1 | #ifndef MINIMP3_H |
2 | #define MINIMP3_H |
3 | /* |
4 | https://github.com/lieff/minimp3 |
5 | To the extent possible under law, the author(s) have dedicated all copyright and related and neighboring rights to this software to the public domain worldwide. |
6 | This software is distributed without any warranty. |
7 | See <http://creativecommons.org/publicdomain/zero/1.0/>. |
8 | */ |
9 | #include <stdint.h> |
10 | |
11 | #define MINIMP3_MAX_SAMPLES_PER_FRAME (1152*2) |
12 | |
13 | typedef struct |
14 | { |
15 | int frame_bytes, frame_offset, channels, hz, layer, bitrate_kbps; |
16 | } mp3dec_frame_info_t; |
17 | |
18 | typedef struct |
19 | { |
20 | float mdct_overlap[2][9*32], qmf_state[15*2*32]; |
21 | int reserv, free_format_bytes; |
22 | unsigned char [4], reserv_buf[511]; |
23 | } mp3dec_t; |
24 | |
25 | #ifdef __cplusplus |
26 | extern "C" { |
27 | #endif /* __cplusplus */ |
28 | |
29 | void mp3dec_init(mp3dec_t *dec); |
30 | #ifndef MINIMP3_FLOAT_OUTPUT |
31 | typedef int16_t mp3d_sample_t; |
32 | #else /* MINIMP3_FLOAT_OUTPUT */ |
33 | typedef float mp3d_sample_t; |
34 | void mp3dec_f32_to_s16(const float *in, int16_t *out, int num_samples); |
35 | #endif /* MINIMP3_FLOAT_OUTPUT */ |
36 | int mp3dec_decode_frame(mp3dec_t *dec, const uint8_t *mp3, int mp3_bytes, mp3d_sample_t *pcm, mp3dec_frame_info_t *info); |
37 | |
38 | #ifdef __cplusplus |
39 | } |
40 | #endif /* __cplusplus */ |
41 | |
42 | #endif /* MINIMP3_H */ |
43 | #if defined(MINIMP3_IMPLEMENTATION) && !defined(_MINIMP3_IMPLEMENTATION_GUARD) |
44 | #define _MINIMP3_IMPLEMENTATION_GUARD |
45 | |
46 | #include <stdlib.h> |
47 | #include <string.h> |
48 | |
49 | #define MAX_FREE_FORMAT_FRAME_SIZE 2304 /* more than ISO spec's */ |
50 | #ifndef MAX_FRAME_SYNC_MATCHES |
51 | #define MAX_FRAME_SYNC_MATCHES 10 |
52 | #endif /* MAX_FRAME_SYNC_MATCHES */ |
53 | |
54 | #define MAX_L3_FRAME_PAYLOAD_BYTES MAX_FREE_FORMAT_FRAME_SIZE /* MUST be >= 320000/8/32000*1152 = 1440 */ |
55 | |
56 | #define MAX_BITRESERVOIR_BYTES 511 |
57 | #define SHORT_BLOCK_TYPE 2 |
58 | #define STOP_BLOCK_TYPE 3 |
59 | #define MODE_MONO 3 |
60 | #define MODE_JOINT_STEREO 1 |
61 | #define HDR_SIZE 4 |
62 | #define HDR_IS_MONO(h) (((h[3]) & 0xC0) == 0xC0) |
63 | #define HDR_IS_MS_STEREO(h) (((h[3]) & 0xE0) == 0x60) |
64 | #define HDR_IS_FREE_FORMAT(h) (((h[2]) & 0xF0) == 0) |
65 | #define HDR_IS_CRC(h) (!((h[1]) & 1)) |
66 | #define HDR_TEST_PADDING(h) ((h[2]) & 0x2) |
67 | #define HDR_TEST_MPEG1(h) ((h[1]) & 0x8) |
68 | #define HDR_TEST_NOT_MPEG25(h) ((h[1]) & 0x10) |
69 | #define HDR_TEST_I_STEREO(h) ((h[3]) & 0x10) |
70 | #define HDR_TEST_MS_STEREO(h) ((h[3]) & 0x20) |
71 | #define HDR_GET_STEREO_MODE(h) (((h[3]) >> 6) & 3) |
72 | #define HDR_GET_STEREO_MODE_EXT(h) (((h[3]) >> 4) & 3) |
73 | #define HDR_GET_LAYER(h) (((h[1]) >> 1) & 3) |
74 | #define HDR_GET_BITRATE(h) ((h[2]) >> 4) |
75 | #define HDR_GET_SAMPLE_RATE(h) (((h[2]) >> 2) & 3) |
76 | #define HDR_GET_MY_SAMPLE_RATE(h) (HDR_GET_SAMPLE_RATE(h) + (((h[1] >> 3) & 1) + ((h[1] >> 4) & 1))*3) |
77 | #define HDR_IS_FRAME_576(h) ((h[1] & 14) == 2) |
78 | #define HDR_IS_LAYER_1(h) ((h[1] & 6) == 6) |
79 | |
80 | #define BITS_DEQUANTIZER_OUT -1 |
81 | #define MAX_SCF (255 + BITS_DEQUANTIZER_OUT*4 - 210) |
82 | #define MAX_SCFI ((MAX_SCF + 3) & ~3) |
83 | |
84 | #define MINIMP3_MIN(a, b) ((a) > (b) ? (b) : (a)) |
85 | #define MINIMP3_MAX(a, b) ((a) < (b) ? (b) : (a)) |
86 | |
87 | #if !defined(MINIMP3_NO_SIMD) |
88 | |
89 | #if !defined(MINIMP3_ONLY_SIMD) && (defined(_M_X64) || defined(__x86_64__) || defined(__aarch64__) || defined(_M_ARM64)) |
90 | /* x64 always have SSE2, arm64 always have neon, no need for generic code */ |
91 | #define MINIMP3_ONLY_SIMD |
92 | #endif /* SIMD checks... */ |
93 | |
94 | #if (defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_X64))) || ((defined(__i386__) || defined(__x86_64__)) && defined(__SSE2__)) |
95 | #if defined(_MSC_VER) |
96 | #include <intrin.h> |
97 | #endif /* defined(_MSC_VER) */ |
98 | #include <immintrin.h> |
99 | #define HAVE_SSE 1 |
100 | #define HAVE_SIMD 1 |
101 | #define VSTORE _mm_storeu_ps |
102 | #define VLD _mm_loadu_ps |
103 | #define VSET _mm_set1_ps |
104 | #define VADD _mm_add_ps |
105 | #define VSUB _mm_sub_ps |
106 | #define VMUL _mm_mul_ps |
107 | #define VMAC(a, x, y) _mm_add_ps(a, _mm_mul_ps(x, y)) |
108 | #define VMSB(a, x, y) _mm_sub_ps(a, _mm_mul_ps(x, y)) |
109 | #define VMUL_S(x, s) _mm_mul_ps(x, _mm_set1_ps(s)) |
110 | #define VREV(x) _mm_shuffle_ps(x, x, _MM_SHUFFLE(0, 1, 2, 3)) |
111 | typedef __m128 f4; |
112 | #if defined(_MSC_VER) || defined(MINIMP3_ONLY_SIMD) |
113 | #define minimp3_cpuid __cpuid |
114 | #else /* defined(_MSC_VER) || defined(MINIMP3_ONLY_SIMD) */ |
115 | static __inline__ __attribute__((always_inline)) void minimp3_cpuid(int CPUInfo[], const int InfoType) |
116 | { |
117 | #if defined(__PIC__) |
118 | __asm__ __volatile__( |
119 | #if defined(__x86_64__) |
120 | "push %%rbx\n" |
121 | "cpuid\n" |
122 | "xchgl %%ebx, %1\n" |
123 | "pop %%rbx\n" |
124 | #else /* defined(__x86_64__) */ |
125 | "xchgl %%ebx, %1\n" |
126 | "cpuid\n" |
127 | "xchgl %%ebx, %1\n" |
128 | #endif /* defined(__x86_64__) */ |
129 | : "=a" (CPUInfo[0]), "=r" (CPUInfo[1]), "=c" (CPUInfo[2]), "=d" (CPUInfo[3]) |
130 | : "a" (InfoType)); |
131 | #else /* defined(__PIC__) */ |
132 | __asm__ __volatile__( |
133 | "cpuid" |
134 | : "=a" (CPUInfo[0]), "=b" (CPUInfo[1]), "=c" (CPUInfo[2]), "=d" (CPUInfo[3]) |
135 | : "a" (InfoType)); |
136 | #endif /* defined(__PIC__)*/ |
137 | } |
138 | #endif /* defined(_MSC_VER) || defined(MINIMP3_ONLY_SIMD) */ |
139 | static int have_simd(void) |
140 | { |
141 | #ifdef MINIMP3_ONLY_SIMD |
142 | return 1; |
143 | #else /* MINIMP3_ONLY_SIMD */ |
144 | static int g_have_simd; |
145 | int CPUInfo[4]; |
146 | #ifdef MINIMP3_TEST |
147 | static int g_counter; |
148 | if (g_counter++ > 100) |
149 | return 0; |
150 | #endif /* MINIMP3_TEST */ |
151 | if (g_have_simd) |
152 | goto end; |
153 | minimp3_cpuid(CPUInfo, 0); |
154 | g_have_simd = 1; |
155 | if (CPUInfo[0] > 0) |
156 | { |
157 | minimp3_cpuid(CPUInfo, 1); |
158 | g_have_simd = (CPUInfo[3] & (1 << 26)) + 1; /* SSE2 */ |
159 | } |
160 | end: |
161 | return g_have_simd - 1; |
162 | #endif /* MINIMP3_ONLY_SIMD */ |
163 | } |
164 | #elif defined(__ARM_NEON) || defined(__aarch64__) || defined(_M_ARM64) |
165 | #include <arm_neon.h> |
166 | #define HAVE_SSE 0 |
167 | #define HAVE_SIMD 1 |
168 | #define VSTORE vst1q_f32 |
169 | #define VLD vld1q_f32 |
170 | #define VSET vmovq_n_f32 |
171 | #define VADD vaddq_f32 |
172 | #define VSUB vsubq_f32 |
173 | #define VMUL vmulq_f32 |
174 | #define VMAC(a, x, y) vmlaq_f32(a, x, y) |
175 | #define VMSB(a, x, y) vmlsq_f32(a, x, y) |
176 | #define VMUL_S(x, s) vmulq_f32(x, vmovq_n_f32(s)) |
177 | #define VREV(x) vcombine_f32(vget_high_f32(vrev64q_f32(x)), vget_low_f32(vrev64q_f32(x))) |
178 | typedef float32x4_t f4; |
179 | static int have_simd() |
180 | { /* TODO: detect neon for !MINIMP3_ONLY_SIMD */ |
181 | return 1; |
182 | } |
183 | #else /* SIMD checks... */ |
184 | #define HAVE_SSE 0 |
185 | #define HAVE_SIMD 0 |
186 | #ifdef MINIMP3_ONLY_SIMD |
187 | #error MINIMP3_ONLY_SIMD used, but SSE/NEON not enabled |
188 | #endif /* MINIMP3_ONLY_SIMD */ |
189 | #endif /* SIMD checks... */ |
190 | #else /* !defined(MINIMP3_NO_SIMD) */ |
191 | #define HAVE_SIMD 0 |
192 | #endif /* !defined(MINIMP3_NO_SIMD) */ |
193 | |
194 | #if defined(__ARM_ARCH) && (__ARM_ARCH >= 6) && !defined(__aarch64__) && !defined(_M_ARM64) |
195 | #define HAVE_ARMV6 1 |
196 | static __inline__ __attribute__((always_inline)) int32_t minimp3_clip_int16_arm(int32_t a) |
197 | { |
198 | int32_t x = 0; |
199 | __asm__ ("ssat %0, #16, %1" : "=r" (x) : "r" (a)); |
200 | return x; |
201 | } |
202 | #else |
203 | #define HAVE_ARMV6 0 |
204 | #endif |
205 | |
206 | typedef struct |
207 | { |
208 | const uint8_t *buf; |
209 | int pos, limit; |
210 | } bs_t; |
211 | |
212 | typedef struct |
213 | { |
214 | float scf[3*64]; |
215 | uint8_t total_bands, stereo_bands, bitalloc[64], scfcod[64]; |
216 | } L12_scale_info; |
217 | |
218 | typedef struct |
219 | { |
220 | uint8_t tab_offset, code_tab_width, band_count; |
221 | } L12_subband_alloc_t; |
222 | |
223 | typedef struct |
224 | { |
225 | const uint8_t *sfbtab; |
226 | uint16_t part_23_length, big_values, scalefac_compress; |
227 | uint8_t global_gain, block_type, mixed_block_flag, n_long_sfb, n_short_sfb; |
228 | uint8_t table_select[3], region_count[3], subblock_gain[3]; |
229 | uint8_t preflag, scalefac_scale, count1_table, scfsi; |
230 | } L3_gr_info_t; |
231 | |
232 | typedef struct |
233 | { |
234 | bs_t bs; |
235 | uint8_t maindata[MAX_BITRESERVOIR_BYTES + MAX_L3_FRAME_PAYLOAD_BYTES]; |
236 | L3_gr_info_t gr_info[4]; |
237 | float grbuf[2][576], scf[40], syn[18 + 15][2*32]; |
238 | uint8_t ist_pos[2][39]; |
239 | } mp3dec_scratch_t; |
240 | |
241 | static void bs_init(bs_t *bs, const uint8_t *data, int bytes) |
242 | { |
243 | bs->buf = data; |
244 | bs->pos = 0; |
245 | bs->limit = bytes*8; |
246 | } |
247 | |
248 | static uint32_t get_bits(bs_t *bs, int n) |
249 | { |
250 | uint32_t next, cache = 0, s = bs->pos & 7; |
251 | int shl = n + s; |
252 | const uint8_t *p = bs->buf + (bs->pos >> 3); |
253 | if ((bs->pos += n) > bs->limit) |
254 | return 0; |
255 | next = *p++ & (255 >> s); |
256 | while ((shl -= 8) > 0) |
257 | { |
258 | cache |= next << shl; |
259 | next = *p++; |
260 | } |
261 | return cache | (next >> -shl); |
262 | } |
263 | |
264 | static int hdr_valid(const uint8_t *h) |
265 | { |
266 | return h[0] == 0xff && |
267 | ((h[1] & 0xF0) == 0xf0 || (h[1] & 0xFE) == 0xe2) && |
268 | (HDR_GET_LAYER(h) != 0) && |
269 | (HDR_GET_BITRATE(h) != 15) && |
270 | (HDR_GET_SAMPLE_RATE(h) != 3); |
271 | } |
272 | |
273 | static int hdr_compare(const uint8_t *h1, const uint8_t *h2) |
274 | { |
275 | return hdr_valid(h2) && |
276 | ((h1[1] ^ h2[1]) & 0xFE) == 0 && |
277 | ((h1[2] ^ h2[2]) & 0x0C) == 0 && |
278 | !(HDR_IS_FREE_FORMAT(h1) ^ HDR_IS_FREE_FORMAT(h2)); |
279 | } |
280 | |
281 | static unsigned hdr_bitrate_kbps(const uint8_t *h) |
