, s->windows[bsize], block_len); |
409 |
|
|
} else { |
410 |
|
|
block_len = 1 << s->next_block_len_bits; |
411 |
|
|
n = (s->block_len - block_len) / 2; |
412 |
|
|
bsize = s->frame_len_bits - s->next_block_len_bits; |
413 |
|
|
|
414 |
|
|
memcpy(out, in, n * sizeof(float)); |
415 |
|
|
|
416 |
|
|
s->fdsp->vector_fmul_reverse(out + n, in + n, s->windows[bsize], |
417 |
|
|
block_len); |
418 |
|
|
|
419 |
|
|
memset(out + n + block_len, 0, n * sizeof(float)); |
420 |
|
|
} |
421 |
|
1172 |
} |
422 |
|
|
|
423 |
|
|
/** |
424 |
|
|
* @return 0 if OK. 1 if last block of frame. return -1 if |
425 |
|
|
* unrecoverable error. |
426 |
|
|
*/ |
427 |
|
586 |
static int wma_decode_block(WMACodecContext *s) |
428 |
|
|
{ |
429 |
|
|
int n, v, a, ch, bsize; |
430 |
|
|
int coef_nb_bits, total_gain; |
431 |
|
|
int nb_coefs[MAX_CHANNELS]; |
432 |
|
|
float mdct_norm; |
433 |
|
|
FFTContext *mdct; |
434 |
|
|
|
435 |
|
|
#ifdef TRACE |
436 |
|
|
ff_tlog(s->avctx, "***decode_block: %d:%d\n", |
437 |
|
|
s->frame_count - 1, s->block_num); |
438 |
|
|
#endif /* TRACE */ |
439 |
|
|
|
440 |
|
|
/* compute current block length */ |
441 |
✗✓ |
586 |
if (s->use_variable_block_len) { |
442 |
|
|
n = av_log2(s->nb_block_sizes - 1) + 1; |
443 |
|
|
|
444 |
|
|
if (s->reset_block_lengths) { |
445 |
|
|
s->reset_block_lengths = 0; |
446 |
|
|
v = get_bits(&s->gb, n); |
447 |
|
|
if (v >= s->nb_block_sizes) { |
448 |
|
|
av_log(s->avctx, AV_LOG_ERROR, |
449 |
|
|
"prev_block_len_bits %d out of range\n", |
450 |
|
|
s->frame_len_bits - v); |
451 |
|
|
return -1; |
452 |
|
|
} |
453 |
|
|
s->prev_block_len_bits = s->frame_len_bits - v; |
454 |
|
|
v = get_bits(&s->gb, n); |
455 |
|
|
if (v >= s->nb_block_sizes) { |
456 |
|
|
av_log(s->avctx, AV_LOG_ERROR, |
457 |
|
|
"block_len_bits %d out of range\n", |
458 |
|
|
s->frame_len_bits - v); |
459 |
|
|
return -1; |
460 |
|
|
} |
461 |
|
|
s->block_len_bits = s->frame_len_bits - v; |
462 |
|
|
} else { |
463 |
|
|
/* update block lengths */ |
464 |
|
|
s->prev_block_len_bits = s->block_len_bits; |
465 |
|
|
s->block_len_bits = s->next_block_len_bits; |
466 |
|
|
} |
467 |
|
|
v = get_bits(&s->gb, n); |
468 |
|
|
if (v >= s->nb_block_sizes) { |
469 |
|
|
av_log(s->avctx, AV_LOG_ERROR, |
470 |
|
|
"next_block_len_bits %d out of range\n", |
471 |
|
|
s->frame_len_bits - v); |
472 |
|
|
return -1; |
473 |
|
|
} |
474 |
|
|
s->next_block_len_bits = s->frame_len_bits - v; |
475 |
|
|
} else { |
476 |
|
|
/* fixed block len */ |
477 |
|
586 |
s->next_block_len_bits = s->frame_len_bits; |
478 |
|
586 |
s->prev_block_len_bits = s->frame_len_bits; |
479 |
|
586 |
s->block_len_bits = s->frame_len_bits; |
480 |
|
|
} |
481 |
|
|
|
482 |
✗✓ |
586 |
if (s->frame_len_bits - s->block_len_bits >= s->nb_block_sizes){ |
483 |
|
|
av_log(s->avctx, AV_LOG_ERROR, "block_len_bits not initialized to a valid value\n"); |
