Diffstat (limited to 'core/multimedia/opieplayer/libmad/layer3.c') (more/less context) (ignore whitespace changes)
-rw-r--r-- | core/multimedia/opieplayer/libmad/layer3.c | 232 |
1 files changed, 214 insertions, 18 deletions
diff --git a/core/multimedia/opieplayer/libmad/layer3.c b/core/multimedia/opieplayer/libmad/layer3.c index 03f13fe..3c5dd9e 100644 --- a/core/multimedia/opieplayer/libmad/layer3.c +++ b/core/multimedia/opieplayer/libmad/layer3.c @@ -1,51 +1,51 @@ /* * libmad - MPEG audio decoder library - * Copyright (C) 2000-2001 Robert Leslie + * Copyright (C) 2000-2004 Underbit Technologies, Inc. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program 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 General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * * $Id$ */ # ifdef HAVE_CONFIG_H # include "libmad_config.h" # endif # include "libmad_global.h" # include <stdlib.h> # include <string.h> # ifdef HAVE_ASSERT_H # include <assert.h> # endif # ifdef HAVE_LIMITS_H # include <limits.h> # else # define CHAR_BIT 8 # endif # include "fixed.h" # include "bit.h" # include "stream.h" # include "frame.h" # include "huffman.h" # include "layer3.h" /* --- Layer III ----------------------------------------------------------- */ enum { count1table_select = 0x01, @@ -243,270 +243,270 @@ unsigned char const sfb_24000_mixed[] = { /* long */ 6, 6, 6, 6, 6, 6, /* short */ 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 }; static unsigned char const sfb_22050_mixed[] = { /* long */ 6, 6, 6, 6, 6, 6, /* short */ 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 }; static unsigned char const sfb_16000_mixed[] = { /* long */ 6, 6, 6, 6, 6, 6, /* short */ 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 }; /* * MPEG 2.5 scalefactor band widths * derived from public sources */ # define sfb_12000_long sfb_16000_long # define sfb_11025_long sfb_12000_long static unsigned char const sfb_8000_long[] = { 12, 12, 12, 12, 12, 12, 16, 20, 24, 28, 32, 40, 48, 56, 64, 76, 90, 2, 2, 2, 2, 2 }; # define sfb_12000_short sfb_16000_short # define sfb_11025_short sfb_12000_short static unsigned char const sfb_8000_short[] = { 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 }; # define sfb_12000_mixed sfb_16000_mixed # define sfb_11025_mixed sfb_12000_mixed -/* the 8000 Hz short block scalefactor bands do not break after the first 36 - frequency lines, so this is probably wrong */ +/* the 8000 Hz short block scalefactor bands do not break after + the first 36 frequency lines, so this is probably wrong */ static unsigned char const sfb_8000_mixed[] = { /* long */ 12, 12, 12, /* short */ 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 }; static struct { unsigned char const *l; unsigned char const *s; unsigned char const *m; } const sfbwidth_table[9] = { { sfb_48000_long, sfb_48000_short, sfb_48000_mixed }, { sfb_44100_long, sfb_44100_short, sfb_44100_mixed }, { sfb_32000_long, sfb_32000_short, sfb_32000_mixed }, { sfb_24000_long, sfb_24000_short, sfb_24000_mixed }, { sfb_22050_long, sfb_22050_short, sfb_22050_mixed }, { sfb_16000_long, sfb_16000_short, sfb_16000_mixed }, { sfb_12000_long, sfb_12000_short, sfb_12000_mixed }, { sfb_11025_long, sfb_11025_short, sfb_11025_mixed }, { sfb_8000_long, sfb_8000_short, sfb_8000_mixed } }; /* * scalefactor band preemphasis (used only