/proc/self/cwd/libfaad/output.c

Source
/*
** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
** Copyright (C) 2003-2005 M. Bakker, Nero 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.
**
** The "appropriate copyright message" mentioned in section 2c of the GPLv2
** must read: "Code from FAAD2 is copyright (c) Nero AG, www.nero.com"
**
** Commercial non-GPL licensing of this software is possible.
** For more info contact Nero AG through Mpeg4AAClicense@nero.com.
**
** $Id: output.c,v 1.47 2009/01/26 23:51:15 menno Exp $
**/

#include "common.h"
#include "structs.h"

#include "output.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(NeAACDecStruct *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(NeAACDecStruct *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(NeAACDecStruct *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(NeAACDecStruct *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(NeAACDecStruct *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(NeAACDecStruct *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)
#define BOTH FRAC_CONST(0.2265409196609864215998) // 1/(sqrt(2) + 2 + 1)

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)
{
    real_t C;
    if (up_matrix == 1)
        return input[internal_channel[0]][sample];

    if (!down_matrix)
        return input[internal_channel[channel]][sample];

    C = MUL_F(input[internal_channel[0]][sample], BOTH);
    if (channel == 0)
    {
        real_t L_S = MUL_F(input[internal_channel[3]][sample], BOTH);
        real_t core = MUL_F(input[internal_channel[1]][sample], DM_MUL);
        return core + C + L_S;
    } else {
        real_t R_S = MUL_F(input[internal_channel[4]][sample], BOTH);
        real_t core = MUL_F(input[internal_channel[2]][sample], DM_MUL);
        return core + C + R_S;
    }
}

void* output_to_PCM(NeAACDecStruct *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;
    int32_t exp, half, sat_shift_mask;

    /* 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)
                {
                    if (tmp <= 0x7FFFFFFF - (1 << (REAL_BITS-1)))
                        tmp += (1 << (REAL_BITS-1));
                    if (tmp >= REAL_CONST(32767))
                    {
                        tmp = REAL_CONST(32767);
                    }
                } else {
                    if (tmp >= (int32_t)0x80000000 + (1 << (REAL_BITS-1)))
                        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)
                {
                    if (tmp <= 0x7FFFFFFF - (1 << (REAL_BITS-9)))
                        tmp += (1 << (REAL_BITS-9));
                    tmp >>= (REAL_BITS-8);
                    if (tmp >= 8388607)
                    {
                        tmp = 8388607;
                    }
                } else {
                    if (tmp >= (int32_t)0x80000000 + (1 << (REAL_BITS-9)))
                        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:
            exp = 16 - REAL_BITS;
            half = 1 << (exp - 1);
            sat_shift_mask = SAT_SHIFT_MASK(exp);
            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)
                {
                    if (tmp <= 0x7FFFFFFF - half)
                        tmp += half;
                } else {
                    if (tmp >= (int32_t)0x80000000 + half)
                        tmp += -half;
                }
                tmp = SAT_SHIFT(tmp, exp, sat_shift_mask);
                int_sample_buffer[(i*channels)+ch] = 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;
}