484 |
|
|
return -1; |
485 |
|
|
} |
486 |
|
|
|
487 |
|
|
/* now check if the block length is coherent with the frame length */ |
488 |
|
586 |
s->block_len = 1 << s->block_len_bits; |
489 |
✗✓ |
586 |
if ((s->block_pos + s->block_len) > s->frame_len) { |
490 |
|
|
av_log(s->avctx, AV_LOG_ERROR, "frame_len overflow\n"); |
491 |
|
|
return -1; |
492 |
|
|
} |
493 |
|
|
|
494 |
✓✗ |
586 |
if (s->avctx->channels == 2) |
495 |
|
586 |
s->ms_stereo = get_bits1(&s->gb); |
496 |
|
586 |
v = 0; |
497 |
✓✓ |
1758 |
for (ch = 0; ch < s->avctx->channels; ch++) { |
498 |
|
1172 |
a = get_bits1(&s->gb); |
499 |
|
1172 |
s->channel_coded[ch] = a; |
500 |
|
1172 |
v |= a; |
501 |
|
|
} |
502 |
|
|
|
503 |
|
586 |
bsize = s->frame_len_bits - s->block_len_bits; |
504 |
|
|
|
505 |
|
|
/* if no channel coded, no need to go further */ |
506 |
|
|
/* XXX: fix potential framing problems */ |
507 |
✗✓ |
586 |
if (!v) |
508 |
|
|
goto next; |
509 |
|
|
|
510 |
|
|
/* read total gain and extract corresponding number of bits for |
511 |
|
|
* coef escape coding */ |
512 |
|
586 |
total_gain = 1; |
513 |
|
|
for (;;) { |
514 |
✗✓ |
586 |
if (get_bits_left(&s->gb) < 7) { |
515 |
|
|
av_log(s->avctx, AV_LOG_ERROR, "total_gain overread\n"); |
516 |
|
|
return AVERROR_INVALIDDATA; |
517 |
|
|
} |
518 |
|
586 |
a = get_bits(&s->gb, 7); |
519 |
|
586 |
total_gain += a; |
520 |
✓✗ |
586 |
if (a != 127) |
521 |
|
586 |
break; |
522 |
|
|
} |
523 |
|
|
|
524 |
|
586 |
coef_nb_bits = ff_wma_total_gain_to_bits(total_gain); |
525 |
|
|
|
526 |
|
|
/* compute number of coefficients */ |
527 |
|
586 |
n = s->coefs_end[bsize] - s->coefs_start; |
528 |
✓✓ |
1758 |
for (ch = 0; ch < s->avctx->channels; ch++) |
529 |
|
1172 |
nb_coefs[ch] = n; |
530 |
|
|
|
531 |
|
|
/* complex coding */ |
532 |
✗✓ |
586 |
if (s->use_noise_coding) { |
533 |
|
|
for (ch = 0; ch < s->avctx->channels; ch++) { |
534 |
|
|
if (s->channel_coded[ch]) { |
535 |
|
|
int i, n, a; |
536 |
|
|
n = s->exponent_high_sizes[bsize]; |
537 |
|
|
for (i = 0; i < n; i++) { |
538 |
|
|
a = get_bits1(&s->gb); |
539 |
|
|
s->high_band_coded[ch][i] = a; |
540 |
|
|
/* if noise coding, the coefficients are not transmitted */ |
541 |
|
|
if (a) |
542 |
|
|
nb_coefs[ch] -= s->exponent_high_bands[bsize][i]; |
543 |
|
|
} |
544 |
|
|
} |
545 |
|
|
} |
546 |
|
|
for (ch = 0; ch < s->avctx->channels; ch++) { |
547 |
|
|
if (s->channel_coded[ch]) { |
548 |
|
|
int i, n, val; |
549 |
|
|
|
550 |
|
|
n = s->exponent_high_sizes[bsize]; |
551 |
|
|
val = (int) 0x80000000; |
552 |
|
|
for (i = 0; i < n; i++) { |
553 |
|
|
if (s->high_band_coded[ch][i]) { |
554 |
|
|
if (val == (int) 0x80000000) { |
555 |
|
|
val = get_bits(&s->gb, 7) - 19; |
556 |
|
|
} else { |
557 |
|
|
val += get_vlc2(&s->gb, s->hgain_vlc.table, |
558 |
|
|
HGAINVLCBITS, HGAINMAX); |
559 |
|
|
} |
560 |
