mirror of https://github.com/mpv-player/mpv
604 lines
19 KiB
C
604 lines
19 KiB
C
/*
|
|
** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
|
|
** Copyright (C) 2003 M. Bakker, Ahead Software AG, http://www.nero.com
|
|
**
|
|
** 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.
|
|
**
|
|
** Any non-GPL usage of this software or parts of this software is strictly
|
|
** forbidden.
|
|
**
|
|
** Initially modified for use with MPlayer by Rich Felker on 2005/03/29
|
|
** $Id$
|
|
** detailed changelog at http://svn.mplayerhq.hu/mplayer/trunk/
|
|
**/
|
|
|
|
#include "common.h"
|
|
#include "structs.h"
|
|
|
|
#include "output.h"
|
|
#include "decoder.h"
|
|
|
|
#ifndef FIXED_POINT
|
|
|
|
|
|
#define FLOAT_SCALE (1.0f/(1<<15))
|
|
|
|
#define DM_MUL REAL_CONST(0.3203772410170407) // 1/(1+sqrt(2) + 1/sqrt(2))
|
|
#define RSQRT2 REAL_CONST(0.7071067811865475244) // 1/sqrt(2)
|
|
|
|
|
|
static INLINE real_t get_sample(real_t **input, uint8_t channel, uint16_t sample,
|
|
uint8_t down_matrix, uint8_t *internal_channel)
|
|
{
|
|
if (!down_matrix)
|
|
return input[internal_channel[channel]][sample];
|
|
|
|
if (channel == 0)
|
|
{
|
|
return DM_MUL * (input[internal_channel[1]][sample] +
|
|
input[internal_channel[0]][sample] * RSQRT2 +
|
|
input[internal_channel[3]][sample] * RSQRT2);
|
|
} else {
|
|
return DM_MUL * (input[internal_channel[2]][sample] +
|
|
input[internal_channel[0]][sample] * RSQRT2 +
|
|
input[internal_channel[4]][sample] * RSQRT2);
|
|
}
|
|
}
|
|
|
|
#ifndef HAS_LRINTF
|
|
#define CLIP(sample, max, min) \
|
|
if (sample >= 0.0f) \
|
|
{ \
|
|
sample += 0.5f; \
|
|
if (sample >= max) \
|
|
sample = max; \
|
|
} else { \
|
|
sample += -0.5f; \
|
|
if (sample <= min) \
|
|
sample = min; \
|
|
}
|
|
#else
|
|
#define CLIP(sample, max, min) \
|
|
if (sample >= 0.0f) \
|
|
{ \
|
|
if (sample >= max) \
|
|
sample = max; \
|
|
} else { \
|
|
if (sample <= min) \
|
|
sample = min; \
|
|
}
|
|
#endif
|
|
|
|
#define CONV(a,b) ((a<<1)|(b&0x1))
|
|
|
|
static void to_PCM_16bit(NeAACDecHandle hDecoder, real_t **input,
|
|
uint8_t channels, uint16_t frame_len,
|
|
int16_t **sample_buffer)
|
|
{
|
|
uint8_t ch, ch1;
|
|
uint16_t i;
|
|
|
|
switch (CONV(channels,hDecoder->downMatrix))
|
|
{
|
|
case CONV(1,0):
|
|
case CONV(1,1):
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = input[hDecoder->internal_channel[0]][i];
|
|
|
|
CLIP(inp, 32767.0f, -32768.0f);
|
|
|
|
(*sample_buffer)[i] = (int16_t)lrintf(inp);
|
|
}
|
|
break;
|
|
case CONV(2,0):
|
|
if (hDecoder->upMatrix)
|
|
{
|
|
ch = hDecoder->internal_channel[0];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch][i];
|
|
|
|
CLIP(inp0, 32767.0f, -32768.0f);
|
|
|
|
(*sample_buffer)[(i*2)+0] = (int16_t)lrintf(inp0);
|
|
(*sample_buffer)[(i*2)+1] = (int16_t)lrintf(inp0);
|
|
}
|
|
} else {
|
|
ch = hDecoder->internal_channel[0];
|
|
ch1 = hDecoder->internal_channel[1];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch ][i];
|
|
real_t inp1 = input[ch1][i];
|
|
|
|
CLIP(inp0, 32767.0f, -32768.0f);