#endif

Coverage Report

Created: 2026-10-17 06:34

Line	Count	Source
1		/*
2		** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
3		** Copyright (C) 2003-2005 M. Bakker, Nero AG, http://www.nero.com
4		**
5		** This program is free software; you can redistribute it and/or modify
6		** it under the terms of the GNU General Public License as published by
7		** the Free Software Foundation; either version 2 of the License, or
8		** (at your option) any later version.
9		**
10		** This program is distributed in the hope that it will be useful,
11		** but WITHOUT ANY WARRANTY; without even the implied warranty of
12		** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13		** GNU General Public License for more details.
14		**
15		** You should have received a copy of the GNU General Public License
16		** along with this program; if not, write to the Free Software
17		** Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
18		**
19		** Any non-GPL usage of this software or parts of this software is strictly
20		** forbidden.
21		**
22		** The "appropriate copyright message" mentioned in section 2c of the GPLv2
23		** must read: "Code from FAAD2 is copyright (c) Nero AG, www.nero.com"
24		**
25		** Commercial non-GPL licensing of this software is possible.
26		** For more info contact Nero AG through Mpeg4AAClicense@nero.com.
27		**
28		** $Id: output.c,v 1.47 2009/01/26 23:51:15 menno Exp $
29		**/
30
31		#include "common.h"
32		#include "structs.h"
33
34		#include "output.h"
35
36		#ifndef FIXED_POINT
37
38
39	8.18M	#define FLOAT_SCALE (1.0f/(1<<15))
40
41	1.33M	#define DM_MUL REAL_CONST(0.3203772410170407) // 1/(1+sqrt(2) + 1/sqrt(2))
42	2.66M	#define RSQRT2 REAL_CONST(0.7071067811865475244) // 1/sqrt(2)
43
44
45		static INLINE real_t get_sample(real_t **input, uint8_t channel, uint16_t sample,
46		uint8_t down_matrix, uint8_t *internal_channel)
47	24.1M	{
48	24.1M	if (!down_matrix)
49	22.8M	return input[internal_channel[channel]][sample];
50
51	1.33M	if (channel == 0)
52	653k	{
53	653k	return DM_MUL * (input[internal_channel[1]][sample] +
54	653k	input[internal_channel[0]][sample] * RSQRT2 +
55	653k	input[internal_channel[3]][sample] * RSQRT2);
56	677k	} else {
57	677k	return DM_MUL * (input[internal_channel[2]][sample] +
58	677k	input[internal_channel[0]][sample] * RSQRT2 +
59	677k	input[internal_channel[4]][sample] * RSQRT2);
60	677k	}
61	1.33M	}
62
63		#ifndef HAS_LRINTF
64	17.5M	#define CLIP(sample, max, min) \
65	17.5M	if (sample >= 0.0f) \
66	15.5M	{ \
67	15.5M	sample += 0.5f; \
68	15.5M	if (sample >= max) \
69	15.5M	sample = max; \
70	15.5M	} else { \
71	1.98M	sample += -0.5f; \
72	1.98M	if (sample <= min) \
73	1.98M	sample = min; \
74	1.98M	}
75		#else
76		#define CLIP(sample, max, min) \
77		if (sample >= 0.0f) \
78		{ \
79		if (sample >= max) \
80		sample = max; \
81		} else { \
82		if (sample <= min) \
83		sample = min; \
84		}
85		#endif
86
87	4.06k	#define CONV(a,b) ((a<<1)\|(b&0x1))
88
89		static void to_PCM_16bit(NeAACDecStruct hDecoder, real_t *input,
90		uint8_t channels, uint16_t frame_len,
91		int16_t **sample_buffer)
92	791	{
93	791	uint8_t ch, ch1;
94	791	uint16_t i;
95
96	791	switch (CONV(channels,hDecoder->downMatrix))
97	791	{
98	165	case CONV(1,0):
99	166	case CONV(1,1):
100	200k	for(i = 0; i < frame_len; i++)
101	200k	{