when preflag is set) * derived from Table B.6 of ISO/IEC 11172-3 */ static unsigned char const pretab[22] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 2, 0 }; /* * table for requantization * * rq_table[x].mantissa * 2^(rq_table[x].exponent) = x^(4/3) */ static struct fixedfloat { unsigned long mantissa : 27; unsigned short exponent : 5; } const rq_table[8207] = { # include "rq_table.dat" }; /* * fractional powers of two * used for requantization and joint stereo decoding * * root_table[3 + x] = 2^(x/4) */ static mad_fixed_t const root_table[7] = { MAD_F(0x09837f05) /* 2^(-3/4) == 0.59460355750136 */, MAD_F(0x0b504f33) /* 2^(-2/4) == 0.70710678118655 */, MAD_F(0x0d744fcd) /* 2^(-1/4) == 0.84089641525371 */, MAD_F(0x10000000) /* 2^( 0/4) == 1.00000000000000 */, MAD_F(0x1306fe0a) /* 2^(+1/4) == 1.18920711500272 */, MAD_F(0x16a09e66) /* 2^(+2/4) == 1.41421356237310 */, MAD_F(0x1ae89f99) /* 2^(+3/4) == 1.68179283050743 */ }; /* * coefficients for aliasing reduction * derived from Table B.9 of ISO/IEC 11172-3 * * c[] = { -0.6, -0.535, -0.33, -0.185, -0.095, -0.041, -0.0142, -0.0037 } * cs[i] = 1 / sqrt(1 + c[i]^2) * ca[i] = c[i] / sqrt(1 + c[i]^2) */ static mad_fixed_t const cs[8] = { +MAD_F(0x0db84a81) /* +0.857492926 */, +MAD_F(0x0e1b9d7f) /* +0.881741997 */, +MAD_F(0x0f31adcf) /* +0.949628649 */, +MAD_F(0x0fbba815) /* +0.983314592 */, +MAD_F(0x0feda417) /* +0.995517816 */, +MAD_F(0x0ffc8fc8) /* +0.999160558 */, +MAD_F(0x0fff964c) /* +0.999899195 */, +MAD_F(0x0ffff8d3) /* +0.999993155 */ }; static mad_fixed_t const ca[8] = { -MAD_F(0x083b5fe7) /* -0.514495755 */, -MAD_F(0x078c36d2) /* -0.471731969 */, -MAD_F(0x05039814) /* -0.313377454 */, -MAD_F(0x02e91dd1) /* -0.181913200 */, -MAD_F(0x0183603a) /* -0.094574193 */, -MAD_F(0x00a7cb87) /* -0.040965583 */, -MAD_F(0x003a2847) /* -0.014198569 */, -MAD_F(0x000f27b4) /* -0.003699975 */ }; /* * IMDCT coefficients for short blocks * derived from section 2.4.3.4.10.2 of ISO/IEC 11172-3 * - * imdct_s[i/even][k] = cos((PI / 24) * (2 * (i / 2) + 7) * (2 * k + 1)) + * imdct_s[i/even][k] = cos((PI / 24) * (2 * (i / 2) + 7) * (2 * k + 1)) * imdct_s[i /odd][k] = cos((PI / 24) * (2 * (6 + (i-1)/2) + 7) * (2 * k + 1)) */ static mad_fixed_t const imdct_s[6][6] = { # include "imdct_s.dat" }; # if !defined(ASO_IMDCT) /* * windowing coefficients for long blocks * derived from section 2.4.3.4.10.3 of ISO/IEC 11172-3 * * window_l[i] = sin((PI / 36) * (i + 1/2)) */ static mad_fixed_t const window_l[36] = { MAD_F(0x00b2aa3e) /* 0.043619387 */, MAD_F(0x0216a2a2) /* 0.130526192 */, MAD_F(0x03768962) /* 0.216439614 */, MAD_F(0x04cfb0e2) /* 0.300705800 */, MAD_F(0x061f78aa) /* 0.382683432 */, MAD_F(0x07635284) /* 0.461748613 */, MAD_F(0x0898c779) /* 0.537299608 */, MAD_F(0x09bd7ca0) /* 0.608761429 */, MAD_F(0x0acf37ad) /* 0.675590208 */, MAD_F(0x0bcbe352) /* 0.737277337 */, MAD_F(0x0cb19346) /* 0.793353340 */, MAD_F(0x0d7e8807) /* 0.843391446 */, MAD_F(0x0e313245) /* 0.887010833 */, MAD_F(0x0ec835e8) /* 0.923879533 */, MAD_F(0x0f426cb5) /* 0.953716951 */, MAD_F(0x0f9ee890) /* 0.976296007 */, MAD_F(0x0fdcf549) /* 0.991444861 */, MAD_F(0x0ffc19fd) /* 0.999048222 */, MAD_F(0x0ffc19fd) /* 0.999048222 */, MAD_F(0x0fdcf549) /* 0.991444861 */, MAD_F(0x0f9ee890) /* 0.976296007 */, MAD_F(0x0f426cb5) /* 0.953716951 */, MAD_F(0x0ec835e8) /* 0.923879533 */, MAD_F(0x0e313245) /* 0.887010833 */, MAD_F(0x0d7e8807) /* 0.843391446 */, MAD_F(0x0cb19346) /* 0.793353340 */, MAD_F(0x0bcbe352) /* 0.737277337 */, MAD_F(0x0acf37ad) /* 0.675590208 */, MAD_F(0x09bd7ca0) /* 0.608761429 */, MAD_F(0x0898c779) /* 0.537299608 */, MAD_F(0x07635284) /* 0.461748613 */, MAD_F(0x061f78aa) /* 0.382683432 */, MAD_F(0x04cfb0e2) /* 0.300705800 */, MAD_F(0x03768962) /* 0.216439614 */, MAD_F(0x0216a2a2) /* 0.130526192 */, MAD_F(0x00b2aa3e) /* 0.043619387 */, }; # endif /* ASO_IMDCT */ /* * windowing coefficients for short blocks * derived from section 2.4.3.4.10.3 of ISO/IEC 11172-3 * * window_s[i] = sin((PI / 12) * (i + 1/2)) */ static mad_fixed_t const window_s[12] = { MAD_F(0x0216a2a2) /* 0.130526192 */, MAD_F(0x061f78aa) /* 0.382683432 */, MAD_F(0x09bd7ca0) /* 0.608761429 */, MAD_F(0x0cb19346) /* 0.793353340 */, MAD_F(0x0ec835e8) /* 0.923879533 */, MAD_F(0x0fdcf549) /* 0.991444861 */, MAD_F(0x0fdcf549) /* 0.991444861 */, MAD_F(0x0ec835e8) /* 0.923879533 */, MAD_F(0x0cb19346) /* 0.793353340 */, MAD_F(0x09bd7ca0) /* 0.608761429 */, MAD_F(0x061f78aa) /* 0.382683432 */, MAD_F(0x0216a2a2) /* 0.130526192 */, }; /* * coefficients for intensity stereo processing * derived from section 2.4.3.4.9.3 of ISO/IEC 11172-3 * * is_ratio[i] = tan(i * (PI / 12)) * is_table[i] = is_ratio[i] / (1 + is_ratio[i]) */ static mad_fixed_t const is_table[7] = { MAD_F(0x00000000) /* 0.000000000 */, MAD_F(0x0361962f) /* 0.211324865 */, MAD_F(0x05db3d74) /* 0.366025404 */, MAD_F(0x08000000) /* 0.500000000 */, MAD_F(0x0a24c28c) /* 0.633974596 */, MAD_F(0x0c9e69d1) /* 0.788675135 */, MAD_F(0x10000000) /* 1.000000000 */ }; /* * coefficients for LSF intensity stereo processing * derived from section 2.4.3.2 of ISO/IEC 13818-3 * * is_lsf_table[0][i] = (1 / sqrt(sqrt(2)))^(i + 1) - * is_lsf_table[1][i] = (1 / sqrt(2))^(i + 1) + * is_lsf_table[1][i] = (1 / sqrt(2)) ^(i + 1) */ static mad_fixed_t const is_lsf_table[2][15] = { { MAD_F(0x0d744fcd) /* 0.840896415 */, MAD_F(0x0b504f33) /* 0.707106781 */, MAD_F(0x09837f05) /* 0.594603558 */, MAD_F(0x08000000) /* 0.500000000 */, MAD_F(0x06ba27e6) /* 0.420448208 */, MAD_F(0x05a8279a) /* 0.353553391 */, MAD_F(0x04c1bf83) /* 0.297301779 */, MAD_F(0x04000000) /* 0.250000000 */, MAD_F(0x035d13f3) /* 0.210224104 */, MAD_F(0x02d413cd) /* 0.176776695 */, MAD_F(0x0260dfc1) /* 0.148650889 */, MAD_F(0x02000000) /* 0.125000000 */, MAD_F(0x01ae89fa) /* 0.105112052 */, MAD_F(0x016a09e6) /* 0.088388348 */, MAD_F(0x01306fe1) /* 0.074325445 */ }, { MAD_F(0x0b504f33) /* 0.707106781 */, MAD_F(0x08000000) /* 0.500000000 */, MAD_F(0x05a8279a) /* 0.353553391 */, MAD_F(0x04000000) /* 0.250000000 */, MAD_F(0x02d413cd) /* 0.176776695 */, MAD_F(0x02000000) /* 0.125000000 */, MAD_F(0x016a09e6) /* 0.088388348 */, MAD_F(0x01000000) /* 0.062500000 */, MAD_F(0x00b504f3) /* 0.044194174 */, MAD_F(0x00800000) /* 0.031250000 */, MAD_F(0x005a827a) /* 0.022097087 */, MAD_F(0x00400000) /* 0.015625000 */, MAD_F(0x002d413d) /* 0.011048543 */, MAD_F(0x00200000) /* 0.007812500 */, MAD_F(0x0016a09e) /* 0.005524272 */ } }; /* * NAME: III_sideinfo() * DESCRIPTION: decode frame side information from a bitstream */ static enum mad_error III_sideinfo(struct mad_bitptr *ptr, unsigned int nch, int lsf, struct sideinfo *si, unsigned int *data_bitlen, unsigned int *priv_bitlen) { @@ -1530,96 +1530,283 @@ enum mad_error III_stereo(mad_fixed_t xr[2][576], } } return MAD_ERROR_NONE; } /* * NAME: III_aliasreduce() * DESCRIPTION: perform frequency line alias reduction */ static void III_aliasreduce(mad_fixed_t xr[576], int lines) { mad_fixed_t const *bound; int i; bound = &xr[lines]; for (xr += 18; xr < bound; xr += 18) { for (i = 0; i < 8; ++i) { register mad_fixed_t a, b; register mad_fixed64hi_t hi; register mad_fixed64lo_t lo; a = xr[-1 - i]; b = xr[ i]; # if defined(ASO_ZEROCHECK) if (a | b) { # endif MAD_F_ML0(hi, lo, a, cs[i]); MAD_F_MLA(hi, lo, -b, ca[i]); xr[-1 - i] = MAD_F_MLZ(hi, lo); MAD_F_ML0(hi, lo, b, cs[i]); MAD_F_MLA(hi, lo, a, ca[i]); xr[ i] = MAD_F_MLZ(hi, lo); # if defined(ASO_ZEROCHECK) } # endif } } } # if defined(ASO_IMDCT) void III_imdct_l(mad_fixed_t const [18], mad_fixed_t [36], unsigned int); # else +# if 1 +static +void fastsdct(mad_fixed_t const x[9], mad_fixed_t y[18]) +{ + mad_fixed_t a0, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10, a11, a12; + mad_fixed_t a13, a14, a15, a16, a17, a18, a19, a20, a21, a22, a23, a24, a25; + mad_fixed_t m0, m1, m2, m3, m4, m5, m6, m7; + + enum { + c0 = MAD_F(0x1f838b8d), /* 2 * cos( 1 * PI / 18) */ + c1 = MAD_F(0x1bb67ae8), /* 2 * cos( 3 * PI / 18) */ + c2 = MAD_F(0x18836fa3), /* 2 * cos( 4 * PI / 18) */ + c3 = MAD_F(0x1491b752), /* 2 * cos( 5 * PI / 18) */ + c4 = MAD_F(0x0af1d43a), /* 2 * cos( 7 * PI / 18) */ + c5 = MAD_F(0x058e86a0), /* 2 * cos( 8 * PI / 18) */ + c6 = -MAD_F(0x1e11f642) /* 2 * cos(16 * PI / 18) */ + }; + + a0 = x[3] + x[5]; + a1 = x[3] - x[5]; + a2 = x[6] + x[2]; + a3 = x[6] - x[2]; + a4 = x[1] + x[7]; + a5 = x[1] - x[7]; + a6 = x[8] + x[0]; + a7 = x[8] - x[0]; + + a8 = a0 + a2; + a9 = a0 - a2; + a10 = a0 - a6; + a11 = a2 - a6; + a12 = a8 + a6; + a13 = a1 - a3; + a14 = a13 + a7; + a15 = a3 + a7; + a16 = a1 - a7; + a17 = a1 + a3; + + m0 = mad_f_mul(a17, -c3); + m1 = mad_f_mul(a16, -c0); + m2 = mad_f_mul(a15, -c4); + m3 = mad_f_mul(a14, -c1); + m4 = mad_f_mul(a5, -c1); + m5 = mad_f_mul(a11, -c6); + m6 = mad_f_mul(a10, -c5); + m7 = mad_f_mul(a9, -c2); + + a18 = x[4] + a4; + a19 = 2 * x[4] - a4; + a20 = a19 + m5; + a21 = a19 - m5; + a22 = a19 + m6; + a23 = m4 + m2; + a24 = m4 - m2; + a25 = m4 + m1; + + /* output to every other slot for convenience */ + + y[ 0] = a18 + a12; + y[ 2] = m0 - a25; + y[ 4] = m7 - a20; + y[ 6] = m3; + y[ 8] = a21 - m6; + y[10] = a24 - m1; + y[12] = a12 - 2 * a18; + y[14] = a23 + m0; + y[16] = a22 + m7; +} + +static inline +void sdctII(mad_fixed_t const x[18], mad_fixed_t X[18]) +{ + mad_fixed_t tmp[9]; + int i; + + /* scale[i] = 2 * cos(PI * (2 * i + 1) / (2 * 18)) */ + static mad_fixed_t const scale[9] = { + MAD_F(0x1fe0d3b4), MAD_F(0x1ee8dd47), MAD_F(0x1d007930), + MAD_F(0x1a367e59), MAD_F(0x16a09e66), MAD_F(0x125abcf8), + MAD_F(0x0d8616bc), MAD_F(0x08483ee1), MAD_F(0x02c9fad7) + }; + + /* divide the 18-point SDCT-II into two 9-point SDCT-IIs */ + + /* even input butterfly */ + + for (i = 0; i < 9; i += 3) { + tmp[i + 0] = x[i + 0] + x[18 - (i + 0) - 1]; + tmp[i + 1] = x[i + 1] + x[18 - (i + 1) - 1]; + tmp[i + 2] = x[i + 2] + x[18 - (i + 2) - 1]; + } + + fastsdct(tmp, &X[0]); + + /* odd input butterfly and scaling */ + + for (i = 0; i < 9; i += 3) { + tmp[i + 0] = mad_f_mul(x[i + 0] - x[18 - (i + 0) - 1], scale[i + 0]); + tmp[i + 1] = mad_f_mul(x[i + 1] - x[18 - (i + 