diff options
Diffstat (limited to 'libavcodec/mpegaudiodec.c')
-rw-r--r-- | libavcodec/mpegaudiodec.c | 210 |
1 files changed, 47 insertions, 163 deletions
diff --git a/libavcodec/mpegaudiodec.c b/libavcodec/mpegaudiodec.c index 1b50f9fd92..eae030ce44 100644 --- a/libavcodec/mpegaudiodec.c +++ b/libavcodec/mpegaudiodec.c @@ -2,20 +2,20 @@ * MPEG Audio decoder * Copyright (c) 2001, 2002 Fabrice Bellard * - * This file is part of Libav. + * This file is part of FFmpeg. * - * Libav is free software; you can redistribute it and/or + * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * - * Libav is distributed in the hope that it will be useful, + * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with Libav; if not, write to the Free Software + * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -24,6 +24,8 @@ * MPEG Audio decoder */ +#define UNCHECKED_BITSTREAM_READER 1 + #include "libavutil/audioconvert.h" #include "avcodec.h" #include "get_bits.h" @@ -58,7 +60,7 @@ typedef struct GranuleDef { int preflag; int short_start, long_end; /* long/short band indexes */ uint8_t scale_factors[40]; - INTFLOAT sb_hybrid[SBLIMIT * 18]; /* 576 samples */ + DECLARE_ALIGNED(16, INTFLOAT, sb_hybrid)[SBLIMIT * 18]; /* 576 samples */ } GranuleDef; typedef struct MPADecodeContext { @@ -130,7 +132,10 @@ static uint16_t band_index_long[9][23]; static INTFLOAT is_table[2][16]; static INTFLOAT is_table_lsf[2][2][16]; static INTFLOAT csa_table[8][4]; -static INTFLOAT mdct_win[8][36]; +/** Window for MDCT. Note that only the component [0,17] and [20,37] are used, + the components 18 and 19 are there only to assure 128-bit alignment for asm + */ +DECLARE_ALIGNED(16, static INTFLOAT, mdct_win)[8][40]; static int16_t division_tab3[1<<6 ]; static int16_t division_tab5[1<<8 ]; @@ -441,15 +446,17 @@ static av_cold void decode_init_static(void) if (j == 2) mdct_win[j][i/3] = FIXHR((d / (1<<5))); - else - mdct_win[j][i ] = FIXHR((d / (1<<5))); + else { + int idx = i < 18 ? i : i + 2; + mdct_win[j][idx] = FIXHR((d / (1<<5))); + } } } /* NOTE: we do frequency inversion adter the MDCT by changing the sign of the right window coefs */ for (j = 0; j < 4; j++) { - for (i = 0; i < 36; i += 2) { + for (i = 0; i < 40; i += 2) { mdct_win[j + 4][i ] = mdct_win[j][i ]; mdct_win[j + 4][i + 1] = -mdct_win[j][i + 1]; } @@ -483,32 +490,9 @@ static av_cold int decode_init(AVCodecContext * avctx) } #define C3 FIXHR(0.86602540378443864676/2) - -/* 0.5 / cos(pi*(2*i+1)/36) */ -static const INTFLOAT icos36[9] = { - FIXR(0.50190991877167369479), - FIXR(0.51763809020504152469), //0 - FIXR(0.55168895948124587824), - FIXR(0.61038729438072803416), - FIXR(0.70710678118654752439), //1 - FIXR(0.87172339781054900991), - FIXR(1.18310079157624925896), - FIXR(1.93185165257813657349), //2 - FIXR(5.73685662283492756461), -}; - -/* 0.5 / cos(pi*(2*i+1)/36) */ -static const INTFLOAT icos36h[9] = { - FIXHR(0.50190991877167369479/2), - FIXHR(0.51763809020504152469/2), //0 - FIXHR(0.55168895948124587824/2), - FIXHR(0.61038729438072803416/2), - FIXHR(0.70710678118654752439/2), //1 - FIXHR(0.87172339781054900991/2), - FIXHR(1.18310079157624925896/4), - FIXHR(1.93185165257813657349/4), //2 -// FIXHR(5.73685662283492756461), -}; +#define C4 FIXHR(0.70710678118654752439/2) //0.5 / cos(pi*(9)/36) +#define C5 FIXHR(0.51763809020504152469/2) //0.5 / cos(pi*(5)/36) +#define C6 FIXHR(1.93185165257813657349/4) //0.5 / cos(pi*(15)/36) /* 12 points IMDCT. We compute it "by hand" by factorizing obvious cases. */ @@ -529,7 +513,7 @@ static void imdct12(INTFLOAT *out, INTFLOAT *in) in3 = MULH3(in3, C3, 4); t1 = in0 - in4; - t2 = MULH3(in1 - in5, icos36h[4], 2); + t2 = MULH3(in1 - in5, C4, 2); out[ 7] = out[10] = t1 + t2; @@ -539,112 +523,20 @@ static void imdct12(INTFLOAT *out, INTFLOAT *in) in0 += SHR(in4, 1); in4 = in0 + in2; in5 += 2*in1; - in1 = MULH3(in5 + in3, icos36h[1], 1); + in1 = MULH3(in5 + in3, C5, 1); out[ 8] = out[ 9] = in4 + in1; out[ 2] = out[ 3] = in4 - in1; in0 -= in2; - in5 = MULH3(in5 - in3, icos36h[7], 2); + in5 = MULH3(in5 - in3, C6, 2); out[ 0] = out[ 5] = in0 - in5; out[ 6] = out[11] = in0 + in5; } -/* cos(pi*i/18) */ -#define C1 FIXHR(0.98480775301220805936/2) -#define C2 FIXHR(0.93969262078590838405/2) -#define C3 FIXHR(0.86602540378443864676/2) -#define C4 FIXHR(0.76604444311897803520/2) -#define C5 FIXHR(0.64278760968653932632/2) -#define C6 FIXHR(0.5/2) -#define C7 FIXHR(0.34202014332566873304/2) -#define C8 FIXHR(0.17364817766693034885/2) - - -/* using Lee like decomposition followed by hand coded 9 points DCT */ -static void imdct36(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in, INTFLOAT *win) -{ - int i, j; - INTFLOAT t0, t1, t2, t3, s0, s1, s2, s3; - INTFLOAT tmp[18], *tmp1, *in1; - - for (i = 17; i >= 1; i--) - in[i] += in[i-1]; - for (i = 17; i >= 3; i -= 2) - in[i] += in[i-2]; - - for (j = 0; j < 2; j++) { - tmp1 = tmp + j; - in1 = in + j; - - t2 = in1[2*4] + in1[2*8] - in1[2*2]; - - t3 = in1[2*0] + SHR(in1[2*6],1); - t1 = in1[2*0] - in1[2*6]; - tmp1[ 6] = t1 - SHR(t2,1); - tmp1[16] = t1 + t2; - - t0 = MULH3(in1[2*2] + in1[2*4] , C2, 2); - t1 = MULH3(in1[2*4] - in1[2*8] , -2*C8, 1); - t2 = MULH3(in1[2*2] + in1[2*8] , -C4, 2); - - tmp1[10] = t3 - t0 - t2; - tmp1[ 2] = t3 + t0 + t1; - tmp1[14] = t3 + t2 - t1; - - tmp1[ 4] = MULH3(in1[2*5] + in1[2*7] - in1[2*1], -C3, 2); - t2 = MULH3(in1[2*1] + in1[2*5], C1, 2); - t3 = MULH3(in1[2*5] - in1[2*7], -2*C7, 1); - t0 = MULH3(in1[2*3], C3, 2); - - t1 = MULH3(in1[2*1] + in1[2*7], -C5, 