/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)
                {
                    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:
            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)
                {
                    tmp += half;
                } else {
                    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: 2025-10-13 06:47

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	6.93M	#define FLOAT_SCALE (1.0f/(1<<15))
40
41	1.29M	#define DM_MUL REAL_CONST(0.3203772410170407) // 1/(1+sqrt(2) + 1/sqrt(2))
42	2.59M	#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	24.0M	{
48	24.0M	if (!down_matrix)
49	22.7M	return input[internal_channel[channel]][sample];
50
51	1.29M	if (channel == 0)
52	636k	{
53	636k	return DM_MUL * (input[internal_channel[1]][sample] +
54	636k	input[internal_channel[0]][sample] * RSQRT2 +
55	636k	input[internal_channel[3]][sample] * RSQRT2);
56	660k	} else {
57	660k	return DM_MUL * (input[internal_channel[2]][sample] +
58	660k	input[internal_channel[0]][sample] * RSQRT2 +
59	660k	input[internal_channel[4]][sample] * RSQRT2);
60	660k	}
61	1.29M	}
62
63		#ifndef HAS_LRINTF
64	20.3M	#define CLIP(sample, max, min) \
65	20.3M	if (sample >= 0.0f) \
66	17.6M	{ \
67	17.6M	sample += 0.5f; \
68	17.6M	if (sample >= max) \
69	17.6M	sample = max; \
70	17.6M	} else { \
71	2.66M	sample += -0.5f; \
72	2.66M	if (sample <= min) \
73	2.66M	sample = min; \
74	2.66M	}
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	4.22k	#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	979	{
93	979	uint8_t ch, ch1;
94	979	uint16_t i;
95
96	979	switch (CONV(channels,hDecoder->downMatrix))
97	979	{
98	0	case CONV(1,0):
99	0	case CONV(1,1):
100	0	for(i = 0; i < frame_len; i++)
101	0	{
102	0	real_t inp = input[hDecoder->internal_channel[0]][i];
103
104	0	CLIP(inp, 32767.0f, -32768.0f);
105
106	0	(*sample_buffer)[i] = (int16_t)lrintf(inp);
107	0	}
108	0	break;
109	385	case CONV(2,0):
110	385	if (hDecoder->upMatrix)
111	68	{
112	68	ch = hDecoder->internal_channel[0];
113	115k	for(i = 0; i < frame_len; i++)
114	115k	{
115	115k	real_t inp0 = input[ch][i];
116
117	115k	CLIP(inp0, 32767.0f, -32768.0f);
118
119	115k	(sample_buffer)[(i2)+0] = (int16_t)lrintf(inp0);
120	115k	(sample_buffer)[(i2)+1] = (int16_t)lrintf(inp0);
121	115k	}
122	317	} else {
123	317	ch = hDecoder->internal_channel[0];
124	317	ch1 = hDecoder->internal_channel[1];
125	451k	for(i = 0; i < frame_len; i++)
126	450k	{
127	450k	real_t inp0 = input[ch ][i];
128	450k	real_t inp1 = input[ch1][i];
129
130	450k	CLIP(inp0, 32767.0f, -32768.0f);
131	450k	CLIP(inp1, 32767.0f, -32768.0f);
132
133	450k	(sample_buffer)[(i2)+0] = (int16_t)lrintf(inp0);
134	450k	(sample_buffer)[(i2)+1] = (int16_t)lrintf(inp1);
135	450k	}
136	317	}
137	385	break;
138	594	default:
139	6.41k	for (ch = 0; ch < channels; ch++)
140	5.82k	{
141	8.36M	for(i = 0; i < frame_len; i++)
142	8.35M	{
143	8.35M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
144
145	8.35M	CLIP(inp, 32767.0f, -32768.0f);
146
147	8.35M	(sample_buffer)[(ichannels)+ch] = (int16_t)lrintf(inp);
148	8.35M	}
149	5.82k	}
150	594	break;
151	979	}
152	979	}
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	709	{
158	709	uint8_t ch, ch1;
159	709	uint16_t i;
160
161	709	switch (CONV(channels,hDecoder->downMatrix))
162	709	{
163	0	case CONV(1,0):
164	0	case CONV(1,1):
165	0	for(i = 0; i < frame_len; i++)
166	0	{
167	0	real_t inp = input[hDecoder->internal_channel[0]][i];
168
169	0	inp *= 256.0f;
170	0	CLIP(inp, 8388607.0f, -8388608.0f);
171
172	0	(*sample_buffer)[i] = (int32_t)lrintf(inp);
173	0	}
174	0	break;
175	313	case CONV(2,0):
176	313	if (hDecoder->upMatrix)
177	36	{
178	36	ch = hDecoder->internal_channel[0];
179	52.1k	for(i = 0; i < frame_len; i++)
180	52.0k	{
181	52.0k	real_t inp0 = input[ch][i];
182
183	52.0k	inp0 *= 256.0f;
184	52.0k	CLIP(inp0, 8388607.0f, -8388608.0f);
185
186	52.0k	(sample_buffer)[(i2)+0] = (int32_t)lrintf(inp0);
187	52.0k	(sample_buffer)[(i2)+1] = (int32_t)lrintf(inp0);