1) - 1], scale[i + 1]); + tmp[i + 2] = mad_f_mul(x[i + 2] - x[18 - (i + 2) - 1], scale[i + 2]); + } + + fastsdct(tmp, &X[1]); + + /* output accumulation */ + + for (i = 3; i < 18; i += 8) { + X[i + 0] -= X[(i + 0) - 2]; + X[i + 2] -= X[(i + 2) - 2]; + X[i + 4] -= X[(i + 4) - 2]; + X[i + 6] -= X[(i + 6) - 2]; + } +} + +static inline +void dctIV(mad_fixed_t const y[18], mad_fixed_t X[18]) +{ + mad_fixed_t tmp[18]; + int i; + + /* scale[i] = 2 * cos(PI * (2 * i + 1) / (4 * 18)) */ + static mad_fixed_t const scale[18] = { + MAD_F(0x1ff833fa), MAD_F(0x1fb9ea93), MAD_F(0x1f3dd120), + MAD_F(0x1e84d969), MAD_F(0x1d906bcf), MAD_F(0x1c62648b), + MAD_F(0x1afd100f), MAD_F(0x1963268b), MAD_F(0x1797c6a4), + MAD_F(0x159e6f5b), MAD_F(0x137af940), MAD_F(0x11318ef3), + MAD_F(0x0ec6a507), MAD_F(0x0c3ef153), MAD_F(0x099f61c5), + MAD_F(0x06ed12c5), MAD_F(0x042d4544), MAD_F(0x0165547c) + }; + + /* scaling */ + + for (i = 0; i < 18; i += 3) { + tmp[i + 0] = mad_f_mul(y[i + 0], scale[i + 0]); + tmp[i + 1] = mad_f_mul(y[i + 1], scale[i + 1]); + tmp[i + 2] = mad_f_mul(y[i + 2], scale[i + 2]); + } + + /* SDCT-II */ + + sdctII(tmp, X); + + /* scale reduction and output accumulation */ + + X[0] /= 2; + for (i = 1; i < 17; i += 4) { + X[i + 0] = X[i + 0] / 2 - X[(i + 0) - 1]; + X[i + 1] = X[i + 1] / 2 - X[(i + 1) - 1]; + X[i + 2] = X[i + 2] / 2 - X[(i + 2) - 1]; + X[i + 3] = X[i + 3] / 2 - X[(i + 3) - 1]; + } + X[17] = X[17] / 2 - X[16]; +} + +/* + * NAME: imdct36 + * DESCRIPTION: perform X[18]->x[36] IMDCT using Szu-Wei Lee's fast algorithm + */ +static inline +void imdct36(mad_fixed_t const x[18], mad_fixed_t y[36]) +{ + mad_fixed_t tmp[18]; + int i; + + /* DCT-IV */ + + dctIV(x, tmp); + + /* convert 18-point DCT-IV to 36-point IMDCT */ + + for (i = 0; i < 9; i += 3) { + y[i + 0] = tmp[9 + (i + 0)]; + y[i + 1] = tmp[9 + (i + 1)]; + y[i + 2] = tmp[9 + (i + 2)]; + } + for (i = 9; i < 27; i += 3) { + y[i + 0] = -tmp[36 - (9 + (i + 0)) - 1]; + y[i + 1] = -tmp[36 - (9 + (i + 1)) - 1]; + y[i + 2] = -tmp[36 - (9 + (i + 2)) - 1]; + } + for (i = 27; i < 36; i += 3) { + y[i + 0] = -tmp[(i + 0) - 27]; + y[i + 1] = -tmp[(i + 1) - 27]; + y[i + 2] = -tmp[(i + 2) - 27]; + } +} +# else /* * NAME: imdct36 * DESCRIPTION: perform X[18]->x[36] IMDCT */ static inline void imdct36(mad_fixed_t const X[18], mad_fixed_t x[36]) { mad_fixed_t t0, t1, t2, t3, t4, t5, t6, t7; mad_fixed_t t8, t9, t10, t11, t12, t13, t14, t15; register mad_fixed64hi_t hi; register mad_fixed64lo_t lo; MAD_F_ML0(hi, lo, X[4], MAD_F(0x0ec835e8)); MAD_F_MLA(hi, lo, X[13], MAD_F(0x061f78aa)); t6 = MAD_F_MLZ(hi, lo); MAD_F_MLA(hi, lo, (t14 = X[1] - X[10]), -MAD_F(0x061f78aa)); MAD_F_MLA(hi, lo, (t15 = X[7] + X[16]), -MAD_F(0x0ec835e8)); t0 = MAD_F_MLZ(hi, lo); MAD_F_MLA(hi, lo, (t8 = X[0] - X[11] - X[12]), MAD_F(0x0216a2a2)); MAD_F_MLA(hi, lo, (t9 = X[2] - X[9] - X[14]), MAD_F(0x09bd7ca0)); MAD_F_MLA(hi, lo, (t10 = X[3] - X[8] - X[15]), -MAD_F(0x0cb19346)); MAD_F_MLA(hi, lo, (t11 = X[5] - X[6] - X[17]), -MAD_F(0x0fdcf549)); x[7] = MAD_F_MLZ(hi, lo); x[10] = -x[7]; MAD_F_ML0(hi, lo, t8, -MAD_F(0x0cb19346)); MAD_F_MLA(hi, lo, t9, MAD_F(0x0fdcf549)); MAD_F_MLA(hi, lo, t10, MAD_F(0x0216a2a2)); MAD_F_MLA(hi, lo, t11, -MAD_F(0x09bd7ca0)); x[19] = x[34] = MAD_F_MLZ(hi, lo) - t0; t12 = X[0] - X[3] + X[8] - X[11] - X[12] + X[15]; t13 = X[2] + X[5] - X[6] - X[9] - X[14] - X[17]; MAD_F_ML0(hi, lo, t12, -MAD_F(0x0ec835e8)); MAD_F_MLA(hi, lo, t13, MAD_F(0x061f78aa)); x[22] = x[31] = MAD_F_MLZ(hi, lo) + t0; MAD_F_ML0(hi, lo, X[1], -MAD_F(0x09bd7ca0)); MAD_F_MLA(hi, lo, X[7], MAD_F(0x0216a2a2)); MAD_F_MLA(hi, lo, X[10], -MAD_F(0x0fdcf549)); @@ -1820,171 +2007,180 @@ void imdct36(mad_fixed_t const X[18], mad_fixed_t x[36]) MAD_F_ML0(hi, lo, X[0], MAD_F(0x0898c779)); MAD_F_MLA(hi, lo, X[2], MAD_F(0x04cfb0e2)); MAD_F_MLA(hi, lo, X[3], MAD_F(0x0bcbe352)); MAD_F_MLA(hi, lo, X[5], MAD_F(0x00b2aa3e)); MAD_F_MLA(hi, lo, X[6], MAD_F(0x0e313245)); MAD_F_MLA(hi, lo, X[8], -MAD_F(0x03768962)); MAD_F_MLA(hi, lo, X[9], MAD_F(0x0f9ee890)); MAD_F_MLA(hi, lo, X[11], -MAD_F(0x07635284)); MAD_F_MLA(hi, lo, X[12], MAD_F(0x0ffc19fd)); MAD_F_MLA(hi, lo, X[14], -MAD_F(0x0acf37ad)); MAD_F_MLA(hi, lo, X[15], MAD_F(0x0f426cb5)); MAD_F_MLA(hi, lo, X[17], -MAD_F(0x0d7e8807)); x[2] = MAD_F_MLZ(hi, lo) + t5; x[15] = -x[2]; MAD_F_ML0(hi, lo, X[0], MAD_F(0x07635284)); MAD_F_MLA(hi, lo, X[2], MAD_F(0x0acf37ad)); MAD_F_MLA(hi, lo, X[3], MAD_F(0x03768962)); MAD_F_MLA(hi, lo, X[5], MAD_F(0x0d7e8807)); MAD_F_MLA(hi, lo, X[6], -MAD_F(0x00b2aa3e)); MAD_F_MLA(hi, lo, X[8], MAD_F(0x0f426cb5)); MAD_F_MLA(hi, lo, X[9], -MAD_F(0x04cfb0e2)); MAD_F_MLA(hi, lo, X[11], MAD_F(0x0ffc19fd)); MAD_F_MLA(hi, lo, X[12], -MAD_F(0x0898c779)); MAD_F_MLA(hi, lo, X[14], MAD_F(0x0f9ee890)); MAD_F_MLA(hi, lo, X[15], -MAD_F(0x0bcbe352)); MAD_F_MLA(hi, lo, X[17], MAD_F(0x0e313245)); x[3] = MAD_F_MLZ(hi, lo) + t5; x[14] = -x[3]; MAD_F_ML0(hi, lo, X[0], -MAD_F(0x0ffc19fd)); MAD_F_MLA(hi, lo, X[2], -MAD_F(0x0f9ee890)); MAD_F_MLA(hi, lo, X[3], -MAD_F(0x0f426cb5)); MAD_F_MLA(hi, lo, X[5], -MAD_F(0x0e313245)); MAD_F_MLA(hi, lo, X[6], -MAD_F(0x0d7e8807)); MAD_F_MLA(hi, lo, X[8], -MAD_F(0x0bcbe352)); MAD_F_MLA(hi, lo, X[9], -MAD_F(0x0acf37ad)); MAD_F_MLA(hi, lo, X[11], -MAD_F(0x0898c779)); MAD_F_MLA(hi, lo, X[12], -MAD_F(0x07635284)); MAD_F_MLA(hi, lo, X[14], -MAD_F(0x04cfb0e2)); MAD_F_MLA(hi, lo, X[15], -MAD_F(0x03768962)); MAD_F_MLA(hi, lo, X[17], -MAD_F(0x00b2aa3e)); x[26] = x[27] = MAD_F_MLZ(hi, lo) + t5; } +# endif /* * NAME: III_imdct_l() * DESCRIPTION: perform IMDCT and windowing for long blocks */ static void III_imdct_l(mad_fixed_t const X[18], mad_fixed_t z[36], unsigned int block_type) { unsigned int i; /* IMDCT */ imdct36(X, z); /* windowing */ switch (block_type) { case 0: /* normal window */ # if defined(ASO_INTERLEAVE1) { register mad_fixed_t tmp1, tmp2; tmp1 = window_l[0]; tmp2 = window_l[1]; for (i = 0; i < 34; i += 2) { z[i + 0] = mad_f_mul(z[i + 0], tmp1); tmp1 = window_l[i + 2]; z[i + 1] = mad_f_mul(z[i + 1], tmp2); tmp2 = window_l[i + 3]; } z[34] = mad_f_mul(z[34], tmp1); z[35] = mad_f_mul(z[35], tmp2); } # elif defined(ASO_INTERLEAVE2) { register mad_fixed_t tmp1, tmp2; tmp1 = z[0]; tmp2 = window_l[0]; for (i = 0; i < 35; ++i) { z[i] = mad_f_mul(tmp1, tmp2); tmp1 = z[i + 1]; tmp2 = window_l[i + 1]; } z[35] = mad_f_mul(tmp1, tmp2); } # elif 1 for (i = 0; i < 36; i += 4) { z[i + 0] = mad_f_mul(z[i + 0], window_l[i + 0]); z[i + 1] = mad_f_mul(z[i + 1], window_l[i + 1]); z[i + 2] = mad_f_mul(z[i + 2], window_l[i + 2]); z[i + 3] = mad_f_mul(z[i + 3], window_l[i + 3]); } # else for (i = 0; i < 36; ++i) z[i] = mad_f_mul(z[i], window_l[i]); # endif break; case 1: /* start block */ - for (i = 0; i < 