2); - - tmp1[ 0] = t2 + t3 + t0; - tmp1[12] = t2 + t1 - t0; - tmp1[ 8] = t3 - t1 - t0; - } - - i = 0; - for (j = 0; j < 4; j++) { - t0 = tmp[i]; - t1 = tmp[i + 2]; - s0 = t1 + t0; - s2 = t1 - t0; - - t2 = tmp[i + 1]; - t3 = tmp[i + 3]; - s1 = MULH3(t3 + t2, icos36h[ j], 2); - s3 = MULLx(t3 - t2, icos36 [8 - j], FRAC_BITS); - - t0 = s0 + s1; - t1 = s0 - s1; - out[(9 + j) * SBLIMIT] = MULH3(t1, win[ 9 + j], 1) + buf[9 + j]; - out[(8 - j) * SBLIMIT] = MULH3(t1, win[ 8 - j], 1) + buf[8 - j]; - buf[ 9 + j ] = MULH3(t0, win[18 + 9 + j], 1); - buf[ 8 - j ] = MULH3(t0, win[18 + 8 - j], 1); - - t0 = s2 + s3; - t1 = s2 - s3; - out[(9 + 8 - j) * SBLIMIT] = MULH3(t1, win[ 9 + 8 - j], 1) + buf[9 + 8 - j]; - out[ j * SBLIMIT] = MULH3(t1, win[ j], 1) + buf[ j]; - buf[ 9 + 8 - j ] = MULH3(t0, win[18 + 9 + 8 - j], 1); - buf[ j ] = MULH3(t0, win[18 + j], 1); - i += 4; - } - - s0 = tmp[16]; - s1 = MULH3(tmp[17], icos36h[4], 2); - t0 = s0 + s1; - t1 = s0 - s1; - out[(9 + 4) * SBLIMIT] = MULH3(t1, win[ 9 + 4], 1) + buf[9 + 4]; - out[(8 - 4) * SBLIMIT] = MULH3(t1, win[ 8 - 4], 1) + buf[8 - 4]; - buf[ 9 + 4 ] = MULH3(t0, win[18 + 9 + 4], 1); - buf[ 8 - 4 ] = MULH3(t0, win[18 + 8 - 4], 1); -} - /* return the number of decoded frames */ static int mp_decode_layer1(MPADecodeContext *s) { @@ -1108,7 +1000,7 @@ static int huffman_decode(MPADecodeContext *s, GranuleDef *g, s_index -= 4; skip_bits_long(&s->gb, last_pos - pos); av_log(s->avctx, AV_LOG_INFO, "overread, skip %d enddists: %d %d\n", last_pos - pos, end_pos-pos, end_pos2-pos); - if(s->err_recognition & AV_EF_BITSTREAM) + if(s->err_recognition & (AV_EF_BITSTREAM|AV_EF_COMPLIANT)) s_index=0; break; } @@ -1138,10 +1030,10 @@ static int huffman_decode(MPADecodeContext *s, GranuleDef *g, /* skip extension bits */ bits_left = end_pos2 - get_bits_count(&s->gb); //av_log(NULL, AV_LOG_ERROR, "left:%d buf:%p\n", bits_left, s->in_gb.buffer); - if (bits_left < 0 && (s->err_recognition & AV_EF_BITSTREAM)) { + if (bits_left < 0 && (s->err_recognition & (AV_EF_BITSTREAM|AV_EF_COMPLIANT))) { av_log(s->avctx, AV_LOG_ERROR, "bits_left=%d\n", bits_left); s_index=0; - } else if (bits_left > 0 && (s->err_recognition & AV_EF_BUFFER)) { + } else if (bits_left > 0 && (s->err_recognition & (AV_EF_BITSTREAM|AV_EF_AGGRESSIVE))) { av_log(s->avctx, AV_LOG_ERROR, "bits_left=%d\n", bits_left); s_index = 0; } @@ -1366,7 +1258,7 @@ static void compute_antialias(MPADecodeContext *s, GranuleDef *g) static void compute_imdct(MPADecodeContext *s, GranuleDef *g, INTFLOAT *sb_samples, INTFLOAT *mdct_buf) { - INTFLOAT *win, *win1, *out_ptr, *ptr, *buf, *ptr1; + INTFLOAT *win, *out_ptr, *ptr, *buf, *ptr1; INTFLOAT out2[12]; int i, j, mdct_long_end, sblimit; @@ -1395,23 +1287,19 @@ static void compute_imdct(MPADecodeContext *s, GranuleDef *g, buf = mdct_buf; ptr = g->sb_hybrid; for (j = 0; j < mdct_long_end; j++) { + int win_idx = (g->switch_point && j < 2) ? 