282 | { |
283 | static const uint8_t halfrate[2][3][15] = { |
284 | { { 0,4,8,12,16,20,24,28,32,40,48,56,64,72,80 }, { 0,4,8,12,16,20,24,28,32,40,48,56,64,72,80 }, { 0,16,24,28,32,40,48,56,64,72,80,88,96,112,128 } }, |
285 | { { 0,16,20,24,28,32,40,48,56,64,80,96,112,128,160 }, { 0,16,24,28,32,40,48,56,64,80,96,112,128,160,192 }, { 0,16,32,48,64,80,96,112,128,144,160,176,192,208,224 } }, |
286 | }; |
287 | return 2*halfrate[!!HDR_TEST_MPEG1(h)][HDR_GET_LAYER(h) - 1][HDR_GET_BITRATE(h)]; |
288 | } |
289 | |
290 | static unsigned hdr_sample_rate_hz(const uint8_t *h) |
291 | { |
292 | static const unsigned g_hz[3] = { 44100, 48000, 32000 }; |
293 | return g_hz[HDR_GET_SAMPLE_RATE(h)] >> (int)!HDR_TEST_MPEG1(h) >> (int)!HDR_TEST_NOT_MPEG25(h); |
294 | } |
295 | |
296 | static unsigned hdr_frame_samples(const uint8_t *h) |
297 | { |
298 | return HDR_IS_LAYER_1(h) ? 384 : (1152 >> (int)HDR_IS_FRAME_576(h)); |
299 | } |
300 | |
301 | static int hdr_frame_bytes(const uint8_t *h, int free_format_size) |
302 | { |
303 | int frame_bytes = hdr_frame_samples(h)*hdr_bitrate_kbps(h)*125/hdr_sample_rate_hz(h); |
304 | if (HDR_IS_LAYER_1(h)) |
305 | { |
306 | frame_bytes &= ~3; /* slot align */ |
307 | } |
308 | return frame_bytes ? frame_bytes : free_format_size; |
309 | } |
310 | |
311 | static int hdr_padding(const uint8_t *h) |
312 | { |
313 | return HDR_TEST_PADDING(h) ? (HDR_IS_LAYER_1(h) ? 4 : 1) : 0; |
314 | } |
315 | |
316 | #ifndef MINIMP3_ONLY_MP3 |
317 | static const L12_subband_alloc_t *L12_subband_alloc_table(const uint8_t *hdr, L12_scale_info *sci) |
318 | { |
319 | const L12_subband_alloc_t *alloc; |
320 | int mode = HDR_GET_STEREO_MODE(hdr); |
321 | int nbands, stereo_bands = (mode == MODE_MONO) ? 0 : (mode == MODE_JOINT_STEREO) ? (HDR_GET_STEREO_MODE_EXT(hdr) << 2) + 4 : 32; |
322 | |
323 | if (HDR_IS_LAYER_1(hdr)) |
324 | { |
325 | static const L12_subband_alloc_t g_alloc_L1[] = { { 76, 4, 32 } }; |
326 | alloc = g_alloc_L1; |
327 | nbands = 32; |
328 | } else if (!HDR_TEST_MPEG1(hdr)) |
329 | { |
330 | static const L12_subband_alloc_t g_alloc_L2M2[] = { { 60, 4, 4 }, { 44, 3, 7 }, { 44, 2, 19 } }; |
331 | alloc = g_alloc_L2M2; |
332 | nbands = 30; |
333 | } else |
334 | { |
335 | static const L12_subband_alloc_t g_alloc_L2M1[] = { { 0, 4, 3 }, { 16, 4, 8 }, { 32, 3, 12 }, { 40, 2, 7 } }; |
336 | int sample_rate_idx = HDR_GET_SAMPLE_RATE(hdr); |
337 | unsigned kbps = hdr_bitrate_kbps(hdr) >> (int)(mode != MODE_MONO); |
338 | if (!kbps) /* free-format */ |
339 | { |
340 | kbps = 192; |
341 | } |
342 | |
343 | alloc = g_alloc_L2M1; |
344 | nbands = 27; |
345 | if (kbps < 56) |
346 | { |
347 | static const L12_subband_alloc_t g_alloc_L2M1_lowrate[] = { { 44, 4, 2 }, { 44, 3, 10 } }; |
348 | alloc = g_alloc_L2M1_lowrate; |
349 | nbands = sample_rate_idx == 2 ? 12 : 8; |
350 | } else if (kbps >= 96 && sample_rate_idx != 1) |
351 | { |
352 | nbands = 30; |
353 | } |
354 | } |
355 | |
356 | sci->total_bands = (uint8_t)nbands; |
357 | sci->stereo_bands = (uint8_t)MINIMP3_MIN(stereo_bands, nbands); |
358 | |
359 | return alloc; |
360 | } |
361 | |
362 | static void L12_read_scalefactors(bs_t *bs, uint8_t *pba, uint8_t *scfcod, int bands, float *scf) |
363 | { |
364 | static const float g_deq_L12[18*3] = { |
365 | #define DQ(x) 9.53674316e-07f/x, 7.56931807e-07f/x, 6.00777173e-07f/x |
366 | DQ(3),DQ(7),DQ(15),DQ(31),DQ(63),DQ(127),DQ(255),DQ(511),DQ(1023),DQ(2047),DQ(4095),DQ(8191),DQ(16383),DQ(32767),DQ(65535),DQ(3),DQ(5),DQ(9) |
367 | }; |
368 | int i, m; |
369 | for (i = 0; i < bands; i++) |
370 | { |
371 | float s = 0; |
372 | int ba = *pba++; |
373 | int mask = ba ? 4 + ((19 >> scfcod[i]) & 3) : 0; |
374 | for (m = 4; m; m >>= 1) |
375 | { |
376 | if (mask & m) |
377 | { |
378 | int b = get_bits(bs, 6); |
379 | s = g_deq_L12[ba*3 - 6 + b % 3]*(1 << 21 >> b/3); |
380 | } |
381 | *scf++ = s; |
382 | } |
383 | } |
384 | } |
385 | |
386 | static void L12_read_scale_info(const uint8_t *hdr, bs_t *bs, L12_scale_info *sci) |
387 | { |
388 | static const uint8_t g_bitalloc_code_tab[] = { |
389 | 0,17, 3, 4, 5,6,7, 8,9,10,11,12,13,14,15,16, |
390 | 0,17,18, 3,19,4,5, 6,7, 8, 9,10,11,12,13,16, |
391 | 0,17,18, 3,19,4,5,16, |
392 | 0,17,18,16, |
393 | 0,17,18,19, 4,5,6, 7,8, 9,10,11,12,13,14,15, |
394 | 0,17,18, 3,19,4,5, 6,7, 8, 9,10,11,12,13,14, |
395 | 0, 2, 3, 4, 5,6,7, 8,9,10,11,12,13,14,15,16 |
396 | }; |
397 | const L12_subband_alloc_t *subband_alloc = L12_subband_alloc_table(hdr, sci); |
398 | |
399 | int i, k = 0, ba_bits = 0; |
400 | const uint8_t *ba_code_tab = g_bitalloc_code_tab; |
401 | |
402 | for (i = 0; i < sci->total_bands; i++) |
403 | { |
404 | uint8_t ba; |
405 | if (i == k) |
406 | { |
407 | k += subband_alloc->band_count; |
408 | ba_bits = subband_alloc->code_tab_width; |
409 | ba_code_tab = g_bitalloc_code_tab + subband_alloc->tab_offset; |
410 | subband_alloc++; |
411 | } |
412 | ba = ba_code_tab[get_bits(bs, ba_bits)]; |
413 | sci->bitalloc[2*i] = ba; |
414 | if (i < sci->stereo_bands) |
415 | { |
416 | ba = ba_code_tab[get_bits(bs, ba_bits)]; |
417 | } |
418 | sci->bitalloc[2*i + 1] = sci->stereo_bands ? ba : 0; |
419 | } |
420 | |
421 | for (i = 0; i < 2*sci->total_bands; i++) |
422 | { |
423 | sci->scfcod[i] = sci->bitalloc[i] ? HDR_IS_LAYER_1(hdr) ? 2 : get_bits(bs, 2) : 6; |
424 | } |
425 | |
426 | L12_read_scalefactors(bs, sci->bitalloc, sci->scfcod, sci->total_bands*2, sci->scf); |
427 | |
428 | for (i = sci->stereo_bands; i < sci->total_bands; i++) |
429 | { |
430 | sci->bitalloc[2*i + 1] = 0; |
431 | } |
432 | } |
433 | |
434 | static int L12_dequantize_granule(float *grbuf, bs_t *bs, L12_scale_info *sci, int group_size) |
435 | { |
436 | int i, j, k, choff = 576; |
437 | for (j = 0; j < 4; j++) |
438 | { |
439 | float *dst = grbuf + group_size*j; |
440 | for (i = 0; i < 2*sci->total_bands; i++) |
441 | { |
442 | int ba = sci->bitalloc[i]; |
443 | if (ba != 0) |
444 | { |
445 | if (ba < 17) |
446 | { |
447 | int half = (1 << (ba - 1)) - 1; |
448 | for (k = 0; k < group_size; k++) |
449 | { |
450 | dst[k] = (float)((int)get_bits(bs, ba) - half); |
451 | } |
452 | } else |
453 | { |
454 | unsigned mod = (2 << (ba - 17)) + 1; /* 3, 5, 9 */ |
455 | unsigned code = get_bits(bs, mod + 2 - (mod >> 3)); /* 5, 7, 10 */ |
456 | for (k = 0; k < group_size; k++, code /= mod) |
457 | { |
458 | dst[k] = (float)((int)(code % mod - mod/2)); |
459 | } |
460 | } |
461 | } |
462 | dst += choff; |
463 | choff = 18 - choff; |
464 | } |
465 | } |
466 | return group_size*4; |
467 | } |
468 | |
469 | static void L12_apply_scf_384(L12_scale_info *sci, const float *scf, float *dst) |
470 | { |
471 | int i, k; |
472 | memcpy(dst + 576 + sci->stereo_bands*18, dst + sci->stereo_bands*18, (sci->total_bands - sci->stereo_bands)*18*sizeof(float)); |
473 | for (i = 0; i < sci->total_bands; i++, dst += 18, scf += 6) |
474 | { |
475 | for (k = 0; k < 12; k++) |
476 | { |
477 | dst[k + 0] *= scf[0]; |
478 | dst[k + 576] *= scf[3]; |
479 | } |
480 | } |
481 | } |
482 | #endif /* MINIMP3_ONLY_MP3 */ |
483 | |
484 | static int L3_read_side_info(bs_t *bs, L3_gr_info_t *gr, const uint8_t *hdr) |
485 | { |
486 | static const uint8_t g_scf_long[8][23] = { |
487 | { 6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54,0 }, |
488 | { 12,12,12,12,12,12,16,20,24,28,32,40,48,56,64,76,90,2,2,2,2,2,0 }, |
489 | { 6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54,0 }, |
490 | { 6,6,6,6,6,6,8,10,12,14,16,18,22,26,32,38,46,54,62,70,76,36,0 }, |
491 | { 6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54,0 }, |
492 | { 4,4,4,4,4,4,6,6,8,8,10,12,16,20,24,28,34,42,50,54,76,158,0 }, |
493 | { 4,4,4,4,4,4,6,6,6,8,10,12,16,18,22,28,34,40,46,54,54,192,0 }, |
494 | { 4,4,4,4,4,4,6,6,8,10,12,16,20,24,30,38,46,56,68,84,102,26,0 } |
495 | }; |
496 | static const uint8_t g_scf_short[8][40] = { |
497 | { 4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,30,30,30,40,40,40,18,18,18,0 }, |
498 | { 8,8,8,8,8,8,8,8,8,12,12,12,16,16,16,20,20,20,24,24,24,28,28,28,36,36,36,2,2,2,2,2,2,2,2,2,26,26,26,0 }, |
499 | { 4,4,4,4,4,4,4,4,4,6,6,6,6,6,6,8,8,8,10,10,10,14,14,14,18,18,18,26,26,26,32,32,32,42,42,42,18,18,18,0 }, |
500 | { 4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,32,32,32,44,44,44,12,12,12,0 }, |
501 | { 4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,30,30,30,40,40,40,18,18,18,0 }, |
502 | { 4,4,4,4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,22,22,22,30,30,30,56,56,56,0 }, |
503 | { 4,4,4,4,4,4,4,4,4,4,4,4,6,6,6,6,6,6,10,10,10,12,12,12,14,14,14,16,16,16,20,20,20,26,26,26,66,66,66,0 }, |
504 | { 4,4,4,4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,12,12,12,16,16,16,20,20,20,26,26,26,34,34,34,42,42,42,12,12,12,0 } |
505 | }; |
506 | static const uint8_t g_scf_mixed[8][40] = { |
507 | { 6,6,6,6,6,6,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,30,30,30,40,40,40,18,18,18,0 }, |
508 | { 12,12,12,4,4,4,8,8,8,12,12,12,16,16,16,20,20,20,24,24,24,28,28,28,36,36,36,2,2,2,2,2,2,2,2,2,26,26,26,0 }, |
509 | { 6,6,6,6,6,6,6,6,6,6,6,6,8,8,8,10,10,10,14,14,14,18,18,18,26,26,26,32,32,32,42,42,42,18,18,18,0 }, |
510 | { 6,6,6,6,6,6,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,32,32,32,44,44,44,12,12,12,0 }, |
511 | { 6,6,6,6,6,6,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,30,30,30,40,40,40,18,18,18,0 }, |
512 | { 4,4,4,4,4,4,6,6,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,22,22,22,30,30,30,56,56,56,0 }, |
513 | { 4,4,4,4,4,4,6,6,4,4,4,6,6,6,6,6,6,10,10,10,12,12,12,14,14,14,16,16,16,20,20,20,26,26,26,66,66,66,0 }, |
514 | { 4,4,4,4,4,4,6,6,4,4,4,6,6,6,8,8,8,12,12,12,16,16,16,20,20,20,26,26,26,34,34,34,42,42,42,12,12,12,0 } |
515 | }; |
516 | |
517 | unsigned tables, scfsi = 0; |
518 | int main_data_begin, part_23_sum = 0; |
519 | int sr_idx = HDR_GET_MY_SAMPLE_RATE(hdr); sr_idx -= (sr_idx != 0); |
520 | int gr_count = HDR_IS_MONO(hdr) ? 1 : 2; |
521 | |
522 | if (HDR_TEST_MPEG1(hdr)) |
523 | { |
524 | gr_count *= 2; |
525 | main_data_begin = get_bits(bs, 9); |
526 | scfsi = get_bits(bs, 7 + gr_count); |
527 | } else |
528 | { |
529 | main_data_begin = get_bits(bs, 8 + gr_count) >> gr_count; |
530 | } |