102	200k	real_t inp = input[hDecoder->internal_channel[0]][i];
103
104	200k	CLIP(inp, 32767.0f, -32768.0f);
105
106	200k	(*sample_buffer)[i] = (int16_t)lrintf(inp);
107	200k	}
108	166	break;
109	114	case CONV(2,0):
110	114	if (hDecoder->upMatrix)
111	0	{
112	0	ch = hDecoder->internal_channel[0];
113	0	for(i = 0; i < frame_len; i++)
114	0	{
115	0	real_t inp0 = input[ch][i];
116
117	0	CLIP(inp0, 32767.0f, -32768.0f);
118
119	0	(sample_buffer)[(i2)+0] = (int16_t)lrintf(inp0);
120	0	(sample_buffer)[(i2)+1] = (int16_t)lrintf(inp0);
121	0	}
122	114	} else {
123	114	ch = hDecoder->internal_channel[0];
124	114	ch1 = hDecoder->internal_channel[1];
125	166k	for(i = 0; i < frame_len; i++)
126	166k	{
127	166k	real_t inp0 = input[ch ][i];
128	166k	real_t inp1 = input[ch1][i];
129
130	166k	CLIP(inp0, 32767.0f, -32768.0f);
131	166k	CLIP(inp1, 32767.0f, -32768.0f);
132
133	166k	(sample_buffer)[(i2)+0] = (int16_t)lrintf(inp0);
134	166k	(sample_buffer)[(i2)+1] = (int16_t)lrintf(inp1);
135	166k	}
136	114	}
137	114	break;
138	511	default:
139	5.22k	for (ch = 0; ch < channels; ch++)
140	4.71k	{
141	6.90M	for(i = 0; i < frame_len; i++)
142	6.89M	{
143	6.89M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
144
145	6.89M	CLIP(inp, 32767.0f, -32768.0f);
146
147	6.89M	(sample_buffer)[(ichannels)+ch] = (int16_t)lrintf(inp);
148	6.89M	}
149	4.71k	}
150	511	break;
151	791	}
152	791	}
153
154		static void to_PCM_24bit(NeAACDecStruct hDecoder, real_t *input,
155		uint8_t channels, uint16_t frame_len,
156		int32_t **sample_buffer)
157	670	{
158	670	uint8_t ch, ch1;
159	670	uint16_t i;
160
161	670	switch (CONV(channels,hDecoder->downMatrix))
162	670	{
163	154	case CONV(1,0):
164	155	case CONV(1,1):
165	184k	for(i = 0; i < frame_len; i++)
166	183k	{
167	183k	real_t inp = input[hDecoder->internal_channel[0]][i];
168
169	183k	inp *= 256.0f;
170	183k	CLIP(inp, 8388607.0f, -8388608.0f);
171
172	183k	(*sample_buffer)[i] = (int32_t)lrintf(inp);
173	183k	}
174	155	break;
175	89	case CONV(2,0):
176	89	if (hDecoder->upMatrix)
177	0	{
178	0	ch = hDecoder->internal_channel[0];
179	0	for(i = 0; i < frame_len; i++)
180	0	{
181	0	real_t inp0 = input[ch][i];
182
183	0	inp0 *= 256.0f;
184	0	CLIP(inp0, 8388607.0f, -8388608.0f);
185
186	0	(sample_buffer)[(i2)+0] = (int32_t)lrintf(inp0);
187	0	(sample_buffer)[(i2)+1] = (int32_t)lrintf(inp0);
188	0	}
189	89	} else {
190	89	ch = hDecoder->internal_channel[0];
191	89	ch1 = hDecoder->internal_channel[1];
192	138k	for(i = 0; i < frame_len; i++)
193	138k	{
194	138k	real_t inp0 = input[ch ][i];
195	138k	real_t inp1 = input[ch1][i];
196
197	138k	inp0 *= 256.0f;
198	138k	inp1 *= 256.0f;
199	138k	CLIP(inp0, 8388607.0f, -8388608.0f);
200	138k	CLIP(inp1, 8388607.0f, -8388608.0f);
201
202	138k	(sample_buffer)[(i2)+0] = (int32_t)lrintf(inp0);
203	138k	(sample_buffer)[(i2)+1] = (int32_t)lrintf(inp1);
204	138k	}
205	89	}
206	89	break;
207	426	default:
208	3.70k	for (ch = 0; ch < channels; ch++)
209	3.27k	{
210	4.58M	for(i = 0; i < frame_len; i++)
211	4.58M	{
212	4.58M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
213
214	4.58M	inp *= 256.0f;
215	4.58M	CLIP(inp, 8388607.0f, -8388608.0f);
216
217	4.58M	(sample_buffer)[(ichannels)+ch] = (int32_t)lrintf(inp);
218	4.58M	}
219	3.27k	}
220	426	break;
221	670	}
222	670	}
223
224		static void to_PCM_32bit(NeAACDecStruct hDecoder, real_t *input,