0 : g->block_type; /* apply window & overlap with previous buffer */ out_ptr = sb_samples + j; /* select window */ - if (g->switch_point && j < 2) - win1 = mdct_win[0]; - else - win1 = mdct_win[g->block_type]; - /* select frequency inversion */ - win = win1 + ((4 * 36) & -(j & 1)); - imdct36(out_ptr, buf, ptr, win); + win = mdct_win[win_idx + (4 & -(j & 1))]; + s->mpadsp.RENAME(imdct36)(out_ptr, buf, ptr, win); out_ptr += 18 * SBLIMIT; ptr += 18; buf += 18; } for (j = mdct_long_end; j < sblimit; j++) { /* select frequency inversion */ - win = mdct_win[2] + ((4 * 36) & -(j & 1)); + win = mdct_win[2 + (4 & -(j & 1))]; out_ptr = sb_samples + j; for (i = 0; i < 6; i++) { @@ -1552,29 +1440,25 @@ static int mp_decode_layer3(MPADecodeContext *s) #if !UNCHECKED_BITSTREAM_READER s->gb.size_in_bits_plus8 += EXTRABYTES * 8; #endif - s->last_buf_size <<= 3; - for (gr = 0; gr < nb_granules && (s->last_buf_size >> 3) < main_data_begin; gr++) { - for (ch = 0; ch < s->nb_channels; ch++) { - g = &s->granules[ch][gr]; - s->last_buf_size += g->part2_3_length; - memset(g->sb_hybrid, 0, sizeof(g->sb_hybrid)); - } - } - skip = s->last_buf_size - 8 * main_data_begin; - if (skip >= s->gb.size_in_bits && s->in_gb.buffer) { - skip_bits_long(&s->in_gb, skip - s->gb.size_in_bits); - s->gb = s->in_gb; - s->in_gb.buffer = NULL; - } else { - skip_bits_long(&s->gb, skip); - } - } else { - gr = 0; + skip_bits_long(&s->gb, 8*(s->last_buf_size - main_data_begin)); } - for (; gr < nb_granules; gr++) { + for (gr = 0; gr < nb_granules; gr++) { for (ch = 0; ch < s->nb_channels; ch++) { g = &s->granules[ch][gr]; + if (get_bits_count(&s->gb) < 0) { + av_log(s->avctx, AV_LOG_DEBUG, "mdb:%d, lastbuf:%d skipping granule %d\n", + main_data_begin, s->last_buf_size, gr); + skip_bits_long(&s->gb, g->part2_3_length); + memset(g->sb_hybrid, 0, sizeof(g->sb_hybrid)); + if (get_bits_count(&s->gb) >= s->gb.size_in_bits && s->in_gb.buffer) { + skip_bits_long(&s->in_gb, get_bits_count(&s->gb) - s->gb.size_in_bits); + s->gb = s->in_gb; + s->in_gb.buffer = NULL; + } + continue; + } + bits_pos = get_bits_count(&s->gb); if (!s->lsf) { @@ -1818,8 +1702,8 @@ static int decode_frame(AVCodecContext * avctx, void *data, int *got_frame_ptr, if (s->frame_size <= 0 || s->frame_size > buf_size) { av_log(avctx, AV_LOG_ERROR, "incomplete frame\n"); return AVERROR_INVALIDDATA; - } else if (s->frame_size < buf_size) { - av_log(avctx, AV_LOG_ERROR, "incorrect frame size\n"); + }else if(s->frame_size < buf_size){ + av_log(avctx, AV_LOG_DEBUG, "incorrect frame size - multiple frames in buffer?\n"); buf_size= s->frame_size; } |