|
|
s->high_band_values[ch][i] = val; |
561 |
|
|
} |
562 |
|
|
} |
563 |
|
|
} |
564 |
|
|
} |
565 |
|
|
} |
566 |
|
|
|
567 |
|
|
/* exponents can be reused in short blocks. */ |
568 |
✗✓✗✗
|
586 |
if ((s->block_len_bits == s->frame_len_bits) || get_bits1(&s->gb)) { |
569 |
✓✓ |
1758 |
for (ch = 0; ch < s->avctx->channels; ch++) { |
570 |
✓✓ |
1172 |
if (s->channel_coded[ch]) { |
571 |
✓✗ |
1136 |
if (s->use_exp_vlc) { |
572 |
✗✓ |
1136 |
if (decode_exp_vlc(s, ch) < 0) |
573 |
|
|
return -1; |
574 |
|
|
} else { |
575 |
|
|
decode_exp_lsp(s, ch); |
576 |
|
|
} |
577 |
|
1136 |
s->exponents_bsize[ch] = bsize; |
578 |
|
1136 |
s->exponents_initialized[ch] = 1; |
579 |
|
|
} |
580 |
|
|
} |
581 |
|
|
} |
582 |
|
|
|
583 |
✓✓ |
1758 |
for (ch = 0; ch < s->avctx->channels; ch++) { |
584 |
✓✓✗✓
|
1172 |
if (s->channel_coded[ch] && !s->exponents_initialized[ch]) |
585 |
|
|
return AVERROR_INVALIDDATA; |
586 |
|
|
} |
587 |
|
|
|
588 |
|
|
/* parse spectral coefficients : just RLE encoding */ |
589 |
✓✓ |
1758 |
for (ch = 0; ch < s->avctx->channels; ch++) { |
590 |
✓✓ |
1172 |
if (s->channel_coded[ch]) { |
591 |
|
|
int tindex; |
592 |
|
1136 |
WMACoef *ptr = &s->coefs1[ch][0]; |
593 |
|
|
|
594 |
|
|
/* special VLC tables are used for ms stereo because |
595 |
|
|
* there is potentially less energy there */ |
596 |
✓✓✓✓
|
1136 |
tindex = (ch == 1 && s->ms_stereo); |
597 |
|
1136 |
memset(ptr, 0, s->block_len * sizeof(WMACoef)); |
598 |
|
1136 |
ff_wma_run_level_decode(s->avctx, &s->gb, &s->coef_vlc[tindex], |
599 |
|
1136 |
s->level_table[tindex], s->run_table[tindex], |
600 |
|
|
0, ptr, 0, nb_coefs[ch], |
601 |
|
|
s->block_len, s->frame_len_bits, coef_nb_bits); |
602 |
|
|
} |
603 |
✓✓✓✗
|
1172 |
if (s->version == 1 && s->avctx->channels >= 2) |
604 |
|
410 |
align_get_bits(&s->gb); |
605 |
|
|
} |
606 |
|
|
|
607 |
|
|
/* normalize */ |
608 |
|
|
{ |
609 |
|
586 |
int n4 = s->block_len / 2; |
610 |
|
586 |
mdct_norm = 1.0 / (float) n4; |
611 |
✓✓ |
586 |
if (s->version == 1) |
612 |
|
205 |
mdct_norm *= sqrt(n4); |
613 |
|
|
} |
614 |
|
|
|
615 |
|
|
/* finally compute the MDCT coefficients */ |
616 |
✓✓ |
1758 |
for (ch = 0; ch < s->avctx->channels; ch++) { |
617 |
✓✓ |
1172 |
if (s->channel_coded[ch]) { |
618 |
|
|
WMACoef *coefs1; |
619 |
|
|
float *coefs, *exponents, mult, mult1, noise; |
620 |
|
|
int i, j, n, n1, last_high_band, esize; |
621 |
|
|
float exp_power[HIGH_BAND_MAX_SIZE]; |
622 |
|
|
|
623 |
|
1136 |
coefs1 = s->coefs1[ch]; |
624 |
|
1136 |
exponents = s->exponents[ch]; |
625 |
|
1136 |
esize = s->exponents_bsize[ch]; |
626 |
|
1136 |
mult = ff_exp10(total_gain * 0.05) / s->max_exponent[ch]; |
627 |
|
1136 |
mult *= mdct_norm; |
628 |
|
1136 |
coefs = s->coefs[ch]; |
629 |
✗✓ |
1136 |
if (s->use_noise_coding) { |
630 |
|
|
mult1 = mult; |
631 |
|
|
/* very low freqs : noise */ |