531 | |
532 | do |
533 | { |
534 | if (HDR_IS_MONO(hdr)) |
535 | { |
536 | scfsi <<= 4; |
537 | } |
538 | gr->part_23_length = (uint16_t)get_bits(bs, 12); |
539 | part_23_sum += gr->part_23_length; |
540 | gr->big_values = (uint16_t)get_bits(bs, 9); |
541 | if (gr->big_values > 288) |
542 | { |
543 | return -1; |
544 | } |
545 | gr->global_gain = (uint8_t)get_bits(bs, 8); |
546 | gr->scalefac_compress = (uint16_t)get_bits(bs, HDR_TEST_MPEG1(hdr) ? 4 : 9); |
547 | gr->sfbtab = g_scf_long[sr_idx]; |
548 | gr->n_long_sfb = 22; |
549 | gr->n_short_sfb = 0; |
550 | if (get_bits(bs, 1)) |
551 | { |
552 | gr->block_type = (uint8_t)get_bits(bs, 2); |
553 | if (!gr->block_type) |
554 | { |
555 | return -1; |
556 | } |
557 | gr->mixed_block_flag = (uint8_t)get_bits(bs, 1); |
558 | gr->region_count[0] = 7; |
559 | gr->region_count[1] = 255; |
560 | if (gr->block_type == SHORT_BLOCK_TYPE) |
561 | { |
562 | scfsi &= 0x0F0F; |
563 | if (!gr->mixed_block_flag) |
564 | { |
565 | gr->region_count[0] = 8; |
566 | gr->sfbtab = g_scf_short[sr_idx]; |
567 | gr->n_long_sfb = 0; |
568 | gr->n_short_sfb = 39; |
569 | } else |
570 | { |
571 | gr->sfbtab = g_scf_mixed[sr_idx]; |
572 | gr->n_long_sfb = HDR_TEST_MPEG1(hdr) ? 8 : 6; |
573 | gr->n_short_sfb = 30; |
574 | } |
575 | } |
576 | tables = get_bits(bs, 10); |
577 | tables <<= 5; |
578 | gr->subblock_gain[0] = (uint8_t)get_bits(bs, 3); |
579 | gr->subblock_gain[1] = (uint8_t)get_bits(bs, 3); |
580 | gr->subblock_gain[2] = (uint8_t)get_bits(bs, 3); |
581 | } else |
582 | { |
583 | gr->block_type = 0; |
584 | gr->mixed_block_flag = 0; |
585 | tables = get_bits(bs, 15); |
586 | gr->region_count[0] = (uint8_t)get_bits(bs, 4); |
587 | gr->region_count[1] = (uint8_t)get_bits(bs, 3); |
588 | gr->region_count[2] = 255; |
589 | } |
590 | gr->table_select[0] = (uint8_t)(tables >> 10); |
591 | gr->table_select[1] = (uint8_t)((tables >> 5) & 31); |
592 | gr->table_select[2] = (uint8_t)((tables) & 31); |
593 | gr->preflag = HDR_TEST_MPEG1(hdr) ? get_bits(bs, 1) : (gr->scalefac_compress >= 500); |
594 | gr->scalefac_scale = (uint8_t)get_bits(bs, 1); |
595 | gr->count1_table = (uint8_t)get_bits(bs, 1); |
596 | gr->scfsi = (uint8_t)((scfsi >> 12) & 15); |
597 | scfsi <<= 4; |
598 | gr++; |
599 | } while(--gr_count); |
600 | |
601 | if (part_23_sum + bs->pos > bs->limit + main_data_begin*8) |
602 | { |
603 | return -1; |
604 | } |
605 | |
606 | return main_data_begin; |
607 | } |
608 | |
609 | static void L3_read_scalefactors(uint8_t *scf, uint8_t *ist_pos, const uint8_t *scf_size, const uint8_t *scf_count, bs_t *bitbuf, int scfsi) |
610 | { |
611 | int i, k; |
612 | for (i = 0; i < 4 && scf_count[i]; i++, scfsi *= 2) |
613 | { |
614 | int cnt = scf_count[i]; |
615 | if (scfsi & 8) |
616 | { |
617 | memcpy(scf, ist_pos, cnt); |
618 | } else |
619 | { |
620 | int bits = scf_size[i]; |
621 | if (!bits) |
622 | { |
623 | memset(scf, 0, cnt); |
624 | memset(ist_pos, 0, cnt); |
625 | } else |
626 | { |
627 | int max_scf = (scfsi < 0) ? (1 << bits) - 1 : -1; |
628 | for (k = 0; k < cnt; k++) |
629 | { |
630 | int s = get_bits(bitbuf, bits); |
631 | ist_pos[k] = (s == max_scf ? -1 : s); |
632 | scf[k] = s; |
633 | } |
634 | } |
635 | } |
636 | ist_pos += cnt; |
637 | scf += cnt; |
638 | } |
639 | scf[0] = scf[1] = scf[2] = 0; |
640 | } |
641 | |
642 | static float L3_ldexp_q2(float y, int exp_q2) |
643 | { |
644 | static const float g_expfrac[4] = { 9.31322575e-10f,7.83145814e-10f,6.58544508e-10f,5.53767716e-10f }; |
645 | int e; |
646 | do |
647 | { |
648 | e = MINIMP3_MIN(30*4, exp_q2); |
649 | y *= g_expfrac[e & 3]*(1 << 30 >> (e >> 2)); |
650 | } while ((exp_q2 -= e) > 0); |
651 | return y; |
652 | } |
653 | |
654 | static void L3_decode_scalefactors(const uint8_t *hdr, uint8_t *ist_pos, bs_t *bs, const L3_gr_info_t *gr, float *scf, int ch) |
655 | { |
656 | static const uint8_t g_scf_partitions[3][28] = { |
657 | { 6,5,5, 5,6,5,5,5,6,5, 7,3,11,10,0,0, 7, 7, 7,0, 6, 6,6,3, 8, 8,5,0 }, |
658 | { 8,9,6,12,6,9,9,9,6,9,12,6,15,18,0,0, 6,15,12,0, 6,12,9,6, 6,18,9,0 }, |
659 | { 9,9,6,12,9,9,9,9,9,9,12,6,18,18,0,0,12,12,12,0,12, 9,9,6,15,12,9,0 } |
660 | }; |
661 | const uint8_t *scf_partition = g_scf_partitions[!!gr->n_short_sfb + !gr->n_long_sfb]; |
662 | uint8_t scf_size[4], iscf[40]; |
663 | int i, scf_shift = gr->scalefac_scale + 1, gain_exp, scfsi = gr->scfsi; |
664 | float gain; |
665 | |
666 | if (HDR_TEST_MPEG1(hdr)) |
667 | { |
668 | static const uint8_t g_scfc_decode[16] = { 0,1,2,3, 12,5,6,7, 9,10,11,13, 14,15,18,19 }; |
669 | int part = g_scfc_decode[gr->scalefac_compress]; |
670 | scf_size[1] = scf_size[0] = (uint8_t)(part >> 2); |
671 | scf_size[3] = scf_size[2] = (uint8_t)(part & 3); |
672 | } else |
673 | { |
674 | static const uint8_t g_mod[6*4] = { 5,5,4,4,5,5,4,1,4,3,1,1,5,6,6,1,4,4,4,1,4,3,1,1 }; |
675 | int k, modprod, sfc, ist = HDR_TEST_I_STEREO(hdr) && ch; |
676 | sfc = gr->scalefac_compress >> ist; |
677 | for (k = ist*3*4; sfc >= 0; sfc -= modprod, k += 4) |
678 | { |
679 | for (modprod = 1, i = 3; i >= 0; i--) |
680 | { |
681 | scf_size[i] = (uint8_t)(sfc / modprod % g_mod[k + i]); |
682 | modprod *= g_mod[k + i]; |
683 | } |
684 | } |
685 | scf_partition += k; |
686 | scfsi = -16; |
687 | } |
688 | L3_read_scalefactors(iscf, ist_pos, scf_size, scf_partition, bs, scfsi); |
689 | |
690 | if (gr->n_short_sfb) |
691 | { |
692 | int sh = 3 - scf_shift; |
693 | for (i = 0; i < gr->n_short_sfb; i += 3) |
694 | { |
695 | iscf[gr->n_long_sfb + i + 0] += gr->subblock_gain[0] << sh; |
696 | iscf[gr->n_long_sfb + i + 1] += gr->subblock_gain[1] << sh; |
697 | iscf[gr->n_long_sfb + i + 2] += gr->subblock_gain[2] << sh; |
698 | } |
699 | } else if (gr->preflag) |
700 | { |
701 | static const uint8_t g_preamp[10] = { 1,1,1,1,2,2,3,3,3,2 }; |
702 | for (i = 0; i < 10; i++) |
703 | { |
704 | iscf[11 + i] += g_preamp[i]; |
705 | } |
706 | } |
707 | |
708 | gain_exp = gr->global_gain + BITS_DEQUANTIZER_OUT*4 - 210 - (HDR_IS_MS_STEREO(hdr) ? 2 : 0); |
709 | gain = L3_ldexp_q2(1 << (MAX_SCFI/4), MAX_SCFI - gain_exp); |
710 | for (i = 0; i < (int)(gr->n_long_sfb + gr->n_short_sfb); i++) |
711 | { |
712 | scf[i] = L3_ldexp_q2(gain, iscf[i] << scf_shift); |
713 | } |
714 | } |
715 | |
716 | static const float g_pow43[129 + 16] = { |
717 | 0,-1,-2.519842f,-4.326749f,-6.349604f,-8.549880f,-10.902724f,-13.390518f,-16.000000f,-18.720754f,-21.544347f,-24.463781f,-27.473142f,-30.567351f,-33.741992f,-36.993181f, |
718 | 0,1,2.519842f,4.326749f,6.349604f,8.549880f,10.902724f,13.390518f,16.000000f,18.720754f,21.544347f,24.463781f,27.473142f,30.567351f,33.741992f,36.993181f,40.317474f,43.711787f,47.173345f,50.699631f,54.288352f,57.937408f,61.644865f,65.408941f,69.227979f,73.100443f,77.024898f,81.000000f,85.024491f,89.097188f,93.216975f,97.382800f,101.593667f,105.848633f,110.146801f,114.487321f,118.869381f,123.292209f,127.755065f,132.257246f,136.798076f,141.376907f,145.993119f,150.646117f,155.335327f,160.060199f,164.820202f,169.614826f,174.443577f,179.305980f,184.201575f,189.129918f,194.090580f,199.083145f,204.107210f,209.162385f,214.248292f,219.364564f,224.510845f,229.686789f,234.892058f,240.126328f,245.389280f,250.680604f,256.000000f,261.347174f,266.721841f,272.123723f,277.552547f,283.008049f,288.489971f,293.998060f,299.532071f,305.091761f,310.676898f,316.287249f,321.922592f,327.582707f,333.267377f,338.976394f,344.709550f,350.466646f,356.247482f,362.051866f,367.879608f,373.730522f,379.604427f,385.501143f,391.420496f,397.362314f,403.326427f,409.312672f,415.320884f,421.350905f,427.402579f,433.475750f,439.570269f,445.685987f,451.822757f,457.980436f,464.158883f,470.357960f,476.577530f,482.817459f,489.077615f,495.357868f,501.658090f,507.978156f,514.317941f,520.677324f,527.056184f,533.454404f,539.871867f,546.308458f,552.764065f,559.238575f,565.731879f,572.243870f,578.774440f,585.323483f,591.890898f,598.476581f,605.080431f,611.702349f,618.342238f,625.000000f,631.675540f,638.368763f,645.079578f |
719 | }; |
720 | |
721 | static float L3_pow_43(int x) |
722 | { |
723 | float frac; |
724 | int sign, mult = 256; |
725 | |
726 | if (x < 129) |
727 | { |
728 | return g_pow43[16 + x]; |
729 | } |
730 | |
731 | if (x < 1024) |
732 | { |
733 | mult = 16; |
734 | x <<= 3; |
735 | } |
736 | |
737 | sign = 2*x & 64; |
738 | frac = (float)((x & 63) - sign) / ((x & ~63) + sign); |
739 | return g_pow43[16 + ((x + sign) >> 6)]*(1.f + frac*((4.f/3) + frac*(2.f/9)))*mult; |
740 | } |
741 | |
742 | static void L3_huffman(float *dst, bs_t *bs, const L3_gr_info_t *gr_info, const float *scf, int layer3gr_limit) |
743 | { |
744 | static const int16_t tabs[] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
745 | 785,785,785,785,784,784,784,784,513,513,513,513,513,513,513,513,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256, |
746 | -255,1313,1298,1282,785,785,785,785,784,784,784,784,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,290,288, |
747 | -255,1313,1298,1282,769,769,769,769,529,529,529,529,529,529,529,529,528,528,528,528,528,528,528,528,512,512,512,512,512,512,512,512,290,288, |
748 | -253,-318,-351,-367,785,785,785,785,784,784,784,784,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,819,818,547,547,275,275,275,275,561,560,515,546,289,274,288,258, |
749 | -254,-287,1329,1299,1314,1312,1057,1057,1042,1042,1026,1026,784,784,784,784,529,529,529,529,529,529,529,529,769,769,769,769,768,768,768,768,563,560,306,306,291,259, |
750 | -252,-413,-477,-542,1298,-575,1041,1041,784,784,784,784,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,-383,-399,1107,1092,1106,1061,849,849,789,789,1104,1091,773,773,1076,1075,341,340,325,309,834,804,577,577,532,532,516,516,832,818,803,816,561,561,531,531,515,546,289,289,288,258, |
751 | -252,-429,-493,-559,1057,1057,1042,1042,529,529,529,529,529,529,529,529,784,784,784,784,769,769,769,769,512,512,512,512,512,512,512,512,-382,1077,-415,1106,1061,1104,849,849,789,789,1091,1076,1029,1075,834,834,597,581,340,340,339,324,804,833,532,532,832,772,818,803,817,787,816,771,290,290,290,290,288,258, |