225		uint8_t channels, uint16_t frame_len,
226		int32_t **sample_buffer)
227	673	{
228	673	uint8_t ch, ch1;
229	673	uint16_t i;
230
231	673	switch (CONV(channels,hDecoder->downMatrix))
232	673	{
233	146	case CONV(1,0):
234	147	case CONV(1,1):
235	157k	for(i = 0; i < frame_len; i++)
236	157k	{
237	157k	real_t inp = input[hDecoder->internal_channel[0]][i];
238
239	157k	inp *= 65536.0f;
240	157k	CLIP(inp, 2147483647.0f, -2147483648.0f);
241
242	157k	(*sample_buffer)[i] = (int32_t)lrintf(inp);
243	157k	}
244	147	break;
245	96	case CONV(2,0):
246	96	if (hDecoder->upMatrix)
247	0	{
248	0	ch = hDecoder->internal_channel[0];
249	0	for(i = 0; i < frame_len; i++)
250	0	{
251	0	real_t inp0 = input[ch][i];
252
253	0	inp0 *= 65536.0f;
254	0	CLIP(inp0, 2147483647.0f, -2147483648.0f);
255
256	0	(sample_buffer)[(i2)+0] = (int32_t)lrintf(inp0);
257	0	(sample_buffer)[(i2)+1] = (int32_t)lrintf(inp0);
258	0	}
259	96	} else {
260	96	ch = hDecoder->internal_channel[0];
261	96	ch1 = hDecoder->internal_channel[1];
262	140k	for(i = 0; i < frame_len; i++)
263	140k	{
264	140k	real_t inp0 = input[ch ][i];
265	140k	real_t inp1 = input[ch1][i];
266
267	140k	inp0 *= 65536.0f;
268	140k	inp1 *= 65536.0f;
269	140k	CLIP(inp0, 2147483647.0f, -2147483648.0f);
270	140k	CLIP(inp1, 2147483647.0f, -2147483648.0f);
271
272	140k	(sample_buffer)[(i2)+0] = (int32_t)lrintf(inp0);
273	140k	(sample_buffer)[(i2)+1] = (int32_t)lrintf(inp1);
274	140k	}
275	96	}
276	96	break;
277	430	default:
278	3.81k	for (ch = 0; ch < channels; ch++)
279	3.38k	{
280	4.64M	for(i = 0; i < frame_len; i++)
281	4.64M	{
282	4.64M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
283
284	4.64M	inp *= 65536.0f;
285	4.64M	CLIP(inp, 2147483647.0f, -2147483648.0f);
286
287	4.64M	(sample_buffer)[(ichannels)+ch] = (int32_t)lrintf(inp);
288	4.64M	}
289	3.38k	}
290	430	break;
291	673	}
292	673	}
293
294		static void to_PCM_float(NeAACDecStruct hDecoder, real_t *input,
295		uint8_t channels, uint16_t frame_len,
296		float32_t **sample_buffer)
297	337	{
298	337	uint8_t ch, ch1;
299	337	uint16_t i;
300
301	337	switch (CONV(channels,hDecoder->downMatrix))
302	337	{
303	21	case CONV(1,0):
304	22	case CONV(1,1):
305	40.5k	for(i = 0; i < frame_len; i++)
306	40.5k	{
307	40.5k	real_t inp = input[hDecoder->internal_channel[0]][i];
308	40.5k	(sample_buffer)[i] = inpFLOAT_SCALE;
309	40.5k	}
310	22	break;
311	15	case CONV(2,0):
312	15	if (hDecoder->upMatrix)
313	0	{
314	0	ch = hDecoder->internal_channel[0];
315	0	for(i = 0; i < frame_len; i++)
316	0	{
317	0	real_t inp0 = input[ch][i];
318	0	(sample_buffer)[(i2)+0] = inp0*FLOAT_SCALE;
319	0	(sample_buffer)[(i2)+1] = inp0*FLOAT_SCALE;
320	0	}
321	15	} else {
322	15	ch = hDecoder->internal_channel[0];
323	15	ch1 = hDecoder->internal_channel[1];
324	19.7k	for(i = 0; i < frame_len; i++)
325	19.7k	{
326	19.7k	real_t inp0 = input[ch ][i];
327	19.7k	real_t inp1 = input[ch1][i];
328	19.7k	(sample_buffer)[(i2)+0] = inp0*FLOAT_SCALE;
329	19.7k	(sample_buffer)[(i2)+1] = inp1*FLOAT_SCALE;
330	19.7k	}
331	15	}
332	15	break;
333	300	default:
334	3.08k	for (ch = 0; ch < channels; ch++)
335	2.78k	{
336	4.65M	for(i = 0; i < frame_len; i++)
337	4.65M	{
338	4.65M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
339	4.65M	(sample_buffer)[(ichannels)+ch] = inp*FLOAT_SCALE;
340	4.65M	}