632 |
|
|
for (i = 0; i < s->coefs_start; i++) { |
633 |
|
|
*coefs++ = s->noise_table[s->noise_index] * |
634 |
|
|
exponents[i << bsize >> esize] * mult1; |
635 |
|
|
s->noise_index = (s->noise_index + 1) & |
636 |
|
|
(NOISE_TAB_SIZE - 1); |
637 |
|
|
} |
638 |
|
|
|
639 |
|
|
n1 = s->exponent_high_sizes[bsize]; |
640 |
|
|
|
641 |
|
|
/* compute power of high bands */ |
642 |
|
|
exponents = s->exponents[ch] + |
643 |
|
|
(s->high_band_start[bsize] << bsize >> esize); |
644 |
|
|
last_high_band = 0; /* avoid warning */ |
645 |
|
|
for (j = 0; j < n1; j++) { |
646 |
|
|
n = s->exponent_high_bands[s->frame_len_bits - |
647 |
|
|
s->block_len_bits][j]; |
648 |
|
|
if (s->high_band_coded[ch][j]) { |
649 |
|
|
float e2, v; |
650 |
|
|
e2 = 0; |
651 |
|
|
for (i = 0; i < n; i++) { |
652 |
|
|
v = exponents[i << bsize >> esize]; |
653 |
|
|
e2 += v * v; |
654 |
|
|
} |
655 |
|
|
exp_power[j] = e2 / n; |
656 |
|
|
last_high_band = j; |
657 |
|
|
ff_tlog(s->avctx, "%d: power=%f (%d)\n", j, exp_power[j], n); |
658 |
|
|
} |
659 |
|
|
exponents += n << bsize >> esize; |
660 |
|
|
} |
661 |
|
|
|
662 |
|
|
/* main freqs and high freqs */ |
663 |
|
|
exponents = s->exponents[ch] + (s->coefs_start << bsize >> esize); |
664 |
|
|
for (j = -1; j < n1; j++) { |
665 |
|
|
if (j < 0) |
666 |
|
|
n = s->high_band_start[bsize] - s->coefs_start; |
667 |
|
|
else |
668 |
|
|
n = s->exponent_high_bands[s->frame_len_bits - |
669 |
|
|
s->block_len_bits][j]; |
670 |
|
|
if (j >= 0 && s->high_band_coded[ch][j]) { |
671 |
|
|
/* use noise with specified power */ |
672 |
|
|
mult1 = sqrt(exp_power[j] / exp_power[last_high_band]); |
673 |
|
|
/* XXX: use a table */ |
674 |
|
|
mult1 = mult1 * ff_exp10(s->high_band_values[ch][j] * 0.05); |
675 |
|
|
mult1 = mult1 / (s->max_exponent[ch] * s->noise_mult); |
676 |
|
|
mult1 *= mdct_norm; |
677 |
|
|
for (i = 0; i < n; i++) { |
678 |
|
|
noise = s->noise_table[s->noise_index]; |
679 |
|
|
s->noise_index = (s->noise_index + 1) & (NOISE_TAB_SIZE - 1); |
680 |
|
|
*coefs++ = noise * exponents[i << bsize >> esize] * mult1; |
681 |
|
|
} |
682 |
|
|
exponents += n << bsize >> esize; |
683 |
|
|
} else { |
684 |
|
|
/* coded values + small noise */ |
685 |
|
|
for (i = 0; i < n; i++) { |
686 |
|
|
noise = s->noise_table[s->noise_index]; |
687 |
|
|
s->noise_index = (s->noise_index + 1) & (NOISE_TAB_SIZE - 1); |
688 |
|
|
*coefs++ = ((*coefs1++) + noise) * |
689 |
|
|
exponents[i << bsize >> esize] * mult; |
690 |
|
|
} |
691 |
|
|
exponents += n << bsize >> esize; |
692 |
|
|
} |
693 |
|
|
} |
694 |
|
|
|
695 |
|
|
/* very high freqs : noise */ |
696 |
|
|
n = s->block_len - s->coefs_end[bsize]; |
697 |
|
|
mult1 = mult * exponents[(-(1 << bsize)) >> esize]; |
698 |
|
|
for (i = 0; i < n; i++) { |
699 |
|
|
*coefs++ = s->noise_table[s->noise_index] * mult1; |
700 |
|
|