752 | -253,-349,-414,-447,-463,1329,1299,-479,1314,1312,1057,1057,1042,1042,1026,1026,785,785,785,785,784,784,784,784,769,769,769,769,768,768,768,768,-319,851,821,-335,836,850,805,849,341,340,325,336,533,533,579,579,564,564,773,832,578,548,563,516,321,276,306,291,304,259, |
753 | -251,-572,-733,-830,-863,-879,1041,1041,784,784,784,784,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,-511,-527,-543,1396,1351,1381,1366,1395,1335,1380,-559,1334,1138,1138,1063,1063,1350,1392,1031,1031,1062,1062,1364,1363,1120,1120,1333,1348,881,881,881,881,375,374,359,373,343,358,341,325,791,791,1123,1122,-703,1105,1045,-719,865,865,790,790,774,774,1104,1029,338,293,323,308,-799,-815,833,788,772,818,803,816,322,292,307,320,561,531,515,546,289,274,288,258, |
754 | -251,-525,-605,-685,-765,-831,-846,1298,1057,1057,1312,1282,785,785,785,785,784,784,784,784,769,769,769,769,512,512,512,512,512,512,512,512,1399,1398,1383,1367,1382,1396,1351,-511,1381,1366,1139,1139,1079,1079,1124,1124,1364,1349,1363,1333,882,882,882,882,807,807,807,807,1094,1094,1136,1136,373,341,535,535,881,775,867,822,774,-591,324,338,-671,849,550,550,866,864,609,609,293,336,534,534,789,835,773,-751,834,804,308,307,833,788,832,772,562,562,547,547,305,275,560,515,290,290, |
755 | -252,-397,-477,-557,-622,-653,-719,-735,-750,1329,1299,1314,1057,1057,1042,1042,1312,1282,1024,1024,785,785,785,785,784,784,784,784,769,769,769,769,-383,1127,1141,1111,1126,1140,1095,1110,869,869,883,883,1079,1109,882,882,375,374,807,868,838,881,791,-463,867,822,368,263,852,837,836,-543,610,610,550,550,352,336,534,534,865,774,851,821,850,805,593,533,579,564,773,832,578,578,548,548,577,577,307,276,306,291,516,560,259,259, |
756 | -250,-2107,-2507,-2764,-2909,-2974,-3007,-3023,1041,1041,1040,1040,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,-767,-1052,-1213,-1277,-1358,-1405,-1469,-1535,-1550,-1582,-1614,-1647,-1662,-1694,-1726,-1759,-1774,-1807,-1822,-1854,-1886,1565,-1919,-1935,-1951,-1967,1731,1730,1580,1717,-1983,1729,1564,-1999,1548,-2015,-2031,1715,1595,-2047,1714,-2063,1610,-2079,1609,-2095,1323,1323,1457,1457,1307,1307,1712,1547,1641,1700,1699,1594,1685,1625,1442,1442,1322,1322,-780,-973,-910,1279,1278,1277,1262,1276,1261,1275,1215,1260,1229,-959,974,974,989,989,-943,735,478,478,495,463,506,414,-1039,1003,958,1017,927,942,987,957,431,476,1272,1167,1228,-1183,1256,-1199,895,895,941,941,1242,1227,1212,1135,1014,1014,490,489,503,487,910,1013,985,925,863,894,970,955,1012,847,-1343,831,755,755,984,909,428,366,754,559,-1391,752,486,457,924,997,698,698,983,893,740,740,908,877,739,739,667,667,953,938,497,287,271,271,683,606,590,712,726,574,302,302,738,736,481,286,526,725,605,711,636,724,696,651,589,681,666,710,364,467,573,695,466,466,301,465,379,379,709,604,665,679,316,316,634,633,436,436,464,269,424,394,452,332,438,363,347,408,393,448,331,422,362,407,392,421,346,406,391,376,375,359,1441,1306,-2367,1290,-2383,1337,-2399,-2415,1426,1321,-2431,1411,1336,-2447,-2463,-2479,1169,1169,1049,1049,1424,1289,1412,1352,1319,-2495,1154,1154,1064,1064,1153,1153,416,390,360,404,403,389,344,374,373,343,358,372,327,357,342,311,356,326,1395,1394,1137,1137,1047,1047,1365,1392,1287,1379,1334,1364,1349,1378,1318,1363,792,792,792,792,1152,1152,1032,1032,1121,1121,1046,1046,1120,1120,1030,1030,-2895,1106,1061,1104,849,849,789,789,1091,1076,1029,1090,1060,1075,833,833,309,324,532,532,832,772,818,803,561,561,531,560,515,546,289,274,288,258, |
757 | -250,-1179,-1579,-1836,-1996,-2124,-2253,-2333,-2413,-2477,-2542,-2574,-2607,-2622,-2655,1314,1313,1298,1312,1282,785,785,785,785,1040,1040,1025,1025,768,768,768,768,-766,-798,-830,-862,-895,-911,-927,-943,-959,-975,-991,-1007,-1023,-1039,-1055,-1070,1724,1647,-1103,-1119,1631,1767,1662,1738,1708,1723,-1135,1780,1615,1779,1599,1677,1646,1778,1583,-1151,1777,1567,1737,1692,1765,1722,1707,1630,1751,1661,1764,1614,1736,1676,1763,1750,1645,1598,1721,1691,1762,1706,1582,1761,1566,-1167,1749,1629,767,766,751,765,494,494,735,764,719,749,734,763,447,447,748,718,477,506,431,491,446,476,461,505,415,430,475,445,504,399,460,489,414,503,383,474,429,459,502,502,746,752,488,398,501,473,413,472,486,271,480,270,-1439,-1455,1357,-1471,-1487,-1503,1341,1325,-1519,1489,1463,1403,1309,-1535,1372,1448,1418,1476,1356,1462,1387,-1551,1475,1340,1447,1402,1386,-1567,1068,1068,1474,1461,455,380,468,440,395,425,410,454,364,467,466,464,453,269,409,448,268,432,1371,1473,1432,1417,1308,1460,1355,1446,1459,1431,1083,1083,1401,1416,1458,1445,1067,1067,1370,1457,1051,1051,1291,1430,1385,1444,1354,1415,1400,1443,1082,1082,1173,1113,1186,1066,1185,1050,-1967,1158,1128,1172,1097,1171,1081,-1983,1157,1112,416,266,375,400,1170,1142,1127,1065,793,793,1169,1033,1156,1096,1141,1111,1155,1080,1126,1140,898,898,808,808,897,897,792,792,1095,1152,1032,1125,1110,1139,1079,1124,882,807,838,881,853,791,-2319,867,368,263,822,852,837,866,806,865,-2399,851,352,262,534,534,821,836,594,594,549,549,593,593,533,533,848,773,579,579,564,578,548,563,276,276,577,576,306,291,516,560,305,305,275,259, |
758 | -251,-892,-2058,-2620,-2828,-2957,-3023,-3039,1041,1041,1040,1040,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,-511,-527,-543,-559,1530,-575,-591,1528,1527,1407,1526,1391,1023,1023,1023,1023,1525,1375,1268,1268,1103,1103,1087,1087,1039,1039,1523,-604,815,815,815,815,510,495,509,479,508,463,507,447,431,505,415,399,-734,-782,1262,-815,1259,1244,-831,1258,1228,-847,-863,1196,-879,1253,987,987,748,-767,493,493,462,477,414,414,686,669,478,446,461,445,474,429,487,458,412,471,1266,1264,1009,1009,799,799,-1019,-1276,-1452,-1581,-1677,-1757,-1821,-1886,-1933,-1997,1257,1257,1483,1468,1512,1422,1497,1406,1467,1496,1421,1510,1134,1134,1225,1225,1466,1451,1374,1405,1252,1252,1358,1480,1164,1164,1251,1251,1238,1238,1389,1465,-1407,1054,1101,-1423,1207,-1439,830,830,1248,1038,1237,1117,1223,1148,1236,1208,411,426,395,410,379,269,1193,1222,1132,1235,1221,1116,976,976,1192,1162,1177,1220,1131,1191,963,963,-1647,961,780,-1663,558,558,994,993,437,408,393,407,829,978,813,797,947,-1743,721,721,377,392,844,950,828,890,706,706,812,859,796,960,948,843,934,874,571,571,-1919,690,555,689,421,346,539,539,944,779,918,873,932,842,903,888,570,570,931,917,674,674,-2575,1562,-2591,1609,-2607,1654,1322,1322,1441,1441,1696,1546,1683,1593,1669,1624,1426,1426,1321,1321,1639,1680,1425,1425,1305,1305,1545,1668,1608,1623,1667,1592,1638,1666,1320,1320,1652,1607,1409,1409,1304,1304,1288,1288,1664,1637,1395,1395,1335,1335,1622,1636,1394,1394,1319,1319,1606,1621,1392,1392,1137,1137,1137,1137,345,390,360,375,404,373,1047,-2751,-2767,-2783,1062,1121,1046,-2799,1077,-2815,1106,1061,789,789,1105,1104,263,355,310,340,325,354,352,262,339,324,1091,1076,1029,1090,1060,1075,833,833,788,788,1088,1028,818,818,803,803,561,561,531,531,816,771,546,546,289,274,288,258, |
759 | -253,-317,-381,-446,-478,-509,1279,1279,-811,-1179,-1451,-1756,-1900,-2028,-2189,-2253,-2333,-2414,-2445,-2511,-2526,1313,1298,-2559,1041,1041,1040,1040,1025,1025,1024,1024,1022,1007,1021,991,1020,975,1019,959,687,687,1018,1017,671,671,655,655,1016,1015,639,639,758,758,623,623,757,607,756,591,755,575,754,559,543,543,1009,783,-575,-621,-685,-749,496,-590,750,749,734,748,974,989,1003,958,988,973,1002,942,987,957,972,1001,926,986,941,971,956,1000,910,985,925,999,894,970,-1071,-1087,-1102,1390,-1135,1436,1509,1451,1374,-1151,1405,1358,1480,1420,-1167,1507,1494,1389,1342,1465,1435,1450,1326,1505,1310,1493,1373,1479,1404,1492,1464,1419,428,443,472,397,736,526,464,464,486,457,442,471,484,482,1357,1449,1434,1478,1388,1491,1341,1490,1325,1489,1463,1403,1309,1477,1372,1448,1418,1433,1476,1356,1462,1387,-1439,1475,1340,1447,1402,1474,1324,1461,1371,1473,269,448,1432,1417,1308,1460,-1711,1459,-1727,1441,1099,1099,1446,1386,1431,1401,-1743,1289,1083,1083,1160,1160,1458,1445,1067,1067,1370,1457,1307,1430,1129,1129,1098,1098,268,432,267,416,266,400,-1887,1144,1187,1082,1173,1113,1186,1066,1050,1158,1128,1143,1172,1097,1171,1081,420,391,1157,1112,1170,1142,1127,1065,1169,1049,1156,1096,1141,1111,1155,1080,1126,1154,1064,1153,1140,1095,1048,-2159,1125,1110,1137,-2175,823,823,1139,1138,807,807,384,264,368,263,868,838,853,791,867,822,852,837,866,806,865,790,-2319,851,821,836,352,262,850,805,849,-2399,533,533,835,820,336,261,578,548,563,577,532,532,832,772,562,562,547,547,305,275,560,515,290,290,288,258 }; |
760 | static const uint8_t tab32[] = { 130,162,193,209,44,28,76,140,9,9,9,9,9,9,9,9,190,254,222,238,126,94,157,157,109,61,173,205 }; |
761 | static const uint8_t tab33[] = { 252,236,220,204,188,172,156,140,124,108,92,76,60,44,28,12 }; |
762 | static const int16_t tabindex[2*16] = { 0,32,64,98,0,132,180,218,292,364,426,538,648,746,0,1126,1460,1460,1460,1460,1460,1460,1460,1460,1842,1842,1842,1842,1842,1842,1842,1842 }; |
763 | static const uint8_t g_linbits[] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,2,3,4,6,8,10,13,4,5,6,7,8,9,11,13 }; |
764 | |
765 | #define PEEK_BITS(n) (bs_cache >> (32 - n)) |
766 | #define FLUSH_BITS(n) { bs_cache <<= (n); bs_sh += (n); } |
767 | #define CHECK_BITS while (bs_sh >= 0) { bs_cache |= (uint32_t)*bs_next_ptr++ << bs_sh; bs_sh -= 8; } |
768 | #define BSPOS ((bs_next_ptr - bs->buf)*8 - 24 + bs_sh) |
769 | |
770 | float one = 0.0f; |
771 | int ireg = 0, big_val_cnt = gr_info->big_values; |
772 | const uint8_t *sfb = gr_info->sfbtab; |
773 | const uint8_t *bs_next_ptr = bs->buf + bs->pos/8; |
774 | uint32_t bs_cache = (((bs_next_ptr[0]*256u + bs_next_ptr[1])*256u + bs_next_ptr[2])*256u + bs_next_ptr[3]) << (bs->pos & 7); |
775 | int pairs_to_decode, np, bs_sh = (bs->pos & 7) - 8; |
776 | bs_next_ptr += 4; |
777 | |
778 | while (big_val_cnt > 0) |
779 | { |
780 | int tab_num = gr_info->table_select[ireg]; |
781 | int sfb_cnt = gr_info->region_count[ireg++]; |
782 | const int16_t *codebook = tabs + tabindex[tab_num]; |
783 | int linbits = g_linbits[tab_num]; |
784 | if (linbits) |
785 | { |
786 | do |
787 | { |
788 | np = *sfb++ / 2; |
789 | pairs_to_decode = MINIMP3_MIN(big_val_cnt, np); |
790 | one = *scf++; |
791 | do |
792 | { |
793 | int j, w = 5; |
794 | int leaf = codebook[PEEK_BITS(w)]; |
795 | while (leaf < 0) |
796 | { |
797 | FLUSH_BITS(w); |
798 | w = leaf & 7; |
799 | leaf = codebook[PEEK_BITS(w) - (leaf >> 3)]; |
800 | } |
801 | FLUSH_BITS(leaf >> 8); |
802 | |
803 | for (j = 0; j < 2; j++, dst++, leaf >>= 4) |