18; ++i) z[i] = mad_f_mul(z[i], window_l[i]); + for (i = 0; i < 18; i += 3) { + z[i + 0] = mad_f_mul(z[i + 0], window_l[i + 0]); + z[i + 1] = mad_f_mul(z[i + 1], window_l[i + 1]); + z[i + 2] = mad_f_mul(z[i + 2], window_l[i + 2]); + } /* (i = 18; i < 24; ++i) z[i] unchanged */ for (i = 24; i < 30; ++i) z[i] = mad_f_mul(z[i], window_s[i - 18]); for (i = 30; i < 36; ++i) z[i] = 0; break; case 3: /* stop block */ for (i = 0; i < 6; ++i) z[i] = 0; for (i = 6; i < 12; ++i) z[i] = mad_f_mul(z[i], window_s[i - 6]); /* (i = 12; i < 18; ++i) z[i] unchanged */ - for (i = 18; i < 36; ++i) z[i] = mad_f_mul(z[i], window_l[i]); + for (i = 18; i < 36; i += 3) { + z[i + 0] = mad_f_mul(z[i + 0], window_l[i + 0]); + z[i + 1] = mad_f_mul(z[i + 1], window_l[i + 1]); + z[i + 2] = mad_f_mul(z[i + 2], window_l[i + 2]); + } break; } } # endif /* ASO_IMDCT */ /* * NAME: III_imdct_s() * DESCRIPTION: perform IMDCT and windowing for short blocks */ static void III_imdct_s(mad_fixed_t const X[18], mad_fixed_t z[36]) { mad_fixed_t y[36], *yptr; mad_fixed_t const *wptr; int w, i; register mad_fixed64hi_t hi; register mad_fixed64lo_t lo; /* IMDCT */ yptr = &y[0]; for (w = 0; w < 3; ++w) { register mad_fixed_t const (*s)[6]; s = imdct_s; for (i = 0; i < 3; ++i) { MAD_F_ML0(hi, lo, X[0], (*s)[0]); MAD_F_MLA(hi, lo, X[1], (*s)[1]); MAD_F_MLA(hi, lo, X[2], (*s)[2]); MAD_F_MLA(hi, lo, X[3], (*s)[3]); MAD_F_MLA(hi, lo, X[4], (*s)[4]); MAD_F_MLA(hi, lo, X[5], (*s)[5]); yptr[i + 0] = MAD_F_MLZ(hi, lo); yptr[5 - i] = -yptr[i + 0]; ++s; MAD_F_ML0(hi, lo, X[0], (*s)[0]); MAD_F_MLA(hi, lo, X[1], (*s)[1]); MAD_F_MLA(hi, lo, X[2], (*s)[2]); MAD_F_MLA(hi, lo, X[3], (*s)[3]); MAD_F_MLA(hi, lo, X[4], (*s)[4]); MAD_F_MLA(hi, lo, X[5], (*s)[5]); yptr[ i + 6] = MAD_F_MLZ(hi, lo); @@ -1997,121 +2193,121 @@ void III_imdct_s(mad_fixed_t const X[18], mad_fixed_t z[36]) X += 6; } /* windowing, overlapping and concatenation */ yptr = &y[0]; wptr = &window_s[0]; for (i = 0; i < 6; ++i) { z[i + 0] = 0; z[i + 6] = mad_f_mul(yptr[ 0 + 0], wptr[0]); MAD_F_ML0(hi, lo, yptr[ 0 + 6], wptr[6]); MAD_F_MLA(hi, lo, yptr[12 + 0], wptr[0]); z[i + 12] = MAD_F_MLZ(hi, lo); MAD_F_ML0(hi, lo, yptr[12 + 6], wptr[6]); MAD_F_MLA(hi, lo, yptr[24 + 0], wptr[0]); z[i + 18] = MAD_F_MLZ(hi, lo); z[i + 24] = mad_f_mul(yptr[24 + 6], wptr[6]); z[i + 30] = 0; ++yptr; ++wptr; } } /* * NAME: III_overlap() * DESCRIPTION: perform overlap-add of windowed IMDCT outputs */ static void III_overlap(mad_fixed_t const output[36], mad_fixed_t overlap[18], mad_fixed_t sample[18][32], unsigned int sb) { unsigned int i; # if defined(ASO_INTERLEAVE2) { register mad_fixed_t tmp1, tmp2; tmp1 = overlap[0]; tmp2 = overlap[1]; for (i = 0; i < 16; i += 2) { - sample[i + 0][sb] = output[i + 0] + tmp1; + sample[i + 0][sb] = output[i + 0 + 0] + tmp1; overlap[i + 0] = output[i + 0 + 18]; tmp1 = overlap[i + 2]; - sample[i + 1][sb] = output[i + 1] + tmp2; + sample[i + 1][sb] = output[i + 1 + 0] + tmp2; overlap[i + 1] = output[i + 1 + 18]; tmp2 = overlap[i + 3]; } - sample[16][sb] = output[16] + tmp1; + sample[16][sb] = output[16 + 0] + tmp1; overlap[16] = output[16 + 18]; - sample[17][sb] = output[17] + tmp2; + sample[17][sb] = output[17 + 0] + tmp2; overlap[17] = output[17 + 18]; } # elif 0 for (i = 0; i < 18; i += 2) { - sample[i + 0][sb] = output[i + 0] + overlap[i + 0]; + sample[i + 