s->noise_index = (s->noise_index + 1) & (NOISE_TAB_SIZE - 1); |
701 |
|
|
} |
702 |
|
|
} else { |
703 |
|
|
/* XXX: optimize more */ |
704 |
✓✓ |
2366 |
for (i = 0; i < s->coefs_start; i++) |
705 |
|
1230 |
*coefs++ = 0.0; |
706 |
|
1136 |
n = nb_coefs[ch]; |
707 |
✓✓ |
2117410 |
for (i = 0; i < n; i++) |
708 |
|
2116274 |
*coefs++ = coefs1[i] * exponents[i << bsize >> esize] * mult; |
709 |
|
1136 |
n = s->block_len - s->coefs_end[bsize]; |
710 |
✓✓ |
210160 |
for (i = 0; i < n; i++) |
711 |
|
209024 |
*coefs++ = 0.0; |
712 |
|
|
} |
713 |
|
|
} |
714 |
|
|
} |
715 |
|
|
|
716 |
|
|
#ifdef TRACE |
717 |
|
|
for (ch = 0; ch < s->avctx->channels; ch++) { |
718 |
|
|
if (s->channel_coded[ch]) { |
719 |
|
|
dump_floats(s, "exponents", 3, s->exponents[ch], s->block_len); |
720 |
|
|
dump_floats(s, "coefs", 1, s->coefs[ch], s->block_len); |
721 |
|
|
} |
722 |
|
|
} |
723 |
|
|
#endif /* TRACE */ |
724 |
|
|
|
725 |
✓✓✓✓
|
586 |
if (s->ms_stereo && s->channel_coded[1]) { |
726 |
|
|
/* nominal case for ms stereo: we do it before mdct */ |
727 |
|
|
/* no need to optimize this case because it should almost |
728 |
|
|
* never happen */ |
729 |
✗✓ |
480 |
if (!s->channel_coded[0]) { |
730 |
|
|
ff_tlog(s->avctx, "rare ms-stereo case happened\n"); |
731 |
|
|
memset(s->coefs[0], 0, sizeof(float) * s->block_len); |
732 |
|
|
s->channel_coded[0] = 1; |
733 |
|
|
} |
734 |
|
|
|
735 |
|
480 |
s->fdsp->butterflies_float(s->coefs[0], s->coefs[1], s->block_len); |
736 |
|
|
} |
737 |
|
|
|
738 |
|
106 |
next: |
739 |
|
586 |
mdct = &s->mdct_ctx[bsize]; |
740 |
|
|
|
741 |
✓✓ |
1758 |
for (ch = 0; ch < s->avctx->channels; ch++) { |
742 |
|
|
int n4, index; |
743 |
|
|
|
744 |
|
1172 |
n4 = s->block_len / 2; |
745 |
✓✓ |
1172 |
if (s->channel_coded[ch]) |
746 |
|
1136 |
mdct->imdct_calc(mdct, s->output, s->coefs[ch]); |
747 |
✓✓✗✓
|
36 |
else if (!(s->ms_stereo && ch == 1)) |
748 |
|
35 |
memset(s->output, 0, sizeof(s->output)); |
749 |
|
|
|
750 |
|
|
/* multiply by the window and add in the frame */ |
751 |
|
1172 |
index = (s->frame_len / 2) + s->block_pos - n4; |
752 |
|
1172 |
wma_window(s, &s->frame_out[ch][index]); |
753 |
|
|
} |
754 |
|
|
|
755 |
|
|
/* update block number */ |
756 |
|
586 |
s->block_num++; |
757 |
|
586 |
s->block_pos += s->block_len; |
758 |
✓✗ |
586 |
if (s->block_pos >= s->frame_len) |
759 |
|
586 |
return 1; |
760 |
|
|
else |
761 |
|
|
return 0; |
762 |
|
|
} |
763 |
|
|
|
764 |
|
|
/* decode a frame of frame_len samples */ |
765 |
|
586 |
static int wma_decode_frame(WMACodecContext *s, float **samples, |
766 |
|
|
int samples_offset) |
767 |
|
|
{ |
768 |
|
|
int ret, ch; |
769 |
|
|
|
770 |
|
|
#ifdef TRACE |
771 |
|
|
ff_tlog(s->avctx, "***decode_frame: %d size=%d\n", |
772 |
|
|
s->frame_count++, s->frame_len); |
773 |
|
|
#endif /* TRACE */ |
774 |
|
|
|
775 |
|
|
/* read each block */ |
776 |