804 | { |
805 | int lsb = leaf & 0x0F; |
806 | if (lsb == 15) |
807 | { |
808 | lsb += PEEK_BITS(linbits); |
809 | FLUSH_BITS(linbits); |
810 | CHECK_BITS; |
811 | *dst = one*L3_pow_43(lsb)*((int32_t)bs_cache < 0 ? -1: 1); |
812 | } else |
813 | { |
814 | *dst = g_pow43[16 + lsb - 16*(bs_cache >> 31)]*one; |
815 | } |
816 | FLUSH_BITS(lsb ? 1 : 0); |
817 | } |
818 | CHECK_BITS; |
819 | } while (--pairs_to_decode); |
820 | } while ((big_val_cnt -= np) > 0 && --sfb_cnt >= 0); |
821 | } else |
822 | { |
823 | do |
824 | { |
825 | np = *sfb++ / 2; |
826 | pairs_to_decode = MINIMP3_MIN(big_val_cnt, np); |
827 | one = *scf++; |
828 | do |
829 | { |
830 | int j, w = 5; |
831 | int leaf = codebook[PEEK_BITS(w)]; |
832 | while (leaf < 0) |
833 | { |
834 | FLUSH_BITS(w); |
835 | w = leaf & 7; |
836 | leaf = codebook[PEEK_BITS(w) - (leaf >> 3)]; |
837 | } |
838 | FLUSH_BITS(leaf >> 8); |
839 | |
840 | for (j = 0; j < 2; j++, dst++, leaf >>= 4) |
841 | { |
842 | int lsb = leaf & 0x0F; |
843 | *dst = g_pow43[16 + lsb - 16*(bs_cache >> 31)]*one; |
844 | FLUSH_BITS(lsb ? 1 : 0); |
845 | } |
846 | CHECK_BITS; |
847 | } while (--pairs_to_decode); |
848 | } while ((big_val_cnt -= np) > 0 && --sfb_cnt >= 0); |
849 | } |
850 | } |
851 | |
852 | for (np = 1 - big_val_cnt;; dst += 4) |
853 | { |
854 | const uint8_t *codebook_count1 = (gr_info->count1_table) ? tab33 : tab32; |
855 | int leaf = codebook_count1[PEEK_BITS(4)]; |
856 | if (!(leaf & 8)) |
857 | { |
858 | leaf = codebook_count1[(leaf >> 3) + (bs_cache << 4 >> (32 - (leaf & 3)))]; |
859 | } |
860 | FLUSH_BITS(leaf & 7); |
861 | if (BSPOS > layer3gr_limit) |
862 | { |
863 | break; |
864 | } |
865 | #define RELOAD_SCALEFACTOR if (!--np) { np = *sfb++/2; if (!np) break; one = *scf++; } |
866 | #define DEQ_COUNT1(s) if (leaf & (128 >> s)) { dst[s] = ((int32_t)bs_cache < 0) ? -one : one; FLUSH_BITS(1) } |
867 | RELOAD_SCALEFACTOR; |
868 | DEQ_COUNT1(0); |
869 | DEQ_COUNT1(1); |
870 | RELOAD_SCALEFACTOR; |
871 | DEQ_COUNT1(2); |
872 | DEQ_COUNT1(3); |
873 | CHECK_BITS; |
874 | } |
875 | |
876 | bs->pos = layer3gr_limit; |
877 | } |
878 | |
879 | static void L3_midside_stereo(float *left, int n) |
880 | { |
881 | int i = 0; |
882 | float *right = left + 576; |
883 | #if HAVE_SIMD |
884 | if (have_simd()) |
885 | { |
886 | for (; i < n - 3; i += 4) |
887 | { |
888 | f4 vl = VLD(left + i); |
889 | f4 vr = VLD(right + i); |
890 | VSTORE(left + i, VADD(vl, vr)); |
891 | VSTORE(right + i, VSUB(vl, vr)); |
892 | } |
893 | #ifdef __GNUC__ |
894 | /* Workaround for spurious -Waggressive-loop-optimizations warning from gcc. |
895 | * For more info see: https://github.com/lieff/minimp3/issues/88 |
896 | */ |
897 | if (__builtin_constant_p(n % 4 == 0) && n % 4 == 0) |
898 | return; |
899 | #endif |
900 | } |
901 | #endif /* HAVE_SIMD */ |
902 | for (; i < n; i++) |
903 | { |
904 | float a = left[i]; |
905 | float b = right[i]; |
906 | left[i] = a + b; |
907 | right[i] = a - b; |
908 | } |
909 | } |
910 | |
911 | static void L3_intensity_stereo_band(float *left, int n, float kl, float kr) |
912 | { |
913 | int i; |
914 | for (i = 0; i < n; i++) |
915 | { |
916 | left[i + 576] = left[i]*kr; |
917 | left[i] = left[i]*kl; |
918 | } |
919 | } |
920 | |
921 | static void L3_stereo_top_band(const float *right, const uint8_t *sfb, int nbands, int max_band[3]) |
922 | { |
923 | int i, k; |
924 | |
925 | max_band[0] = max_band[1] = max_band[2] = -1; |
926 | |
927 | for (i = 0; i < nbands; i++) |
928 | { |
929 | for (k = 0; k < sfb[i]; k += 2) |
930 | { |
931 | if (right[k] != 0 || right[k + 1] != 0) |
932 | { |
933 | max_band[i % 3] = i; |
934 | break; |
935 | } |
936 | } |
937 | right += sfb[i]; |
938 | } |
939 | } |
940 | |
941 | static void L3_stereo_process(float *left, const uint8_t *ist_pos, const uint8_t *sfb, const uint8_t *hdr, int max_band[3], int mpeg2_sh) |
942 | { |
943 | static const float g_pan[7*2] = { 0,1,0.21132487f,0.78867513f,0.36602540f,0.63397460f,0.5f,0.5f,0.63397460f,0.36602540f,0.78867513f,0.21132487f,1,0 }; |
944 | unsigned i, max_pos = HDR_TEST_MPEG1(hdr) ? 7 : 64; |
945 | |
946 | for (i = 0; sfb[i]; i++) |
947 | { |
948 | unsigned ipos = ist_pos[i]; |
949 | if ((int)i > max_band[i % 3] && ipos < max_pos) |
950 | { |
951 | float kl, kr, s = HDR_TEST_MS_STEREO(hdr) ? 1.41421356f : 1; |
952 | if (HDR_TEST_MPEG1(hdr)) |
953 | { |
954 | kl = g_pan[2*ipos]; |
955 | kr = g_pan[2*ipos + 1]; |
956 | } else |
957 | { |
958 | kl = 1; |
959 | kr = L3_ldexp_q2(1, (ipos + 1) >> 1 << mpeg2_sh); |
960 | if (ipos & 1) |
961 | { |
962 | kl = kr; |
963 | kr = 1; |
964 | } |
965 | } |
966 | L3_intensity_stereo_band(left, sfb[i], kl*s, kr*s); |
967 | } else if (HDR_TEST_MS_STEREO(hdr)) |
968 | { |
969 | L3_midside_stereo(left, sfb[i]); |
970 | } |
971 | left += sfb[i]; |
972 | } |
973 | } |
974 | |
975 | static void L3_intensity_stereo(float *left, uint8_t *ist_pos, const L3_gr_info_t *gr, const uint8_t *hdr) |
976 | { |
977 | int max_band[3], n_sfb = gr->n_long_sfb + gr->n_short_sfb; |
978 | int i, max_blocks = gr->n_short_sfb ? 3 : 1; |
979 | |
980 | L3_stereo_top_band(left + 576, gr->sfbtab, n_sfb, max_band); |
981 | if (gr->n_long_sfb) |
982 | { |
983 | max_band[0] = max_band[1] = max_band[2] = MINIMP3_MAX(MINIMP3_MAX(max_band[0], max_band[1]), max_band[2]); |
984 | } |
985 | for (i = 0; i < max_blocks; i++) |
986 | { |
987 | int default_pos = HDR_TEST_MPEG1(hdr) ? 3 : 0; |
988 | int itop = n_sfb - max_blocks + i; |
989 | int prev = itop - max_blocks; |
990 | ist_pos[itop] = max_band[i] >= prev ? default_pos : ist_pos[prev]; |
991 | } |
992 | L3_stereo_process(left, ist_pos, gr->sfbtab, hdr, max_band, gr[1].scalefac_compress & 1); |
993 | } |
994 | |
995 | static void L3_reorder(float *grbuf, float *scratch, const uint8_t *sfb) |
996 | { |
997 | int i, len; |
998 | float *src = grbuf, *dst = scratch; |
999 | |
1000 | for (;0 != (len = *sfb); sfb += 3, src += 2*len) |
1001 | { |
1002 | for (i = 0; i < len; i++, src++) |
1003 | { |
1004 | *dst++ = src[0*len]; |
1005 | *dst++ = src[1*len]; |
1006 | *dst++ = src[2*len]; |
1007 | } |
1008 | } |
1009 | memcpy(grbuf, scratch, (dst - scratch)*sizeof(float)); |
1010 | } |
1011 | |
1012 | static void L3_antialias(float *grbuf, int nbands) |
1013 | { |
1014 | static const float g_aa[2][8] = { |
1015 | {0.85749293f,0.88174200f,0.94962865f,0.98331459f,0.99551782f,0.99916056f,0.99989920f,0.99999316f}, |
1016 | {0.51449576f,0.47173197f,0.31337745f,0.18191320f,0.09457419f,0.04096558f,0.01419856f,0.00369997f} |
1017 | }; |
1018 | |
1019 | for (; nbands > 0; nbands--, grbuf += 18) |
1020 | { |
1021 | int i = 0; |
1022 | #if HAVE_SIMD |
1023 | if (have_simd()) for (; i < 8; i += 4) |
1024 | { |
1025 | f4 vu = VLD(grbuf + 18 + i); |
1026 | f4 vd = VLD(grbuf + 14 - i); |
1027 | f4 vc0 = VLD(g_aa[0] + i); |
1028 | f4 vc1 = VLD(g_aa[1] + i); |
1029 | vd = VREV(vd); |
1030 | VSTORE(grbuf + 18 + i, VSUB(VMUL(vu, vc0), VMUL(vd, vc1))); |
1031 | vd = VADD(VMUL(vu, vc1), VMUL(vd, vc0)); |
1032 | VSTORE(grbuf + 14 - i, VREV(vd)); |
1033 | } |
1034 | #endif /* HAVE_SIMD */ |
1035 | #ifndef MINIMP3_ONLY_SIMD |
1036 | for(; i < 8; i++) |
1037 | { |
1038 | float u = grbuf[18 + i]; |
1039 | float d = grbuf[17 - i]; |
1040 | grbuf[18 + i] = u*g_aa[0][i] - d*g_aa[1][i]; |
1041 | grbuf[17 - i] = u*g_aa[1][i] + d*g_aa[0][i]; |
1042 | } |
1043 | #endif /* MINIMP3_ONLY_SIMD */ |
1044 | } |
1045 | } |
1046 | |
1047 | static void L3_dct3_9(float *y) |
1048 | { |
1049 | float s0, s1, s2, s3, s4, s5, s6, s7, s8, t0, t2, t4; |
1050 | |
1051 | s0 = y[0]; s2 = y[2]; s4 = y[4]; s6 = y[6]; s8 = y[8]; |
1052 | t0 = s0 + s6*0.5f; |
1053 | s0 -= s6; |
1054 | t4 = (s4 + s2)*0.93969262f; |
1055 | t2 = (s8 + s2)*0.76604444f; |
1056 | s6 = (s4 - s8)*0.17364818f; |
1057 | s4 += s8 - s2; |
1058 | |
1059 | s2 = s0 - s4*0.5f; |
1060 | y[4] = s4 + s0; |
1061 | s8 = t0 - t2 + s6; |
1062 | s0 = t0 - t4 + t2; |
1063 | s4 = t0 + t4 - s6; |
1064 | |
1065 | s1 = y[1]; s3 = y[3]; s5 = y[5]; s7 = y[7]; |
1066 | |
1067 | s3 *= 0.86602540f; |
1068 | t0 = (s5 + s1)*0.98480775f; |
1069 | t4 = (s5 - s7)*0.34202014f; |
1070 | t2 = (s1 + s7)*0.64278761f; |
1071 | s1 = (s1 - s5 - s7)*0.86602540f; |
1072 | |
1073 | s5 = t0 - s3 - t2; |
1074 | s7 = t4 - s3 - t0; |
1075 | s3 = t4 + s3 - t2; |
1076 | |
1077 | y[0] = s4 - s7; |
1078 | y[1] = s2 + s1; |
1079 | y[2] = s0 - s3; |
1080 | y[3] = s8 + s5; |
1081 | y[5] = s8 - s5; |
1082 | y[6] = s0 + s3; |
1083 | y[7] = s2 - s1; |
1084 | y[8] = s4 + s7; |
1085 | } |
1086 | |
1087 | static void L3_imdct36(float *grbuf, float *overlap, const float *window, int nbands) |
1088 | { |
1089 | int i, j; |
1090 | static const float g_twid9[18] = { |
1091 | 0.73727734f,0.79335334f,0.84339145f,0.88701083f,0.92387953f,0.95371695f,0.97629601f,0.99144486f,0.99904822f,0.67559021f,0.60876143f,0.53729961f,0.46174861f,0.38268343f,0.30070580f,0.21643961f,0.13052619f,0.04361938f |
1092 | }; |
1093 | |
1094 | for (j = 0; j < nbands; j++, grbuf += 18, overlap += 9) |
1095 | { |
1096 | float co[9], si[9]; |
1097 | co[0] = -grbuf[0]; |
1098 | si[0] = grbuf[17]; |
1099 | for (i = 0; i < 4; i++) |
1100 | { |
1101 | si[8 - 2*i] = grbuf[4*i + 1] - grbuf[4*i + 2]; |
1102 | co[1 + 2*i] = grbuf[4*i + 1] + grbuf[4*i + 2]; |
1103 | si[7 - 2*i] = grbuf[4*i + 4] - grbuf[4*i + 3]; |
1104 | co[2 + 2*i] = -(grbuf[4*i + 3] + grbuf[4*i + 4]); |
1105 | } |
1106 | L3_dct3_9(co); |
1107 | L3_dct3_9(si); |
1108 | |
1109 | si[1] = -si[1]; |
1110 | si[3] = -si[3]; |
1111 | si[5] = -si[5]; |
1112 | si[7] = -si[7]; |
1113 | |
1114 | i = 0; |
1115 | |
1116 | #if HAVE_SIMD |
1117 | if (have_simd()) for (; i < 8; i += 4) |
1118 | { |
1119 | f4 vovl = VLD(overlap + i); |
1120 | f4 vc = VLD(co + i); |
1121 | f4 vs = VLD(si + i); |
1122 | f4 vr0 = VLD(g_twid9 + i); |
1123 | f4 vr1 = VLD(g_twid9 + 9 + i); |
1124 | f4 vw0 = VLD(window + i); |
1125 | f4 vw1 = VLD(window + 9 + i); |
1126 | f4 vsum = VADD(VMUL(vc, vr1), VMUL(vs, vr0)); |
1127 | VSTORE(overlap + i, VSUB(VMUL(vc, vr0), VMUL(vs, vr1))); |
1128 | VSTORE(grbuf + i, VSUB(VMUL(vovl, vw0), VMUL(vsum, vw1))); |
1129 | vsum = VADD(VMUL(vovl, vw1), VMUL(vsum, vw0)); |