|
|
CLIP(inp1, 32767.0f, -32768.0f);
|
|
|
|
(*sample_buffer)[(i*2)+0] = (int16_t)lrintf(inp0);
|
|
(*sample_buffer)[(i*2)+1] = (int16_t)lrintf(inp1);
|
|
}
|
|
}
|
|
break;
|
|
default:
|
|
for (ch = 0; ch < channels; ch++)
|
|
{
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
|
|
|
|
CLIP(inp, 32767.0f, -32768.0f);
|
|
|
|
(*sample_buffer)[(i*channels)+ch] = (int16_t)lrintf(inp);
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
static void to_PCM_24bit(NeAACDecHandle hDecoder, real_t **input,
|
|
uint8_t channels, uint16_t frame_len,
|
|
int32_t **sample_buffer)
|
|
{
|
|
uint8_t ch, ch1;
|
|
uint16_t i;
|
|
|
|
switch (CONV(channels,hDecoder->downMatrix))
|
|
{
|
|
case CONV(1,0):
|
|
case CONV(1,1):
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = input[hDecoder->internal_channel[0]][i];
|
|
|
|
inp *= 256.0f;
|
|
CLIP(inp, 8388607.0f, -8388608.0f);
|
|
|
|
(*sample_buffer)[i] = (int32_t)lrintf(inp);
|
|
}
|
|
break;
|
|
case CONV(2,0):
|
|
if (hDecoder->upMatrix)
|
|
{
|
|
ch = hDecoder->internal_channel[0];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch][i];
|
|
|
|
inp0 *= 256.0f;
|
|
CLIP(inp0, 8388607.0f, -8388608.0f);
|
|
|
|
(*sample_buffer)[(i*2)+0] = (int32_t)lrintf(inp0);
|
|
(*sample_buffer)[(i*2)+1] = (int32_t)lrintf(inp0);
|
|
}
|
|
} else {
|
|
ch = hDecoder->internal_channel[0];
|
|
ch1 = hDecoder->internal_channel[1];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch ][i];
|
|
real_t inp1 = input[ch1][i];
|
|
|
|
inp0 *= 256.0f;
|
|
inp1 *= 256.0f;
|
|
CLIP(inp0, 8388607.0f, -8388608.0f);
|
|
CLIP(inp1, 8388607.0f, -8388608.0f);
|
|
|
|
(*sample_buffer)[(i*2)+0] = (int32_t)lrintf(inp0);
|
|
(*sample_buffer)[(i*2)+1] = (int32_t)lrintf(inp1);
|
|
}
|
|
}
|
|
break;
|
|
default:
|
|
for (ch = 0; ch < channels; ch++)
|
|
{
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
|
|
|
|
inp *= 256.0f;
|
|
CLIP(inp, 8388607.0f, -8388608.0f);
|
|
|
|
(*sample_buffer)[(i*channels)+ch] = (int32_t)lrintf(inp);
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
static void to_PCM_32bit(NeAACDecHandle hDecoder, real_t **input,
|
|
uint8_t channels, uint16_t frame_len,
|
|
int32_t **sample_buffer)
|
|
{
|
|
uint8_t ch, ch1;
|
|
uint16_t i;
|
|
|
|
switch (CONV(channels,hDecoder->downMatrix))
|
|
{
|
|
case CONV(1,0):
|
|
case CONV(1,1):
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = input[hDecoder->internal_channel[0]][i];
|
|
|
|
inp *= 65536.0f;
|
|
CLIP(inp, 2147483647.0f, -2147483648.0f);
|
|
|
|
(*sample_buffer)[i] = (int32_t)lrintf(inp);
|
|
}
|
|
break;
|
|
case CONV(2,0):
|
|
if (hDecoder->upMatrix)
|
|
{
|
|
ch = hDecoder->internal_channel[0];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch][i];
|
|
|
|
inp0 *= 65536.0f;
|
|
CLIP(inp0, 2147483647.0f, -2147483648.0f);
|
|
|
|
(*sample_buffer)[(i*2)+0] = (int32_t)lrintf(inp0);
|
|
(*sample_buffer)[(i*2)+1] = (int32_t)lrintf(inp0);
|
|
}
|
|
} else {
|
|
ch = hDecoder->internal_channel[0];
|
|
ch1 = hDecoder->internal_channel[1];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch ][i];