|
586 |
s->block_num = 0; |
777 |
|
586 |
s->block_pos = 0; |
778 |
|
|
for (;;) { |
779 |
|
586 |
ret = wma_decode_block(s); |
780 |
✗✓ |
586 |
if (ret < 0) |
781 |
|
|
return -1; |
782 |
✓✗ |
586 |
if (ret) |
783 |
|
586 |
break; |
784 |
|
|
} |
785 |
|
|
|
786 |
✓✓ |
1758 |
for (ch = 0; ch < s->avctx->channels; ch++) { |
787 |
|
|
/* copy current block to output */ |
788 |
|
1172 |
memcpy(samples[ch] + samples_offset, s->frame_out[ch], |
789 |
|
1172 |
s->frame_len * sizeof(*s->frame_out[ch])); |
790 |
|
|
/* prepare for next block */ |
791 |
|
1172 |
memmove(&s->frame_out[ch][0], &s->frame_out[ch][s->frame_len], |
792 |
|
1172 |
s->frame_len * sizeof(*s->frame_out[ch])); |
793 |
|
|
|
794 |
|
|
#ifdef TRACE |
795 |
|
|
dump_floats(s, "samples", 6, samples[ch] + samples_offset, |
796 |
|
|
s->frame_len); |
797 |
|
|
#endif /* TRACE */ |
798 |
|
|
} |
799 |
|
|
|
800 |
|
586 |
return 0; |
801 |
|
|
} |
802 |
|
|
|
803 |
|
587 |
static int wma_decode_superframe(AVCodecContext *avctx, void *data, |
804 |
|
|
int *got_frame_ptr, AVPacket *avpkt) |
805 |
|
|
{ |
806 |
|
587 |
AVFrame *frame = data; |
807 |
|
587 |
const uint8_t *buf = avpkt->data; |
808 |
|
587 |
int buf_size = avpkt->size; |
809 |
|
587 |
WMACodecContext *s = avctx->priv_data; |
810 |
|
|
int nb_frames, bit_offset, i, pos, len, ret; |
811 |
|
|
uint8_t *q; |
812 |
|
|
float **samples; |
813 |
|
|
int samples_offset; |
814 |
|
|
|
815 |
|
|
ff_tlog(avctx, "***decode_superframe:\n"); |
816 |
|
|
|
817 |
✗✓ |
587 |
if (buf_size == 0) { |
818 |
|
|
s->last_superframe_len = 0; |
819 |
|
|
return 0; |
820 |
|
|
} |
821 |
✓✓ |
587 |
if (buf_size < avctx->block_align) { |
822 |
|
1 |
av_log(avctx, AV_LOG_ERROR, |
823 |
|
|
"Input packet size too small (%d < %d)\n", |
824 |
|
|
buf_size, avctx->block_align); |
825 |
|
1 |
return AVERROR_INVALIDDATA; |
826 |
|
|
} |
827 |
✓✗ |
586 |
if (avctx->block_align) |
828 |
|
586 |
buf_size = avctx->block_align; |
829 |
|
|
|
830 |
|
586 |
init_get_bits(&s->gb, buf, buf_size * 8); |
831 |
|
|
|
832 |
✗✓ |
586 |
if (s->use_bit_reservoir) { |
833 |
|
|
/* read super frame header */ |
834 |
|
|
skip_bits(&s->gb, 4); /* super frame index */ |
835 |
|
|
nb_frames = get_bits(&s->gb, 4) - (s->last_superframe_len <= 0); |
836 |
|
|
if (nb_frames <= 0) { |
837 |
|
|
int is_error = nb_frames < 0 || get_bits_left(&s->gb) <= 8; |
838 |
|
|
av_log(avctx, is_error ? AV_LOG_ERROR : AV_LOG_WARNING, |
839 |
|
|
"nb_frames is %d bits left %d\n", |
840 |
|
|
nb_frames, get_bits_left(&s->gb)); |
841 |
|
|
if (is_error) |
842 |
|
|
return AVERROR_INVALIDDATA; |
843 |
|
|
|
844 |
|
|
if ((s->last_superframe_len + buf_size - 1) > |
845 |
|
|
MAX_CODED_SUPERFRAME_SIZE) |
846 |
|
|
goto fail; |
847 |
|
|
|
848 |
|
|
q = s->last_superframe + s->last_superframe_len; |
849 |
|
|
len = buf_size - 1; |
850 |
|
|
while (len > 0) { |