1130 | VSTORE(grbuf + 14 - i, VREV(vsum)); |
1131 | } |
1132 | #endif /* HAVE_SIMD */ |
1133 | for (; i < 9; i++) |
1134 | { |
1135 | float ovl = overlap[i]; |
1136 | float sum = co[i]*g_twid9[9 + i] + si[i]*g_twid9[0 + i]; |
1137 | overlap[i] = co[i]*g_twid9[0 + i] - si[i]*g_twid9[9 + i]; |
1138 | grbuf[i] = ovl*window[0 + i] - sum*window[9 + i]; |
1139 | grbuf[17 - i] = ovl*window[9 + i] + sum*window[0 + i]; |
1140 | } |
1141 | } |
1142 | } |
1143 | |
1144 | static void L3_idct3(float x0, float x1, float x2, float *dst) |
1145 | { |
1146 | float m1 = x1*0.86602540f; |
1147 | float a1 = x0 - x2*0.5f; |
1148 | dst[1] = x0 + x2; |
1149 | dst[0] = a1 + m1; |
1150 | dst[2] = a1 - m1; |
1151 | } |
1152 | |
1153 | static void L3_imdct12(float *x, float *dst, float *overlap) |
1154 | { |
1155 | static const float g_twid3[6] = { 0.79335334f,0.92387953f,0.99144486f, 0.60876143f,0.38268343f,0.13052619f }; |
1156 | float co[3], si[3]; |
1157 | int i; |
1158 | |
1159 | L3_idct3(-x[0], x[6] + x[3], x[12] + x[9], co); |
1160 | L3_idct3(x[15], x[12] - x[9], x[6] - x[3], si); |
1161 | si[1] = -si[1]; |
1162 | |
1163 | for (i = 0; i < 3; i++) |
1164 | { |
1165 | float ovl = overlap[i]; |
1166 | float sum = co[i]*g_twid3[3 + i] + si[i]*g_twid3[0 + i]; |
1167 | overlap[i] = co[i]*g_twid3[0 + i] - si[i]*g_twid3[3 + i]; |
1168 | dst[i] = ovl*g_twid3[2 - i] - sum*g_twid3[5 - i]; |
1169 | dst[5 - i] = ovl*g_twid3[5 - i] + sum*g_twid3[2 - i]; |
1170 | } |
1171 | } |
1172 | |
1173 | static void L3_imdct_short(float *grbuf, float *overlap, int nbands) |
1174 | { |
1175 | for (;nbands > 0; nbands--, overlap += 9, grbuf += 18) |
1176 | { |
1177 | float tmp[18]; |
1178 | memcpy(tmp, grbuf, sizeof(tmp)); |
1179 | memcpy(grbuf, overlap, 6*sizeof(float)); |
1180 | L3_imdct12(tmp, grbuf + 6, overlap + 6); |
1181 | L3_imdct12(tmp + 1, grbuf + 12, overlap + 6); |
1182 | L3_imdct12(tmp + 2, overlap, overlap + 6); |
1183 | } |
1184 | } |
1185 | |
1186 | static void L3_change_sign(float *grbuf) |
1187 | { |
1188 | int b, i; |
1189 | for (b = 0, grbuf += 18; b < 32; b += 2, grbuf += 36) |
1190 | for (i = 1; i < 18; i += 2) |
1191 | grbuf[i] = -grbuf[i]; |
1192 | } |
1193 | |
1194 | static void L3_imdct_gr(float *grbuf, float *overlap, unsigned block_type, unsigned n_long_bands) |
1195 | { |
1196 | static const float g_mdct_window[2][18] = { |
1197 | { 0.99904822f,0.99144486f,0.97629601f,0.95371695f,0.92387953f,0.88701083f,0.84339145f,0.79335334f,0.73727734f,0.04361938f,0.13052619f,0.21643961f,0.30070580f,0.38268343f,0.46174861f,0.53729961f,0.60876143f,0.67559021f }, |
1198 | { 1,1,1,1,1,1,0.99144486f,0.92387953f,0.79335334f,0,0,0,0,0,0,0.13052619f,0.38268343f,0.60876143f } |
1199 | }; |
1200 | if (n_long_bands) |
1201 | { |
1202 | L3_imdct36(grbuf, overlap, g_mdct_window[0], n_long_bands); |
1203 | grbuf += 18*n_long_bands; |
1204 | overlap += 9*n_long_bands; |
1205 | } |
1206 | if (block_type == SHORT_BLOCK_TYPE) |
1207 | L3_imdct_short(grbuf, overlap, 32 - n_long_bands); |
1208 | else |
1209 | L3_imdct36(grbuf, overlap, g_mdct_window[block_type == STOP_BLOCK_TYPE], 32 - n_long_bands); |
1210 | } |
1211 | |
1212 | static void L3_save_reservoir(mp3dec_t *h, mp3dec_scratch_t *s) |
1213 | { |
1214 | int pos = (s->bs.pos + 7)/8u; |
1215 | int remains = s->bs.limit/8u - pos; |
1216 | if (remains > MAX_BITRESERVOIR_BYTES) |
1217 | { |
1218 | pos += remains - MAX_BITRESERVOIR_BYTES; |
1219 | remains = MAX_BITRESERVOIR_BYTES; |
1220 | } |
1221 | if (remains > 0) |
1222 | { |
1223 | memmove(h->reserv_buf, s->maindata + pos, remains); |
1224 | } |
1225 | h->reserv = remains; |
1226 | } |
1227 | |
1228 | static int L3_restore_reservoir(mp3dec_t *h, bs_t *bs, mp3dec_scratch_t *s, int main_data_begin) |
1229 | { |
1230 | int frame_bytes = (bs->limit - bs->pos)/8; |
1231 | int bytes_have = MINIMP3_MIN(h->reserv, main_data_begin); |
1232 | memcpy(s->maindata, h->reserv_buf + MINIMP3_MAX(0, h->reserv - main_data_begin), MINIMP3_MIN(h->reserv, main_data_begin)); |
1233 | memcpy(s->maindata + bytes_have, bs->buf + bs->pos/8, frame_bytes); |
1234 | bs_init(&s->bs, s->maindata, bytes_have + frame_bytes); |
1235 | return h->reserv >= main_data_begin; |
1236 | } |
1237 | |
1238 | static void L3_decode(mp3dec_t *h, mp3dec_scratch_t *s, L3_gr_info_t *gr_info, int nch) |
1239 | { |
1240 | int ch; |
1241 | |
1242 | for (ch = 0; ch < nch; ch++) |
1243 | { |
1244 | int layer3gr_limit = s->bs.pos + gr_info[ch].part_23_length; |
1245 | L3_decode_scalefactors(h->header, s->ist_pos[ch], &s->bs, gr_info + ch, s->scf, ch); |
1246 | L3_huffman(s->grbuf[ch], &s->bs, gr_info + ch, s->scf, layer3gr_limit); |
1247 | } |
1248 | |
1249 | if (HDR_TEST_I_STEREO(h->header)) |
1250 | { |
1251 | L3_intensity_stereo(s->grbuf[0], s->ist_pos[1], gr_info, h->header); |
1252 | } else if (HDR_IS_MS_STEREO(h->header)) |
1253 | { |
1254 | L3_midside_stereo(s->grbuf[0], 576); |
1255 | } |
1256 | |
1257 | for (ch = 0; ch < nch; ch++, gr_info++) |
1258 | { |
1259 | int aa_bands = 31; |
1260 | int n_long_bands = (gr_info->mixed_block_flag ? 2 : 0) << (int)(HDR_GET_MY_SAMPLE_RATE(h->header) == 2); |
1261 | |
1262 | if (gr_info->n_short_sfb) |
1263 | { |
1264 | aa_bands = n_long_bands - 1; |
1265 | L3_reorder(s->grbuf[ch] + n_long_bands*18, s->syn[0], gr_info->sfbtab + gr_info->n_long_sfb); |
1266 | } |
1267 | |
1268 | L3_antialias(s->grbuf[ch], aa_bands); |
1269 | L3_imdct_gr(s->grbuf[ch], h->mdct_overlap[ch], gr_info->block_type, n_long_bands); |
1270 | L3_change_sign(s->grbuf[ch]); |
1271 | } |
1272 | } |
1273 | |
1274 | static void mp3d_DCT_II(float *grbuf, int n) |
1275 | { |
1276 | static const float g_sec[24] = { |
1277 | 10.19000816f,0.50060302f,0.50241929f,3.40760851f,0.50547093f,0.52249861f,2.05778098f,0.51544732f,0.56694406f,1.48416460f,0.53104258f,0.64682180f,1.16943991f,0.55310392f,0.78815460f,0.97256821f,0.58293498f,1.06067765f,0.83934963f,0.62250412f,1.72244716f,0.74453628f,0.67480832f,5.10114861f |
1278 | }; |
1279 | int i, k = 0; |
1280 | #if HAVE_SIMD |
1281 | if (have_simd()) for (; k < n; k += 4) |
1282 | { |
1283 | f4 t[4][8], *x; |
1284 | float *y = grbuf + k; |
1285 | |
1286 | for (x = t[0], i = 0; i < 8; i++, x++) |
1287 | { |
1288 | f4 x0 = VLD(&y[i*18]); |
1289 | f4 x1 = VLD(&y[(15 - i)*18]); |
1290 | f4 x2 = VLD(&y[(16 + i)*18]); |
1291 | f4 x3 = VLD(&y[(31 - i)*18]); |
1292 | f4 t0 = VADD(x0, x3); |
1293 | f4 t1 = VADD(x1, x2); |
1294 | f4 t2 = VMUL_S(VSUB(x1, x2), g_sec[3*i + 0]); |
1295 | f4 t3 = VMUL_S(VSUB(x0, x3), g_sec[3*i + 1]); |
1296 | x[0] = VADD(t0, t1); |
1297 | x[8] = VMUL_S(VSUB(t0, t1), g_sec[3*i + 2]); |
1298 | x[16] = VADD(t3, t2); |
1299 | x[24] = VMUL_S(VSUB(t3, t2), g_sec[3*i + 2]); |
1300 | } |
1301 | for (x = t[0], i = 0; i < 4; i++, x += 8) |
1302 | { |
1303 | f4 x0 = x[0], x1 = x[1], x2 = x[2], x3 = x[3], x4 = x[4], x5 = x[5], x6 = x[6], x7 = x[7], xt; |
1304 | xt = VSUB(x0, x7); x0 = VADD(x0, x7); |
1305 | x7 = VSUB(x1, x6); x1 = VADD(x1, x6); |
1306 | x6 = VSUB(x2, x5); x2 = VADD(x2, x5); |
1307 | x5 = VSUB(x3, x4); x3 = VADD(x3, x4); |
1308 | x4 = VSUB(x0, x3); x0 = VADD(x0, x3); |
1309 | x3 = VSUB(x1, x2); x1 = VADD(x1, x2); |
1310 | x[0] = VADD(x0, x1); |
1311 | x[4] = VMUL_S(VSUB(x0, x1), 0.70710677f); |
1312 | x5 = VADD(x5, x6); |
1313 | x6 = VMUL_S(VADD(x6, x7), 0.70710677f); |
1314 | x7 = VADD(x7, xt); |
1315 | x3 = VMUL_S(VADD(x3, x4), 0.70710677f); |
1316 | x5 = VSUB(x5, VMUL_S(x7, 0.198912367f)); /* rotate by PI/8 */ |
1317 | x7 = VADD(x7, VMUL_S(x5, 0.382683432f)); |
1318 | x5 = VSUB(x5, VMUL_S(x7, 0.198912367f)); |
1319 | x0 = VSUB(xt, x6); xt = VADD(xt, x6); |
1320 | x[1] = VMUL_S(VADD(xt, x7), 0.50979561f); |
1321 | x[2] = VMUL_S(VADD(x4, x3), 0.54119611f); |
1322 | x[3] = VMUL_S(VSUB(x0, x5), 0.60134488f); |
1323 | x[5] = VMUL_S(VADD(x0, x5), 0.89997619f); |
1324 | x[6] = VMUL_S(VSUB(x4, x3), 1.30656302f); |
1325 | x[7] = VMUL_S(VSUB(xt, x7), 2.56291556f); |
1326 | } |
1327 | |
1328 | if (k > n - 3) |
1329 | { |
1330 | #if HAVE_SSE |
1331 | #define VSAVE2(i, v) _mm_storel_pi((__m64 *)(void*)&y[i*18], v) |
1332 | #else /* HAVE_SSE */ |
1333 | #define VSAVE2(i, v) vst1_f32((float32_t *)&y[i*18], vget_low_f32(v)) |
1334 | #endif /* HAVE_SSE */ |
1335 | for (i = 0; i < 7; i++, y += 4*18) |
1336 | { |
1337 | f4 s = VADD(t[3][i], t[3][i + 1]); |
1338 | VSAVE2(0, t[0][i]); |
1339 | VSAVE2(1, VADD(t[2][i], s)); |
1340 | VSAVE2(2, VADD(t[1][i], t[1][i + 1])); |
1341 | VSAVE2(3, VADD(t[2][1 + i], s)); |
1342 | } |
1343 | VSAVE2(0, t[0][7]); |
1344 | VSAVE2(1, VADD(t[2][7], t[3][7])); |
1345 | VSAVE2(2, t[1][7]); |
1346 | VSAVE2(3, t[3][7]); |
1347 | } else |
1348 | { |
1349 | #define VSAVE4(i, v) VSTORE(&y[i*18], v) |
1350 | for (i = 0; i < 7; i++, y += 4*18) |
1351 | { |
1352 | f4 s = VADD(t[3][i], t[3][i + 1]); |
1353 | VSAVE4(0, t[0][i]); |
1354 | VSAVE4(1, VADD(t[2][i], s)); |
1355 | VSAVE4(2, VADD(t[1][i], t[1][i + 1])); |
1356 | VSAVE4(3, VADD(t[2][1 + i], s)); |
1357 | } |
1358 | VSAVE4(0, t[0][7]); |
1359 | VSAVE4(1, VADD(t[2][7], t[3][7])); |
1360 | VSAVE4(2, t[1][7]); |
1361 | VSAVE4(3, t[3][7]); |
1362 | } |
1363 | } else |
1364 | #endif /* HAVE_SIMD */ |
1365 | #ifdef MINIMP3_ONLY_SIMD |
1366 | {} /* for HAVE_SIMD=1, MINIMP3_ONLY_SIMD=1 case we do not need non-intrinsic "else" branch */ |
1367 | #else /* MINIMP3_ONLY_SIMD */ |
1368 | for (; k < n; k++) |
1369 | { |
1370 | float t[4][8], *x, *y = grbuf + k; |
1371 | |
1372 | for (x = t[0], i = 0; i < 8; i++, x++) |
1373 | { |
1374 | float x0 = y[i*18]; |
1375 | float x1 = y[(15 - i)*18]; |
1376 | float x2 = y[(16 + i)*18]; |
1377 | float x3 = y[(31 - i)*18]; |
1378 | float t0 = x0 + x3; |
1379 | float t1 = x1 + x2; |
1380 | float t2 = (x1 - x2)*g_sec[3*i + 0]; |
1381 | float t3 = (x0 - x3)*g_sec[3*i + 1]; |
1382 | x[0] = t0 + t1; |
1383 | x[8] = (t0 - t1)*g_sec[3*i + 2]; |
1384 | x[16] = t3 + t2; |
1385 | x[24] = (t3 - t2)*g_sec[3*i + 2]; |
1386 | } |
1387 | for (x = t[0], i = 0; i < 4; i++, x += 8) |
1388 | { |
1389 | float x0 = x[0], x1 = x[1], x2 = x[2], x3 = x[3], x4 = x[4], x5 = x[5], x6 = x[6], x7 = x[7], xt; |
1390 | xt = x0 - x7; x0 += x7; |
1391 | x7 = x1 - x6; x1 += x6; |
1392 | x6 = x2 - x5; x2 += x5; |