0][sb] = output[i + 0 + 0] + overlap[i + 0]; overlap[i + 0] = output[i + 0 + 18]; - sample[i + 1][sb] = output[i + 1] + overlap[i + 1]; + sample[i + 1][sb] = output[i + 1 + 0] + overlap[i + 1]; overlap[i + 1] = output[i + 1 + 18]; } # else for (i = 0; i < 18; ++i) { - sample[i][sb] = output[i] + overlap[i]; + sample[i][sb] = output[i + 0] + overlap[i]; overlap[i] = output[i + 18]; } # endif } /* * NAME: III_overlap_z() * DESCRIPTION: perform "overlap-add" of zero IMDCT outputs */ static inline void III_overlap_z(mad_fixed_t overlap[18], mad_fixed_t sample[18][32], unsigned int sb) { unsigned int i; # if defined(ASO_INTERLEAVE2) { register mad_fixed_t tmp1, tmp2; tmp1 = overlap[0]; tmp2 = overlap[1]; for (i = 0; i < 16; i += 2) { sample[i + 0][sb] = tmp1; overlap[i + 0] = 0; tmp1 = overlap[i + 2]; sample[i + 1][sb] = tmp2; overlap[i + 1] = 0; tmp2 = overlap[i + 3]; } sample[16][sb] = tmp1; overlap[16] = 0; sample[17][sb] = tmp2; overlap[17] = 0; } # else for (i = 0; i < 18; ++i) { sample[i][sb] = overlap[i]; overlap[i] = 0; } # endif } /* * NAME: III_freqinver() * DESCRIPTION: perform subband frequency inversion for odd sample lines @@ -2409,94 +2605,94 @@ int mad_layer_III(struct mad_stream *stream, struct mad_frame *frame) frame_space = stream->next_frame - mad_bit_nextbyte(&stream->ptr); if (next_md_begin > si.main_data_begin + frame_space) next_md_begin = 0; md_len = si.main_data_begin + frame_space - next_md_begin; frame_used = 0; if (si.main_data_begin == 0) { ptr = stream->ptr; stream->md_len = 0; frame_used = md_len; } else { if (si.main_data_begin > stream->md_len) { if (result == 0) { stream->error = MAD_ERROR_BADDATAPTR; result = -1; } } else { mad_bit_init(&ptr, *stream->main_data + stream->md_len - si.main_data_begin); if (md_len > si.main_data_begin) { assert(stream->md_len + md_len - si.main_data_begin <= MAD_BUFFER_MDLEN); memcpy(*stream->main_data + stream->md_len, mad_bit_nextbyte(&stream->ptr), frame_used = md_len - si.main_data_begin); stream->md_len += frame_used; } } } frame_free = frame_space - frame_used; /* decode main_data */ if (result == 0) { error = III_decode(&ptr, frame, &si, nch); if (error) { stream->error = error; result = -1; } - } - /* designate ancillary bits */ + /* designate ancillary bits */ - stream->anc_ptr = ptr; - stream->anc_bitlen = md_len * CHAR_BIT - data_bitlen; + stream->anc_ptr = ptr; + stream->anc_bitlen = md_len * CHAR_BIT - data_bitlen; + } # if 0 && defined(DEBUG) fprintf(stderr, "main_data_begin:%u, md_len:%u, frame_free:%u, " "data_bitlen:%u, anc_bitlen: %u\n", si.main_data_begin, md_len, frame_free, data_bitlen, stream->anc_bitlen); # endif /* preload main_data buffer with up to 511 bytes for next frame(s) */ if (frame_free >= next_md_begin) { memcpy(*stream->main_data, stream->next_frame - next_md_begin, next_md_begin); stream->md_len = next_md_begin; } else { if (md_len < si.main_data_begin) { unsigned int extra; extra = si.main_data_begin - md_len; if (extra + frame_free > next_md_begin) extra = next_md_begin - frame_free; if (extra < stream->md_len) { memmove(*stream->main_data, *stream->main_data + stream->md_len - extra, extra); stream->md_len = extra; } } else stream->md_len = 0; memcpy(*stream->main_data + stream->md_len, stream->next_frame - frame_free, frame_free); stream->md_len += frame_free; } return result; } |