341	2.78k	}
342	300	break;
343	337	}
344	337	}
345
346		static void to_PCM_double(NeAACDecStruct hDecoder, real_t *input,
347		uint8_t channels, uint16_t frame_len,
348		double **sample_buffer)
349	244	{
350	244	uint8_t ch, ch1;
351	244	uint16_t i;
352
353	244	switch (CONV(channels,hDecoder->downMatrix))
354	244	{
355	20	case CONV(1,0):
356	21	case CONV(1,1):
357	31.3k	for(i = 0; i < frame_len; i++)
358	31.2k	{
359	31.2k	real_t inp = input[hDecoder->internal_channel[0]][i];
360	31.2k	(sample_buffer)[i] = (double)inpFLOAT_SCALE;
361	31.2k	}
362	21	break;
363	16	case CONV(2,0):
364	16	if (hDecoder->upMatrix)
365	0	{
366	0	ch = hDecoder->internal_channel[0];
367	0	for(i = 0; i < frame_len; i++)
368	0	{
369	0	real_t inp0 = input[ch][i];
370	0	(sample_buffer)[(i2)+0] = (double)inp0*FLOAT_SCALE;
371	0	(sample_buffer)[(i2)+1] = (double)inp0*FLOAT_SCALE;
372	0	}
373	16	} else {
374	16	ch = hDecoder->internal_channel[0];
375	16	ch1 = hDecoder->internal_channel[1];
376	21.5k	for(i = 0; i < frame_len; i++)
377	21.5k	{
378	21.5k	real_t inp0 = input[ch ][i];
379	21.5k	real_t inp1 = input[ch1][i];
380	21.5k	(sample_buffer)[(i2)+0] = (double)inp0*FLOAT_SCALE;
381	21.5k	(sample_buffer)[(i2)+1] = (double)inp1*FLOAT_SCALE;
382	21.5k	}
383	16	}
384	16	break;
385	207	default:
386	2.28k	for (ch = 0; ch < channels; ch++)
387	2.08k	{
388	3.38M	for(i = 0; i < frame_len; i++)
389	3.37M	{
390	3.37M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
391	3.37M	(sample_buffer)[(ichannels)+ch] = (double)inp*FLOAT_SCALE;
392	3.37M	}
393	2.08k	}
394	207	break;
395	244	}
396	244	}
397
398		void output_to_PCM(NeAACDecStruct hDecoder,
399		real_t *input, void sample_buffer, uint8_t channels,
400		uint16_t frame_len, uint8_t format)
401	2.71k	{
402	2.71k	int16_t short_sample_buffer = (int16_t)sample_buffer;
403	2.71k	int32_t int_sample_buffer = (int32_t)sample_buffer;
404	2.71k	float32_t float_sample_buffer = (float32_t)sample_buffer;
405	2.71k	double double_sample_buffer = (double)sample_buffer;
406
407		#ifdef PROFILE
408		int64_t count = faad_get_ts();
409		#endif
410
411		/* Copy output to a standard PCM buffer */
412	2.71k	switch (format)
413	2.71k	{
414	791	case FAAD_FMT_16BIT:
415	791	to_PCM_16bit(hDecoder, input, channels, frame_len, &short_sample_buffer);
416	791	break;
417	670	case FAAD_FMT_24BIT:
418	670	to_PCM_24bit(hDecoder, input, channels, frame_len, &int_sample_buffer);
419	670	break;
420	673	case FAAD_FMT_32BIT:
421	673	to_PCM_32bit(hDecoder, input, channels, frame_len, &int_sample_buffer);
422	673	break;
423	337	case FAAD_FMT_FLOAT:
424	337	to_PCM_float(hDecoder, input, channels, frame_len, &float_sample_buffer);
425	337	break;
426	244	case FAAD_FMT_DOUBLE:
427	244	to_PCM_double(hDecoder, input, channels, frame_len, &double_sample_buffer);
428	244	break;
429	2.71k	}
430
431		#ifdef PROFILE
432		count = faad_get_ts() - count;
433		hDecoder->output_cycles += count;
434		#endif
435
436	2.71k	return sample_buffer;
437	2.71k	}
438
439		#else
440
441		#define DM_MUL FRAC_CONST(0.3203772410170407) // 1/(1+sqrt(2) + 1/sqrt(2))
442		#define RSQRT2 FRAC_CONST(0.7071067811865475244) // 1/sqrt(2)
443		#define BOTH FRAC_CONST(0.2265409196609864215998) // 1/(sqrt(2) + 2 + 1)
444
445		static INLINE real_t get_sample(real_t **input, uint8_t channel, uint16_t sample,