1393 | x5 = x3 - x4; x3 += x4; |
1394 | x4 = x0 - x3; x0 += x3; |
1395 | x3 = x1 - x2; x1 += x2; |
1396 | x[0] = x0 + x1; |
1397 | x[4] = (x0 - x1)*0.70710677f; |
1398 | x5 = x5 + x6; |
1399 | x6 = (x6 + x7)*0.70710677f; |
1400 | x7 = x7 + xt; |
1401 | x3 = (x3 + x4)*0.70710677f; |
1402 | x5 -= x7*0.198912367f; /* rotate by PI/8 */ |
1403 | x7 += x5*0.382683432f; |
1404 | x5 -= x7*0.198912367f; |
1405 | x0 = xt - x6; xt += x6; |
1406 | x[1] = (xt + x7)*0.50979561f; |
1407 | x[2] = (x4 + x3)*0.54119611f; |
1408 | x[3] = (x0 - x5)*0.60134488f; |
1409 | x[5] = (x0 + x5)*0.89997619f; |
1410 | x[6] = (x4 - x3)*1.30656302f; |
1411 | x[7] = (xt - x7)*2.56291556f; |
1412 | |
1413 | } |
1414 | for (i = 0; i < 7; i++, y += 4*18) |
1415 | { |
1416 | y[0*18] = t[0][i]; |
1417 | y[1*18] = t[2][i] + t[3][i] + t[3][i + 1]; |
1418 | y[2*18] = t[1][i] + t[1][i + 1]; |
1419 | y[3*18] = t[2][i + 1] + t[3][i] + t[3][i + 1]; |
1420 | } |
1421 | y[0*18] = t[0][7]; |
1422 | y[1*18] = t[2][7] + t[3][7]; |
1423 | y[2*18] = t[1][7]; |
1424 | y[3*18] = t[3][7]; |
1425 | } |
1426 | #endif /* MINIMP3_ONLY_SIMD */ |
1427 | } |
1428 | |
1429 | #ifndef MINIMP3_FLOAT_OUTPUT |
1430 | static int16_t mp3d_scale_pcm(float sample) |
1431 | { |
1432 | #if HAVE_ARMV6 |
1433 | int32_t s32 = (int32_t)(sample + .5f); |
1434 | s32 -= (s32 < 0); |
1435 | int16_t s = (int16_t)minimp3_clip_int16_arm(s32); |
1436 | #else |
1437 | if (sample >= 32766.5) return (int16_t) 32767; |
1438 | if (sample <= -32767.5) return (int16_t)-32768; |
1439 | int16_t s = (int16_t)(sample + .5f); |
1440 | s -= (s < 0); /* away from zero, to be compliant */ |
1441 | #endif |
1442 | return s; |
1443 | } |
1444 | #else /* MINIMP3_FLOAT_OUTPUT */ |
1445 | static float mp3d_scale_pcm(float sample) |
1446 | { |
1447 | return sample*(1.f/32768.f); |
1448 | } |
1449 | #endif /* MINIMP3_FLOAT_OUTPUT */ |
1450 | |
1451 | static void mp3d_synth_pair(mp3d_sample_t *pcm, int nch, const float *z) |
1452 | { |
1453 | float a; |
1454 | a = (z[14*64] - z[ 0]) * 29; |
1455 | a += (z[ 1*64] + z[13*64]) * 213; |
1456 | a += (z[12*64] - z[ 2*64]) * 459; |
1457 | a += (z[ 3*64] + z[11*64]) * 2037; |
1458 | a += (z[10*64] - z[ 4*64]) * 5153; |
1459 | a += (z[ 5*64] + z[ 9*64]) * 6574; |
1460 | a += (z[ 8*64] - z[ 6*64]) * 37489; |
1461 | a += z[ 7*64] * 75038; |
1462 | pcm[0] = mp3d_scale_pcm(a); |
1463 | |
1464 | z += 2; |
1465 | a = z[14*64] * 104; |
1466 | a += z[12*64] * 1567; |
1467 | a += z[10*64] * 9727; |
1468 | a += z[ 8*64] * 64019; |
1469 | a += z[ 6*64] * -9975; |
1470 | a += z[ 4*64] * -45; |
1471 | a += z[ 2*64] * 146; |
1472 | a += z[ 0*64] * -5; |
1473 | pcm[16*nch] = mp3d_scale_pcm(a); |
1474 | } |
1475 | |
1476 | static void mp3d_synth(float *xl, mp3d_sample_t *dstl, int nch, float *lins) |
1477 | { |
1478 | int i; |
1479 | float *xr = xl + 576*(nch - 1); |
1480 | mp3d_sample_t *dstr = dstl + (nch - 1); |
1481 | |
1482 | static const float g_win[] = { |
1483 | -1,26,-31,208,218,401,-519,2063,2000,4788,-5517,7134,5959,35640,-39336,74992, |
1484 | -1,24,-35,202,222,347,-581,2080,1952,4425,-5879,7640,5288,33791,-41176,74856, |
1485 | -1,21,-38,196,225,294,-645,2087,1893,4063,-6237,8092,4561,31947,-43006,74630, |
1486 | -1,19,-41,190,227,244,-711,2085,1822,3705,-6589,8492,3776,30112,-44821,74313, |
1487 | -1,17,-45,183,228,197,-779,2075,1739,3351,-6935,8840,2935,28289,-46617,73908, |
1488 | -1,16,-49,176,228,153,-848,2057,1644,3004,-7271,9139,2037,26482,-48390,73415, |
1489 | -2,14,-53,169,227,111,-919,2032,1535,2663,-7597,9389,1082,24694,-50137,72835, |
1490 | -2,13,-58,161,224,72,-991,2001,1414,2330,-7910,9592,70,22929,-51853,72169, |
1491 | -2,11,-63,154,221,36,-1064,1962,1280,2006,-8209,9750,-998,21189,-53534,71420, |
1492 | -2,10,-68,147,215,2,-1137,1919,1131,1692,-8491,9863,-2122,19478,-55178,70590, |
1493 | -3,9,-73,139,208,-29,-1210,1870,970,1388,-8755,9935,-3300,17799,-56778,69679, |
1494 | -3,8,-79,132,200,-57,-1283,1817,794,1095,-8998,9966,-4533,16155,-58333,68692, |
1495 | -4,7,-85,125,189,-83,-1356,1759,605,814,-9219,9959,-5818,14548,-59838,67629, |
1496 | -4,7,-91,117,177,-106,-1428,1698,402,545,-9416,9916,-7154,12980,-61289,66494, |
1497 | -5,6,-97,111,163,-127,-1498,1634,185,288,-9585,9838,-8540,11455,-62684,65290 |
1498 | }; |
1499 | float *zlin = lins + 15*64; |
1500 | const float *w = g_win; |
1501 | |
1502 | zlin[4*15] = xl[18*16]; |
1503 | zlin[4*15 + 1] = xr[18*16]; |
1504 | zlin[4*15 + 2] = xl[0]; |
1505 | zlin[4*15 + 3] = xr[0]; |
1506 | |
1507 | zlin[4*31] = xl[1 + 18*16]; |
1508 | zlin[4*31 + 1] = xr[1 + 18*16]; |
1509 | zlin[4*31 + 2] = xl[1]; |
1510 | zlin[4*31 + 3] = xr[1]; |
1511 | |
1512 | mp3d_synth_pair(dstr, nch, lins + 4*15 + 1); |
1513 | mp3d_synth_pair(dstr + 32*nch, nch, lins + 4*15 + 64 + 1); |
1514 | mp3d_synth_pair(dstl, nch, lins + 4*15); |
1515 | mp3d_synth_pair(dstl + 32*nch, nch, lins + 4*15 + 64); |
1516 | |
1517 | #if HAVE_SIMD |
1518 | if (have_simd()) for (i = 14; i >= 0; i--) |
1519 | { |
1520 | #define VLOAD(k) f4 w0 = VSET(*w++); f4 w1 = VSET(*w++); f4 vz = VLD(&zlin[4*i - 64*k]); f4 vy = VLD(&zlin[4*i - 64*(15 - k)]); |
1521 | #define V0(k) { VLOAD(k) b = VADD(VMUL(vz, w1), VMUL(vy, w0)) ; a = VSUB(VMUL(vz, w0), VMUL(vy, w1)); } |
1522 | #define V1(k) { VLOAD(k) b = VADD(b, VADD(VMUL(vz, w1), VMUL(vy, w0))); a = VADD(a, VSUB(VMUL(vz, w0), VMUL(vy, w1))); } |
1523 | #define V2(k) { VLOAD(k) b = VADD(b, VADD(VMUL(vz, w1), VMUL(vy, w0))); a = VADD(a, VSUB(VMUL(vy, w1), VMUL(vz, w0))); } |
1524 | f4 a, b; |
1525 | zlin[4*i] = xl[18*(31 - i)]; |
1526 | zlin[4*i + 1] = xr[18*(31 - i)]; |
1527 | zlin[4*i + 2] = xl[1 + 18*(31 - i)]; |
1528 | zlin[4*i + 3] = xr[1 + 18*(31 - i)]; |
1529 | zlin[4*i + 64] = xl[1 + 18*(1 + i)]; |
1530 | zlin[4*i + 64 + 1] = xr[1 + 18*(1 + i)]; |
1531 | zlin[4*i - 64 + 2] = xl[18*(1 + i)]; |
1532 | zlin[4*i - 64 + 3] = xr[18*(1 + i)]; |
1533 | |
1534 | V0(0) V2(1) V1(2) V2(3) V1(4) V2(5) V1(6) V2(7) |
1535 | |
1536 | { |
1537 | #ifndef MINIMP3_FLOAT_OUTPUT |
1538 | #if HAVE_SSE |
1539 | static const f4 g_max = { 32767.0f, 32767.0f, 32767.0f, 32767.0f }; |
1540 | static const f4 g_min = { -32768.0f, -32768.0f, -32768.0f, -32768.0f }; |
1541 | __m128i pcm8 = _mm_packs_epi32(_mm_cvtps_epi32(_mm_max_ps(_mm_min_ps(a, g_max), g_min)), |
1542 | _mm_cvtps_epi32(_mm_max_ps(_mm_min_ps(b, g_max), g_min))); |
1543 | dstr[(15 - i)*nch] = _mm_extract_epi16(pcm8, 1); |
1544 | dstr[(17 + i)*nch] = _mm_extract_epi16(pcm8, 5); |
1545 | dstl[(15 - i)*nch] = _mm_extract_epi16(pcm8, 0); |
1546 | dstl[(17 + i)*nch] = _mm_extract_epi16(pcm8, 4); |
1547 | dstr[(47 - i)*nch] = _mm_extract_epi16(pcm8, 3); |
1548 | dstr[(49 + i)*nch] = _mm_extract_epi16(pcm8, 7); |
1549 | dstl[(47 - i)*nch] = _mm_extract_epi16(pcm8, 2); |
1550 | dstl[(49 + i)*nch] = _mm_extract_epi16(pcm8, 6); |
1551 | #else /* HAVE_SSE */ |
1552 | int16x4_t pcma, pcmb; |
1553 | a = VADD(a, VSET(0.5f)); |
1554 | b = VADD(b, VSET(0.5f)); |
1555 | pcma = vqmovn_s32(vqaddq_s32(vcvtq_s32_f32(a), vreinterpretq_s32_u32(vcltq_f32(a, VSET(0))))); |
1556 | pcmb = vqmovn_s32(vqaddq_s32(vcvtq_s32_f32(b), vreinterpretq_s32_u32(vcltq_f32(b, VSET(0))))); |
1557 | vst1_lane_s16(dstr + (15 - i)*nch, pcma, 1); |
1558 | vst1_lane_s16(dstr + (17 + i)*nch, pcmb, 1); |
1559 | vst1_lane_s16(dstl + (15 - i)*nch, pcma, 0); |
1560 | vst1_lane_s16(dstl + (17 + i)*nch, pcmb, 0); |
1561 | vst1_lane_s16(dstr + (47 - i)*nch, pcma, 3); |
1562 | vst1_lane_s16(dstr + (49 + i)*nch, pcmb, 3); |
1563 | vst1_lane_s16(dstl + (47 - i)*nch, pcma, 2); |
1564 | vst1_lane_s16(dstl + (49 + i)*nch, pcmb, 2); |
1565 | #endif /* HAVE_SSE */ |
1566 | |
1567 | #else /* MINIMP3_FLOAT_OUTPUT */ |
1568 | |
1569 | // -- GODOT start -- |
1570 | #if defined(_MSC_VER) && (defined(_M_ARM64) || defined(_M_ARM64EC) || defined(_M_ARM)) |
1571 | static f4 g_scale; |
1572 | g_scale = vsetq_lane_f32(1.0f/32768.0f, g_scale, 0); |
1573 | g_scale = vsetq_lane_f32(1.0f/32768.0f, g_scale, 1); |
1574 | g_scale = vsetq_lane_f32(1.0f/32768.0f, g_scale, 2); |
1575 | g_scale = vsetq_lane_f32(1.0f/32768.0f, g_scale, 3); |
1576 | #else |
1577 | static const f4 g_scale = { 1.0f/32768.0f, 1.0f/32768.0f, 1.0f/32768.0f, 1.0f/32768.0f }; |
1578 | #endif |
1579 | // -- GODOT end -- |
1580 | |
1581 | a = VMUL(a, g_scale); |
1582 | b = VMUL(b, g_scale); |
1583 | #if HAVE_SSE |
1584 | _mm_store_ss(dstr + (15 - i)*nch, _mm_shuffle_ps(a, a, _MM_SHUFFLE(1, 1, 1, 1))); |
1585 | _mm_store_ss(dstr + (17 + i)*nch, _mm_shuffle_ps(b, b, _MM_SHUFFLE(1, 1, 1, 1))); |
1586 | _mm_store_ss(dstl + (15 - i)*nch, _mm_shuffle_ps(a, a, _MM_SHUFFLE(0, 0, 0, 0))); |
1587 | _mm_store_ss(dstl + (17 + i)*nch, _mm_shuffle_ps(b, b, _MM_SHUFFLE(0, 0, 0, 0))); |
1588 | _mm_store_ss(dstr + (47 - i)*nch, _mm_shuffle_ps(a, a, _MM_SHUFFLE(3, 3, 3, 3))); |
1589 | _mm_store_ss(dstr + (49 + i)*nch, _mm_shuffle_ps(b, b, _MM_SHUFFLE(3, 3, 3, 3))); |
1590 | _mm_store_ss(dstl + (47 - i)*nch, _mm_shuffle_ps(a, a, _MM_SHUFFLE(2, 2, 2, 2))); |
1591 | _mm_store_ss(dstl + (49 + i)*nch, _mm_shuffle_ps(b, b, _MM_SHUFFLE(2, 2, 2, 2))); |
1592 | #else /* HAVE_SSE */ |
1593 | vst1q_lane_f32(dstr + (15 - i)*nch, a, 1); |
1594 | vst1q_lane_f32(dstr + (17 + i)*nch, b, 1); |
1595 | vst1q_lane_f32(dstl + (15 - i)*nch, a, 0); |
1596 | vst1q_lane_f32(dstl + (17 + i)*nch, b, 0); |
1597 | vst1q_lane_f32(dstr + (47 - i)*nch, a, 3); |
1598 | vst1q_lane_f32(dstr + (49 + i)*nch, b, 3); |
1599 | vst1q_lane_f32(dstl + (47 - i)*nch, a, 2); |
1600 | vst1q_lane_f32(dstl + (49 + i)*nch, b, 2); |
1601 | #endif /* HAVE_SSE */ |
1602 | #endif /* MINIMP3_FLOAT_OUTPUT */ |
1603 | } |
1604 | } else |
1605 | #endif /* HAVE_SIMD */ |
1606 | #ifdef MINIMP3_ONLY_SIMD |
1607 | {} /* for HAVE_SIMD=1, MINIMP3_ONLY_SIMD=1 case we do not need non-intrinsic "else" branch */ |
1608 | #else /* MINIMP3_ONLY_SIMD */ |
1609 | for (i = 14; i >= 0; i--) |
1610 | { |
1611 | #define LOAD(k) float w0 = *w++; float w1 = *w++; float *vz = &zlin[4*i - k*64]; float *vy = &zlin[4*i - (15 - k)*64]; |
1612 | #define S0(k) { int j; LOAD(k); for (j = 0; j < 4; j++) b[j] = vz[j]*w1 + vy[j]*w0, a[j] = vz[j]*w0 - vy[j]*w1; } |