851 |
|
|
*q++ = get_bits (&s->gb, 8); |
852 |
|
|
len --; |
853 |
|
|
} |
854 |
|
|
memset(q, 0, AV_INPUT_BUFFER_PADDING_SIZE); |
855 |
|
|
|
856 |
|
|
s->last_superframe_len += 8*buf_size - 8; |
857 |
|
|
// s->reset_block_lengths = 1; //XXX is this needed ? |
858 |
|
|
*got_frame_ptr = 0; |
859 |
|
|
return buf_size; |
860 |
|
|
} |
861 |
|
|
} else |
862 |
|
586 |
nb_frames = 1; |
863 |
|
|
|
864 |
|
|
/* get output buffer */ |
865 |
|
586 |
frame->nb_samples = nb_frames * s->frame_len; |
866 |
✗✓ |
586 |
if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) |
867 |
|
|
return ret; |
868 |
|
586 |
samples = (float **) frame->extended_data; |
869 |
|
586 |
samples_offset = 0; |
870 |
|
|
|
871 |
✗✓ |
586 |
if (s->use_bit_reservoir) { |
872 |
|
|
bit_offset = get_bits(&s->gb, s->byte_offset_bits + 3); |
873 |
|
|
if (bit_offset > get_bits_left(&s->gb)) { |
874 |
|
|
av_log(avctx, AV_LOG_ERROR, |
875 |
|
|
"Invalid last frame bit offset %d > buf size %d (%d)\n", |
876 |
|
|
bit_offset, get_bits_left(&s->gb), buf_size); |
877 |
|
|
goto fail; |
878 |
|
|
} |
879 |
|
|
|
880 |
|
|
if (s->last_superframe_len > 0) { |
881 |
|
|
/* add bit_offset bits to last frame */ |
882 |
|
|
if ((s->last_superframe_len + ((bit_offset + 7) >> 3)) > |
883 |
|
|
MAX_CODED_SUPERFRAME_SIZE) |
884 |
|
|
goto fail; |
885 |
|
|
q = s->last_superframe + s->last_superframe_len; |
886 |
|
|
len = bit_offset; |
887 |
|
|
while (len > 7) { |
888 |
|
|
*q++ = get_bits(&s->gb, 8); |
889 |
|
|
len -= 8; |
890 |
|
|
} |
891 |
|
|
if (len > 0) |
892 |
|
|
*q++ = get_bits(&s->gb, len) << (8 - len); |
893 |
|
|
memset(q, 0, AV_INPUT_BUFFER_PADDING_SIZE); |
894 |
|
|
|
895 |
|
|
/* XXX: bit_offset bits into last frame */ |
896 |
|
|
init_get_bits(&s->gb, s->last_superframe, |
897 |
|
|
s->last_superframe_len * 8 + bit_offset); |
898 |
|
|
/* skip unused bits */ |
899 |
|
|
if (s->last_bitoffset > 0) |
900 |
|
|
skip_bits(&s->gb, s->last_bitoffset); |
901 |
|
|
/* this frame is stored in the last superframe and in the |
902 |
|
|
* current one */ |
903 |
|
|
if (wma_decode_frame(s, samples, samples_offset) < 0) |
904 |
|
|
goto fail; |
905 |
|
|
samples_offset += s->frame_len; |
906 |
|
|
nb_frames--; |
907 |
|
|
} |
908 |
|
|
|
909 |
|
|
/* read each frame starting from bit_offset */ |
910 |
|
|
pos = bit_offset + 4 + 4 + s->byte_offset_bits + 3; |
911 |
|
|
if (pos >= MAX_CODED_SUPERFRAME_SIZE * 8 || pos > buf_size * 8) |
912 |
|
|
return AVERROR_INVALIDDATA; |
913 |
|
|
init_get_bits(&s->gb, buf + (pos >> 3), (buf_size - (pos >> 3)) * 8); |
914 |
|
|
len = pos & 7; |
915 |
|
|
if (len > 0) |
916 |
|
|
skip_bits(&s->gb, len); |
917 |
|
|
|
918 |
|
|
s->reset_block_lengths = 1; |
919 |
|
|
for (i = 0; i < nb_frames; i++) { |
920 |
|
|
if (wma_decode_frame(s, samples, samples_offset) < 0) |
921 |
|
|
goto fail; |
922 |
|
|