|
|
real_t inp1 = input[ch1][i];
|
|
|
|
inp0 *= 65536.0f;
|
|
inp1 *= 65536.0f;
|
|
CLIP(inp0, 2147483647.0f, -2147483648.0f);
|
|
CLIP(inp1, 2147483647.0f, -2147483648.0f);
|
|
|
|
(*sample_buffer)[(i*2)+0] = (int32_t)lrintf(inp0);
|
|
(*sample_buffer)[(i*2)+1] = (int32_t)lrintf(inp1);
|
|
}
|
|
}
|
|
break;
|
|
default:
|
|
for (ch = 0; ch < channels; ch++)
|
|
{
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
|
|
|
|
inp *= 65536.0f;
|
|
CLIP(inp, 2147483647.0f, -2147483648.0f);
|
|
|
|
(*sample_buffer)[(i*channels)+ch] = (int32_t)lrintf(inp);
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
static void to_PCM_float(NeAACDecHandle hDecoder, real_t **input,
|
|
uint8_t channels, uint16_t frame_len,
|
|
float32_t **sample_buffer)
|
|
{
|
|
uint8_t ch, ch1;
|
|
uint16_t i;
|
|
|
|
switch (CONV(channels,hDecoder->downMatrix))
|
|
{
|
|
case CONV(1,0):
|
|
case CONV(1,1):
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = input[hDecoder->internal_channel[0]][i];
|
|
(*sample_buffer)[i] = inp*FLOAT_SCALE;
|
|
}
|
|
break;
|
|
case CONV(2,0):
|
|
if (hDecoder->upMatrix)
|
|
{
|
|
ch = hDecoder->internal_channel[0];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch][i];
|
|
(*sample_buffer)[(i*2)+0] = inp0*FLOAT_SCALE;
|
|
(*sample_buffer)[(i*2)+1] = inp0*FLOAT_SCALE;
|
|
}
|
|
} else {
|
|
ch = hDecoder->internal_channel[0];
|
|
ch1 = hDecoder->internal_channel[1];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch ][i];
|
|
real_t inp1 = input[ch1][i];
|
|
(*sample_buffer)[(i*2)+0] = inp0*FLOAT_SCALE;
|
|
(*sample_buffer)[(i*2)+1] = inp1*FLOAT_SCALE;
|
|
}
|
|
}
|
|
break;
|
|
default:
|
|
for (ch = 0; ch < channels; ch++)
|
|
{
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
|
|
(*sample_buffer)[(i*channels)+ch] = inp*FLOAT_SCALE;
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
static void to_PCM_double(NeAACDecHandle hDecoder, real_t **input,
|
|
uint8_t channels, uint16_t frame_len,
|
|
double **sample_buffer)
|
|
{
|
|
uint8_t ch, ch1;
|
|
uint16_t i;
|
|
|
|
switch (CONV(channels,hDecoder->downMatrix))
|
|
{
|
|
case CONV(1,0):
|
|
case CONV(1,1):
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = input[hDecoder->internal_channel[0]][i];
|
|
(*sample_buffer)[i] = (double)inp*FLOAT_SCALE;
|
|
}
|
|
break;
|
|
case CONV(2,0):
|
|
if (hDecoder->upMatrix)
|
|
{
|
|
ch = hDecoder->internal_channel[0];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch][i];
|
|
(*sample_buffer)[(i*2)+0] = (double)inp0*FLOAT_SCALE;
|
|
(*sample_buffer)[(i*2)+1] = (double)inp0*FLOAT_SCALE;
|
|
}
|
|
} else {
|
|
ch = hDecoder->internal_channel[0];
|
|
ch1 = hDecoder->internal_channel[1];
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp0 = input[ch ][i];
|
|
real_t inp1 = input[ch1][i];
|
|
(*sample_buffer)[(i*2)+0] = (double)inp0*FLOAT_SCALE;
|
|
(*sample_buffer)[(i*2)+1] = (double)inp1*FLOAT_SCALE;
|
|
}
|
|
}
|
|
break;
|
|
default:
|
|
for (ch = 0; ch < channels; ch++)
|
|
{
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