1613 | #define S1(k) { int j; LOAD(k); for (j = 0; j < 4; j++) b[j] += vz[j]*w1 + vy[j]*w0, a[j] += vz[j]*w0 - vy[j]*w1; } |
1614 | #define S2(k) { int j; LOAD(k); for (j = 0; j < 4; j++) b[j] += vz[j]*w1 + vy[j]*w0, a[j] += vy[j]*w1 - vz[j]*w0; } |
1615 | float a[4], b[4]; |
1616 | |
1617 | zlin[4*i] = xl[18*(31 - i)]; |
1618 | zlin[4*i + 1] = xr[18*(31 - i)]; |
1619 | zlin[4*i + 2] = xl[1 + 18*(31 - i)]; |
1620 | zlin[4*i + 3] = xr[1 + 18*(31 - i)]; |
1621 | zlin[4*(i + 16)] = xl[1 + 18*(1 + i)]; |
1622 | zlin[4*(i + 16) + 1] = xr[1 + 18*(1 + i)]; |
1623 | zlin[4*(i - 16) + 2] = xl[18*(1 + i)]; |
1624 | zlin[4*(i - 16) + 3] = xr[18*(1 + i)]; |
1625 | |
1626 | S0(0) S2(1) S1(2) S2(3) S1(4) S2(5) S1(6) S2(7) |
1627 | |
1628 | dstr[(15 - i)*nch] = mp3d_scale_pcm(a[1]); |
1629 | dstr[(17 + i)*nch] = mp3d_scale_pcm(b[1]); |
1630 | dstl[(15 - i)*nch] = mp3d_scale_pcm(a[0]); |
1631 | dstl[(17 + i)*nch] = mp3d_scale_pcm(b[0]); |
1632 | dstr[(47 - i)*nch] = mp3d_scale_pcm(a[3]); |
1633 | dstr[(49 + i)*nch] = mp3d_scale_pcm(b[3]); |
1634 | dstl[(47 - i)*nch] = mp3d_scale_pcm(a[2]); |
1635 | dstl[(49 + i)*nch] = mp3d_scale_pcm(b[2]); |
1636 | } |
1637 | #endif /* MINIMP3_ONLY_SIMD */ |
1638 | } |
1639 | |
1640 | static void mp3d_synth_granule(float *qmf_state, float *grbuf, int nbands, int nch, mp3d_sample_t *pcm, float *lins) |
1641 | { |
1642 | int i; |
1643 | for (i = 0; i < nch; i++) |
1644 | { |
1645 | mp3d_DCT_II(grbuf + 576*i, nbands); |
1646 | } |
1647 | |
1648 | memcpy(lins, qmf_state, sizeof(float)*15*64); |
1649 | |
1650 | for (i = 0; i < nbands; i += 2) |
1651 | { |
1652 | mp3d_synth(grbuf + i, pcm + 32*nch*i, nch, lins + i*64); |
1653 | } |
1654 | #ifndef MINIMP3_NONSTANDARD_BUT_LOGICAL |
1655 | if (nch == 1) |
1656 | { |
1657 | for (i = 0; i < 15*64; i += 2) |
1658 | { |
1659 | qmf_state[i] = lins[nbands*64 + i]; |
1660 | } |
1661 | } else |
1662 | #endif /* MINIMP3_NONSTANDARD_BUT_LOGICAL */ |
1663 | { |
1664 | memcpy(qmf_state, lins + nbands*64, sizeof(float)*15*64); |
1665 | } |
1666 | } |
1667 | |
1668 | static int mp3d_match_frame(const uint8_t *hdr, int mp3_bytes, int frame_bytes) |
1669 | { |
1670 | int i, nmatch; |
1671 | for (i = 0, nmatch = 0; nmatch < MAX_FRAME_SYNC_MATCHES; nmatch++) |
1672 | { |
1673 | i += hdr_frame_bytes(hdr + i, frame_bytes) + hdr_padding(hdr + i); |
1674 | if (i + HDR_SIZE > mp3_bytes) |
1675 | return nmatch > 0; |
1676 | if (!hdr_compare(hdr, hdr + i)) |
1677 | return 0; |
1678 | } |
1679 | return 1; |
1680 | } |
1681 | |
1682 | static int mp3d_find_frame(const uint8_t *mp3, int mp3_bytes, int *free_format_bytes, int *ptr_frame_bytes) |
1683 | { |
1684 | int i, k; |
1685 | for (i = 0; i < mp3_bytes - HDR_SIZE; i++, mp3++) |
1686 | { |
1687 | if (hdr_valid(mp3)) |
1688 | { |
1689 | int frame_bytes = hdr_frame_bytes(mp3, *free_format_bytes); |
1690 | int frame_and_padding = frame_bytes + hdr_padding(mp3); |
1691 | |
1692 | for (k = HDR_SIZE; !frame_bytes && k < MAX_FREE_FORMAT_FRAME_SIZE && i + 2*k < mp3_bytes - HDR_SIZE; k++) |
1693 | { |
1694 | if (hdr_compare(mp3, mp3 + k)) |
1695 | { |
1696 | int fb = k - hdr_padding(mp3); |
1697 | int nextfb = fb + hdr_padding(mp3 + k); |
1698 | if (i + k + nextfb + HDR_SIZE > mp3_bytes || !hdr_compare(mp3, mp3 + k + nextfb)) |
1699 | continue; |
1700 | frame_and_padding = k; |
1701 | frame_bytes = fb; |
1702 | *free_format_bytes = fb; |
1703 | } |
1704 | } |
1705 | if ((frame_bytes && i + frame_and_padding <= mp3_bytes && |
1706 | mp3d_match_frame(mp3, mp3_bytes - i, frame_bytes)) || |
1707 | (!i && frame_and_padding == mp3_bytes)) |
1708 | { |
1709 | *ptr_frame_bytes = frame_and_padding; |
1710 | return i; |
1711 | } |
1712 | *free_format_bytes = 0; |
1713 | } |
1714 | } |
1715 | *ptr_frame_bytes = 0; |
1716 | return mp3_bytes; |
1717 | } |
1718 | |
1719 | void mp3dec_init(mp3dec_t *dec) |
1720 | { |
1721 | dec->header[0] = 0; |
1722 | } |
1723 | |
1724 | int mp3dec_decode_frame(mp3dec_t *dec, const uint8_t *mp3, int mp3_bytes, mp3d_sample_t *pcm, mp3dec_frame_info_t *info) |
1725 | { |
1726 | int i = 0, igr, frame_size = 0, success = 1; |
1727 | const uint8_t *hdr; |
1728 | bs_t bs_frame[1]; |
1729 | mp3dec_scratch_t scratch; |
1730 | |
1731 | if (mp3_bytes > 4 && dec->header[0] == 0xff && hdr_compare(dec->header, mp3)) |
1732 | { |
1733 | frame_size = hdr_frame_bytes(mp3, dec->free_format_bytes) + hdr_padding(mp3); |
1734 | if (frame_size != mp3_bytes && (frame_size + HDR_SIZE > mp3_bytes || !hdr_compare(mp3, mp3 + frame_size))) |
1735 | { |
1736 | frame_size = 0; |
1737 | } |
1738 | } |
1739 | if (!frame_size) |
1740 | { |
1741 | memset(dec, 0, sizeof(mp3dec_t)); |
1742 | i = mp3d_find_frame(mp3, mp3_bytes, &dec->free_format_bytes, &frame_size); |
1743 | if (!frame_size || i + frame_size > mp3_bytes) |
1744 | { |
1745 | info->frame_bytes = i; |
1746 | return 0; |
1747 | } |
1748 | } |
1749 | |
1750 | hdr = mp3 + i; |
1751 | memcpy(dec->header, hdr, HDR_SIZE); |
1752 | info->frame_bytes = i + frame_size; |
1753 | info->frame_offset = i; |
1754 | info->channels = HDR_IS_MONO(hdr) ? 1 : 2; |
1755 | info->hz = hdr_sample_rate_hz(hdr); |
1756 | info->layer = 4 - HDR_GET_LAYER(hdr); |
1757 | info->bitrate_kbps = hdr_bitrate_kbps(hdr); |
1758 | |
1759 | if (!pcm) |
1760 | { |
1761 | return hdr_frame_samples(hdr); |
1762 | } |
1763 | |
1764 | bs_init(bs_frame, hdr + HDR_SIZE, frame_size - HDR_SIZE); |
1765 | if (HDR_IS_CRC(hdr)) |
1766 | { |
1767 | get_bits(bs_frame, 16); |
1768 | } |
1769 | |
1770 | if (info->layer == 3) |
1771 | { |
1772 | int main_data_begin = L3_read_side_info(bs_frame, scratch.gr_info, hdr); |
1773 | if (main_data_begin < 0 || bs_frame->pos > bs_frame->limit) |
1774 | { |
1775 | mp3dec_init(dec); |
1776 | return 0; |
1777 | } |
1778 | success = L3_restore_reservoir(dec, bs_frame, &scratch, main_data_begin); |
1779 | if (success) |
1780 | { |
1781 | for (igr = 0; igr < (HDR_TEST_MPEG1(hdr) ? 2 : 1); igr++, pcm += 576*info->channels) |
1782 | { |
1783 | memset(scratch.grbuf[0], 0, 576*2*sizeof(float)); |
1784 | L3_decode(dec, &scratch, scratch.gr_info + igr*info->channels, info->channels); |
1785 | mp3d_synth_granule(dec->qmf_state, scratch.grbuf[0], 18, info->channels, pcm, scratch.syn[0]); |
1786 | } |
1787 | } |
1788 | L3_save_reservoir(dec, &scratch); |
1789 | } else |
1790 | { |
1791 | #ifdef MINIMP3_ONLY_MP3 |
1792 | return 0; |
1793 | #else /* MINIMP3_ONLY_MP3 */ |
1794 | L12_scale_info sci[1]; |
1795 | L12_read_scale_info(hdr, bs_frame, sci); |
1796 | |
1797 | memset(scratch.grbuf[0], 0, 576*2*sizeof(float)); |
1798 | for (i = 0, igr = 0; igr < 3; igr++) |
1799 | { |
1800 | if (12 == (i += L12_dequantize_granule(scratch.grbuf[0] + i, bs_frame, sci, info->layer | 1))) |
1801 | { |
1802 | i = 0; |
1803 | L12_apply_scf_384(sci, sci->scf + igr, scratch.grbuf[0]); |
1804 | mp3d_synth_granule(dec->qmf_state, scratch.grbuf[0], 12, info->channels, pcm, scratch.syn[0]); |
1805 | memset(scratch.grbuf[0], 0, 576*2*sizeof(float)); |
1806 | pcm += 384*info->channels; |
1807 | } |
1808 | if (bs_frame->pos > bs_frame->limit) |
1809 | { |
1810 | mp3dec_init(dec); |
1811 | return 0; |
1812 | } |
1813 | } |
1814 | #endif /* MINIMP3_ONLY_MP3 */ |
1815 | } |
1816 | return success*hdr_frame_samples(dec->header); |
1817 | } |
1818 | |
1819 | #ifdef MINIMP3_FLOAT_OUTPUT |
1820 | void mp3dec_f32_to_s16(const float *in, int16_t *out, int num_samples) |
1821 | { |
1822 | int i = 0; |
1823 | #if HAVE_SIMD |
1824 | int aligned_count = num_samples & ~7; |
1825 | for(; i < aligned_count; i += 8) |
1826 | { |
1827 | |
1828 | // -- GODOT start -- |
1829 | #if defined(_MSC_VER) && (defined(_M_ARM64) || defined(_M_ARM64EC) || defined(_M_ARM)) |
1830 | static f4 g_scale; |
1831 | g_scale = vsetq_lane_f32(32768.0f, g_scale, 0); |
1832 | g_scale = vsetq_lane_f32(32768.0f, g_scale, 1); |
1833 | g_scale = vsetq_lane_f32(32768.0f, g_scale, 2); |
1834 | g_scale = vsetq_lane_f32(32768.0f, g_scale, 3); |
1835 | #else |
1836 | static const f4 g_scale = { 32768.0f, 32768.0f, 32768.0f, 32768.0f }; |
1837 | #endif |
1838 | // -- GODOT end -- |
1839 | |
1840 | f4 a = VMUL(VLD(&in[i ]), g_scale); |
1841 | f4 b = VMUL(VLD(&in[i+4]), g_scale); |
1842 | #if HAVE_SSE |
1843 | static const f4 g_max = { 32767.0f, 32767.0f, 32767.0f, 32767.0f }; |
1844 | static const f4 g_min = { -32768.0f, -32768.0f, -32768.0f, -32768.0f }; |
1845 | __m128i pcm8 = _mm_packs_epi32(_mm_cvtps_epi32(_mm_max_ps(_mm_min_ps(a, g_max), g_min)), |
1846 | _mm_cvtps_epi32(_mm_max_ps(_mm_min_ps(b, g_max), g_min))); |
1847 | out[i ] = _mm_extract_epi16(pcm8, 0); |
1848 | out[i+1] = _mm_extract_epi16(pcm8, 1); |
1849 | out[i+2] = _mm_extract_epi16(pcm8, 2); |
1850 | out[i+3] = _mm_extract_epi16(pcm8, 3); |
1851 | out[i+4] = _mm_extract_epi16(pcm8, 4); |
1852 | out[i+5] = _mm_extract_epi16(pcm8, 5); |
1853 | out[i+6] = _mm_extract_epi16(pcm8, 6); |
1854 | out[i+7] = _mm_extract_epi16(pcm8, 7); |
1855 | #else /* HAVE_SSE */ |
1856 | int16x4_t pcma, pcmb; |
1857 | a = VADD(a, VSET(0.5f)); |
1858 | b = VADD(b, VSET(0.5f)); |
1859 | pcma = vqmovn_s32(vqaddq_s32(vcvtq_s32_f32(a), vreinterpretq_s32_u32(vcltq_f32(a, VSET(0))))); |
1860 | pcmb = vqmovn_s32(vqaddq_s32(vcvtq_s32_f32(b), vreinterpretq_s32_u32(vcltq_f32(b, VSET(0))))); |
1861 | vst1_lane_s16(out+i , pcma, 0); |
1862 | vst1_lane_s16(out+i+1, pcma, 1); |
1863 | vst1_lane_s16(out+i+2, pcma, 2); |
1864 | vst1_lane_s16(out+i+3, pcma, 3); |
1865 | vst1_lane_s16(out+i+4, pcmb, 0); |
1866 | vst1_lane_s16(out+i+5, pcmb, 1); |
1867 | vst1_lane_s16(out+i+6, pcmb, 2); |
1868 | vst1_lane_s16(out+i+7, pcmb, 3); |
1869 | #endif /* HAVE_SSE */ |
1870 | } |
1871 | #endif /* HAVE_SIMD */ |
1872 | for(; i < num_samples; i++) |
1873 | { |
1874 | float sample = in[i] * 32768.0f; |
1875 | if (sample >= 32766.5) |
1876 | out[i] = (int16_t) 32767; |
1877 | else if (sample <= -32767.5) |
1878 | out[i] = (int16_t)-32768; |
1879 | else |
1880 | { |
1881 | int16_t s = (int16_t)(sample + .5f); |
1882 | s -= (s < 0); /* away from zero, to be compliant */ |
1883 | out[i] = s; |
1884 | } |
1885 | } |
1886 | } |
1887 | #endif /* MINIMP3_FLOAT_OUTPUT */ |
1888 | #endif /* MINIMP3_IMPLEMENTATION && !_MINIMP3_IMPLEMENTATION_GUARD */ |
1889 | |