samples_offset += s->frame_len; |
923 |
|
|
} |
924 |
|
|
|
925 |
|
|
/* we copy the end of the frame in the last frame buffer */ |
926 |
|
|
pos = get_bits_count(&s->gb) + |
927 |
|
|
((bit_offset + 4 + 4 + s->byte_offset_bits + 3) & ~7); |
928 |
|
|
s->last_bitoffset = pos & 7; |
929 |
|
|
pos >>= 3; |
930 |
|
|
len = buf_size - pos; |
931 |
|
|
if (len > MAX_CODED_SUPERFRAME_SIZE || len < 0) { |
932 |
|
|
av_log(s->avctx, AV_LOG_ERROR, "len %d invalid\n", len); |
933 |
|
|
goto fail; |
934 |
|
|
} |
935 |
|
|
s->last_superframe_len = len; |
936 |
|
|
memcpy(s->last_superframe, buf + pos, len); |
937 |
|
|
} else { |
938 |
|
|
/* single frame decode */ |
939 |
✗✓ |
586 |
if (wma_decode_frame(s, samples, samples_offset) < 0) |
940 |
|
|
goto fail; |
941 |
|
586 |
samples_offset += s->frame_len; |
942 |
|
|
} |
943 |
|
|
|
944 |
|
|
ff_dlog(s->avctx, "%d %d %d %d outbytes:%"PTRDIFF_SPECIFIER" eaten:%d\n", |
945 |
|
|
s->frame_len_bits, s->block_len_bits, s->frame_len, s->block_len, |
946 |
|
|
(int8_t *) samples - (int8_t *) data, avctx->block_align); |
947 |
|
|
|
948 |
|
586 |
*got_frame_ptr = 1; |
949 |
|
|
|
950 |
|
586 |
return buf_size; |
951 |
|
|
|
952 |
|
|
fail: |
953 |
|
|
/* when error, we reset the bit reservoir */ |
954 |
|
|
s->last_superframe_len = 0; |
955 |
|
|
return -1; |
956 |
|
|
} |
957 |
|
|
|
958 |
|
|
static av_cold void flush(AVCodecContext *avctx) |
959 |
|
|
{ |
960 |
|
|
WMACodecContext *s = avctx->priv_data; |
961 |
|
|
|
962 |
|
|
s->last_bitoffset = |
963 |
|
|
s->last_superframe_len = 0; |
964 |
|
|
} |
965 |
|
|
|
966 |
|
|
#if CONFIG_WMAV1_DECODER |
967 |
|
|
AVCodec ff_wmav1_decoder = { |
968 |
|
|
.name = "wmav1", |
969 |
|
|
.long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 1"), |
970 |
|
|
.type = AVMEDIA_TYPE_AUDIO, |
971 |
|
|
.id = AV_CODEC_ID_WMAV1, |
972 |
|
|
.priv_data_size = sizeof(WMACodecContext), |
973 |
|
|
.init = wma_decode_init, |
974 |
|
|
.close = ff_wma_end, |
975 |
|
|
.decode = wma_decode_superframe, |
976 |
|
|
.flush = flush, |
977 |
|
|
.capabilities = AV_CODEC_CAP_DR1, |
978 |
|
|
.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP, |
979 |
|
|
AV_SAMPLE_FMT_NONE }, |
980 |
|
|
}; |
981 |
|
|
#endif |
982 |
|
|
#if CONFIG_WMAV2_DECODER |
983 |
|
|
AVCodec ff_wmav2_decoder = { |
984 |
|
|
.name = "wmav2", |
985 |
|
|
.long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 2"), |
986 |
|
|
.type = AVMEDIA_TYPE_AUDIO, |
987 |
|
|
.id = AV_CODEC_ID_WMAV2, |
988 |
|
|
.priv_data_size = sizeof(WMACodecContext), |
989 |
|
|
.init = wma_decode_init, |
990 |
|
|
.close = ff_wma_end, |
991 |
|
|
.decode = wma_decode_superframe, |
992 |
|
|
.flush = flush, |
993 |
|
|
.capabilities = AV_CODEC_CAP_DR1, |
994 |
|
|
.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP, |
995 |
|
|
AV_SAMPLE_FMT_NONE }, |
996 |
|
|
}; |
997 |
|
|
#endif |