|
|
(*sample_buffer)[(i*channels)+ch] = (double)inp*FLOAT_SCALE;
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
void *output_to_PCM(NeAACDecHandle hDecoder,
|
|
real_t **input, void *sample_buffer, uint8_t channels,
|
|
uint16_t frame_len, uint8_t format)
|
|
{
|
|
int16_t *short_sample_buffer = (int16_t*)sample_buffer;
|
|
int32_t *int_sample_buffer = (int32_t*)sample_buffer;
|
|
float32_t *float_sample_buffer = (float32_t*)sample_buffer;
|
|
double *double_sample_buffer = (double*)sample_buffer;
|
|
|
|
#ifdef PROFILE
|
|
int64_t count = faad_get_ts();
|
|
#endif
|
|
|
|
/* Copy output to a standard PCM buffer */
|
|
switch (format)
|
|
{
|
|
case FAAD_FMT_16BIT:
|
|
to_PCM_16bit(hDecoder, input, channels, frame_len, &short_sample_buffer);
|
|
break;
|
|
case FAAD_FMT_24BIT:
|
|
to_PCM_24bit(hDecoder, input, channels, frame_len, &int_sample_buffer);
|
|
break;
|
|
case FAAD_FMT_32BIT:
|
|
to_PCM_32bit(hDecoder, input, channels, frame_len, &int_sample_buffer);
|
|
break;
|
|
case FAAD_FMT_FLOAT:
|
|
to_PCM_float(hDecoder, input, channels, frame_len, &float_sample_buffer);
|
|
break;
|
|
case FAAD_FMT_DOUBLE:
|
|
to_PCM_double(hDecoder, input, channels, frame_len, &double_sample_buffer);
|
|
break;
|
|
}
|
|
|
|
#ifdef PROFILE
|
|
count = faad_get_ts() - count;
|
|
hDecoder->output_cycles += count;
|
|
#endif
|
|
|
|
return sample_buffer;
|
|
}
|
|
|
|
#else
|
|
|
|
#define DM_MUL FRAC_CONST(0.3203772410170407) // 1/(1+sqrt(2) + 1/sqrt(2))
|
|
#define RSQRT2 FRAC_CONST(0.7071067811865475244) // 1/sqrt(2)
|
|
|
|
static INLINE real_t get_sample(real_t **input, uint8_t channel, uint16_t sample,
|
|
uint8_t down_matrix, uint8_t up_matrix,
|
|
uint8_t *internal_channel)
|
|
{
|
|
if (up_matrix == 1)
|
|
return input[internal_channel[0]][sample];
|
|
|
|
if (!down_matrix)
|
|
return input[internal_channel[channel]][sample];
|
|
|
|
if (channel == 0)
|
|
{
|
|
real_t C = MUL_F(input[internal_channel[0]][sample], RSQRT2);
|
|
real_t L_S = MUL_F(input[internal_channel[3]][sample], RSQRT2);
|
|
real_t cum = input[internal_channel[1]][sample] + C + L_S;
|
|
return MUL_F(cum, DM_MUL);
|
|
} else {
|
|
real_t C = MUL_F(input[internal_channel[0]][sample], RSQRT2);
|
|
real_t R_S = MUL_F(input[internal_channel[4]][sample], RSQRT2);
|
|
real_t cum = input[internal_channel[2]][sample] + C + R_S;
|
|
return MUL_F(cum, DM_MUL);
|
|
}
|
|
}
|
|
|
|
void* output_to_PCM_sux(NeAACDecHandle hDecoder,
|
|
real_t **input, void *sample_buffer, uint8_t channels,
|
|
uint16_t frame_len, uint8_t format)
|
|
{
|
|
uint8_t ch;
|
|
uint16_t i;
|
|
int16_t *short_sample_buffer = (int16_t*)sample_buffer;
|
|
int32_t *int_sample_buffer = (int32_t*)sample_buffer;
|
|
|
|
/* Copy output to a standard PCM buffer */
|
|
for (ch = 0; ch < channels; ch++)
|
|
{
|
|
switch (format)
|
|
{
|
|
case FAAD_FMT_16BIT:
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
|
|
hDecoder->internal_channel);
|
|
if (tmp >= 0)
|
|
{
|
|
tmp += (1 << (REAL_BITS-1));
|
|
if (tmp >= REAL_CONST(32767))