446		uint8_t down_matrix, uint8_t up_matrix,
447		uint8_t *internal_channel)
448		{
449		real_t C;
450		if (up_matrix == 1)
451		return input[internal_channel[0]][sample];
452
453		if (!down_matrix)
454		return input[internal_channel[channel]][sample];
455
456		C = MUL_F(input[internal_channel[0]][sample], BOTH);
457		if (channel == 0)
458		{
459		real_t L_S = MUL_F(input[internal_channel[3]][sample], BOTH);
460		real_t core = MUL_F(input[internal_channel[1]][sample], DM_MUL);
461		return core + C + L_S;
462		} else {
463		real_t R_S = MUL_F(input[internal_channel[4]][sample], BOTH);
464		real_t core = MUL_F(input[internal_channel[2]][sample], DM_MUL);
465		return core + C + R_S;
466		}
467		}
468
469		void* output_to_PCM(NeAACDecStruct *hDecoder,
470		real_t *input, void sample_buffer, uint8_t channels,
471		uint16_t frame_len, uint8_t format)
472		{
473		uint8_t ch;
474		uint16_t i;
475		int16_t short_sample_buffer = (int16_t)sample_buffer;
476		int32_t int_sample_buffer = (int32_t)sample_buffer;
477		int32_t exp, half, sat_shift_mask;
478
479		/* Copy output to a standard PCM buffer */
480		for (ch = 0; ch < channels; ch++)
481		{
482		switch (format)
483		{
484		case FAAD_FMT_16BIT:
485		for(i = 0; i < frame_len; i++)
486		{
487		int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
488		hDecoder->internal_channel);
489		if (tmp >= 0)
490		{
491		if (tmp <= 0x7FFFFFFF - (1 << (REAL_BITS-1)))
492		tmp += (1 << (REAL_BITS-1));
493		if (tmp >= REAL_CONST(32767))
494		{
495		tmp = REAL_CONST(32767);
496		}
497		} else {
498		if (tmp >= (int32_t)0x80000000 + (1 << (REAL_BITS-1)))
499		tmp += -(1 << (REAL_BITS-1));
500		if (tmp <= REAL_CONST(-32768))
501		{
502		tmp = REAL_CONST(-32768);
503		}
504		}
505		tmp >>= REAL_BITS;
506		short_sample_buffer[(i*channels)+ch] = (int16_t)tmp;
507		}
508		break;
509		case FAAD_FMT_24BIT:
510		for(i = 0; i < frame_len; i++)
511		{
512		int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
513		hDecoder->internal_channel);
514		if (tmp >= 0)
515		{
516		if (tmp <= 0x7FFFFFFF - (1 << (REAL_BITS-9)))
517		tmp += (1 << (REAL_BITS-9));
518		tmp >>= (REAL_BITS-8);
519		if (tmp >= 8388607)
520		{
521		tmp = 8388607;
522		}
523		} else {
524		if (tmp >= (int32_t)0x80000000 + (1 << (REAL_BITS-9)))
525		tmp += -(1 << (REAL_BITS-9));
526		tmp >>= (REAL_BITS-8);
527		if (tmp <= -8388608)
528		{
529		tmp = -8388608;
530		}
531		}
532		int_sample_buffer[(i*channels)+ch] = (int32_t)tmp;
533		}
534		break;
535		case FAAD_FMT_32BIT:
536		exp = 16 - REAL_BITS;
537		half = 1 << (exp - 1);
538		sat_shift_mask = SAT_SHIFT_MASK(exp);
539		for(i = 0; i < frame_len; i++)
540		{
541		int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
542		hDecoder->internal_channel);
543		if (tmp >= 0)
544		{
545		if (tmp <= 0x7FFFFFFF - half)
546		tmp += half;
547		} else {
548		if (tmp >= (int32_t)0x80000000 + half)
549		tmp += -half;
550		}
551		tmp = SAT_SHIFT(tmp, exp, sat_shift_mask);
552		int_sample_buffer[(i*channels)+ch] = tmp;
553		}
554		break;
555		case FAAD_FMT_FIXED:
556		for(i = 0; i < frame_len; i++)
557		{
558		real_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
559		hDecoder->internal_channel);
560		int_sample_buffer[(i*channels)+ch] = (int32_t)tmp;
561		}
562		break;
563		}
564		}
565
566		return sample_buffer;
567		}
568
569		#endif