|
|
{
|
|
tmp = REAL_CONST(32767);
|
|
}
|
|
} else {
|
|
tmp += -(1 << (REAL_BITS-1));
|
|
if (tmp <= REAL_CONST(-32768))
|
|
{
|
|
tmp = REAL_CONST(-32768);
|
|
}
|
|
}
|
|
tmp >>= REAL_BITS;
|
|
short_sample_buffer[(i*channels)+ch] = (int16_t)tmp;
|
|
}
|
|
break;
|
|
case FAAD_FMT_24BIT:
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
|
|
hDecoder->internal_channel);
|
|
if (tmp >= 0)
|
|
{
|
|
tmp += (1 << (REAL_BITS-9));
|
|
tmp >>= (REAL_BITS-8);
|
|
if (tmp >= 8388607)
|
|
{
|
|
tmp = 8388607;
|
|
}
|
|
} else {
|
|
tmp += -(1 << (REAL_BITS-9));
|
|
tmp >>= (REAL_BITS-8);
|
|
if (tmp <= -8388608)
|
|
{
|
|
tmp = -8388608;
|
|
}
|
|
}
|
|
int_sample_buffer[(i*channels)+ch] = (int32_t)tmp;
|
|
}
|
|
break;
|
|
case FAAD_FMT_32BIT:
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
|
|
hDecoder->internal_channel);
|
|
if (tmp >= 0)
|
|
{
|
|
tmp += (1 << (16-REAL_BITS-1));
|
|
tmp <<= (16-REAL_BITS);
|
|
} else {
|
|
tmp += -(1 << (16-REAL_BITS-1));
|
|
tmp <<= (16-REAL_BITS);
|
|
}
|
|
int_sample_buffer[(i*channels)+ch] = (int32_t)tmp;
|
|
}
|
|
break;
|
|
case FAAD_FMT_FIXED:
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
real_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
|
|
hDecoder->internal_channel);
|
|
int_sample_buffer[(i*channels)+ch] = (int32_t)tmp;
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
return sample_buffer;
|
|
}
|
|
|
|
void* output_to_PCM(NeAACDecHandle hDecoder,
|
|
real_t **input, void *sample_buffer, uint8_t channels,
|
|
uint16_t frame_len, uint8_t format)
|
|
{
|
|
int ch;
|
|
int i;
|
|
int16_t *short_sample_buffer = (int16_t*)sample_buffer;
|
|
real_t *ch0 = input[hDecoder->internal_channel[0]];
|
|
real_t *ch1 = input[hDecoder->internal_channel[1]];
|
|
real_t *ch2 = input[hDecoder->internal_channel[2]];
|
|
real_t *ch3 = input[hDecoder->internal_channel[3]];
|
|
real_t *ch4 = input[hDecoder->internal_channel[4]];
|
|
|
|
if (format != FAAD_FMT_16BIT)
|
|
return output_to_PCM_sux(hDecoder, input, sample_buffer, channels, frame_len, format);
|
|
|
|
if (hDecoder->downMatrix) {
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
int32_t tmp;
|
|
tmp = (ch1[i] + ((ch0[i]+ch3[i])>>1) + ((ch0[i]+ch3[i])>>2) + (1<<(REAL_BITS))) >> (REAL_BITS+1);
|
|
if ((tmp+0x8000) & ~0xffff) tmp = ~(tmp>>31)-0x8000;
|
|
short_sample_buffer[0] = tmp;
|
|
tmp = (ch2[i] + ((ch0[i]+ch4[i])>>1) + ((ch0[i]+ch4[i])>>2) + (1<<(REAL_BITS))) >> (REAL_BITS+1);
|
|
if ((tmp+0x8000) & ~0xffff) tmp = ~(tmp>>31)-0x8000;
|
|
short_sample_buffer[1] = tmp;
|
|
short_sample_buffer += channels;
|
|
}
|
|
return sample_buffer;
|
|
}
|
|
|
|
/* Copy output to a standard PCM buffer */
|
|
for(i = 0; i < frame_len; i++)
|
|
{
|
|
for (ch = 0; ch < channels; ch++)
|
|
{
|
|
int32_t tmp = input[hDecoder->internal_channel[ch]][i];
|
|
tmp += (1 << (REAL_BITS-1));
|
|
tmp >>= REAL_BITS;
|
|
if ((tmp+0x8000) & ~0xffff) tmp = ~(tmp>>31)-0x8000;
|
|
*(short_sample_buffer++) = tmp;
|
|
}
|
|
}
|
|
|
|
return sample_buffer;
|
|
}
|
|
|
|
#endif
|