188	52.0k	}
189	277	} else {
190	277	ch = hDecoder->internal_channel[0];
191	277	ch1 = hDecoder->internal_channel[1];
192	390k	for(i = 0; i < frame_len; i++)
193	390k	{
194	390k	real_t inp0 = input[ch ][i];
195	390k	real_t inp1 = input[ch1][i];
196
197	390k	inp0 *= 256.0f;
198	390k	inp1 *= 256.0f;
199	390k	CLIP(inp0, 8388607.0f, -8388608.0f);
200	390k	CLIP(inp1, 8388607.0f, -8388608.0f);
201
202	390k	(sample_buffer)[(i2)+0] = (int32_t)lrintf(inp0);
203	390k	(sample_buffer)[(i2)+1] = (int32_t)lrintf(inp1);
204	390k	}
205	277	}
206	313	break;
207	396	default:
208	3.50k	for (ch = 0; ch < channels; ch++)
209	3.11k	{
210	4.86M	for(i = 0; i < frame_len; i++)
211	4.86M	{
212	4.86M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
213
214	4.86M	inp *= 256.0f;
215	4.86M	CLIP(inp, 8388607.0f, -8388608.0f);
216
217	4.86M	(sample_buffer)[(ichannels)+ch] = (int32_t)lrintf(inp);
218	4.86M	}
219	3.11k	}
220	396	break;
221	709	}
222	709	}
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	674	{
228	674	uint8_t ch, ch1;
229	674	uint16_t i;
230
231	674	switch (CONV(channels,hDecoder->downMatrix))
232	674	{
233	0	case CONV(1,0):
234	0	case CONV(1,1):
235	0	for(i = 0; i < frame_len; i++)
236	0	{
237	0	real_t inp = input[hDecoder->internal_channel[0]][i];
238
239	0	inp *= 65536.0f;
240	0	CLIP(inp, 2147483647.0f, -2147483648.0f);
241
242	0	(*sample_buffer)[i] = (int32_t)lrintf(inp);
243	0	}
244	0	break;
245	343	case CONV(2,0):
246	343	if (hDecoder->upMatrix)
247	36	{
248	36	ch = hDecoder->internal_channel[0];
249	57.2k	for(i = 0; i < frame_len; i++)
250	57.2k	{
251	57.2k	real_t inp0 = input[ch][i];
252
253	57.2k	inp0 *= 65536.0f;
254	57.2k	CLIP(inp0, 2147483647.0f, -2147483648.0f);
255
256	57.2k	(sample_buffer)[(i2)+0] = (int32_t)lrintf(inp0);
257	57.2k	(sample_buffer)[(i2)+1] = (int32_t)lrintf(inp0);
258	57.2k	}
259	307	} else {
260	307	ch = hDecoder->internal_channel[0];
261	307	ch1 = hDecoder->internal_channel[1];
262	479k	for(i = 0; i < frame_len; i++)
263	479k	{
264	479k	real_t inp0 = input[ch ][i];
265	479k	real_t inp1 = input[ch1][i];
266
267	479k	inp0 *= 65536.0f;
268	479k	inp1 *= 65536.0f;
269	479k	CLIP(inp0, 2147483647.0f, -2147483648.0f);
270	479k	CLIP(inp1, 2147483647.0f, -2147483648.0f);
271
272	479k	(sample_buffer)[(i2)+0] = (int32_t)lrintf(inp0);
273	479k	(sample_buffer)[(i2)+1] = (int32_t)lrintf(inp1);
274	479k	}
275	307	}
276	343	break;
277	331	default:
278	3.22k	for (ch = 0; ch < channels; ch++)
279	2.89k	{
280	4.26M	for(i = 0; i < frame_len; i++)
281	4.25M	{
282	4.25M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
283
284	4.25M	inp *= 65536.0f;
285	4.25M	CLIP(inp, 2147483647.0f, -2147483648.0f);
286
287	4.25M	(sample_buffer)[(ichannels)+ch] = (int32_t)lrintf(inp);
288	4.25M	}
289	2.89k	}
290	331	break;
291	674	}
292	674	}
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	378	{
298	378	uint8_t ch, ch1;
299	378	uint16_t i;
300
301	378	switch (CONV(channels,hDecoder->downMatrix))
302	378	{
303	0	case CONV(1,0):
304	0	case CONV(1,1):
305	0	for(i = 0; i < frame_len; i++)
306	0	{
307	0	real_t inp = input[hDecoder->internal_channel[0]][i];
308	0	(sample_buffer)[i] = inpFLOAT_SCALE;
309	0	}
310	0	break;
311	63	case CONV(2,0):
312	63	if (hDecoder->upMatrix)
313	19	{
314	19	ch = hDecoder->internal_channel[0];
315	31.5k	for(i = 0; i < frame_len; i++)
316	31.4k	{
317	31.4k	real_t inp0 = input[ch][i];
318	31.4k	(sample_buffer)[(i2)+0] = inp0*FLOAT_SCALE;
319	31.4k	(sample_buffer)[(i2)+1] = inp0*FLOAT_SCALE;
320	31.4k	}
321	44	} else {
322	44	ch = hDecoder->internal_channel[0];
323	44	ch1 = hDecoder->internal_channel[1];
324	55.6k	for(i = 0; i < frame_len; i++)
325	55.6k	{
326	55.6k	real_t inp0 = input[ch ][i];
327	55.6k	real_t inp1 = input[ch1][i];
328	55.6k	(sample_buffer)[(i2)+0] = inp0*FLOAT_SCALE;
329	55.6k	(sample_buffer)[(i2)+1] = inp1*FLOAT_SCALE;
330	55.6k	}
331	44	}
332	63	break;
333	315	default:
334	2.53k	for (ch = 0; ch < channels; ch++)
335	2.22k	{
336	3.24M	for(i = 0; i < frame_len; i++)
337	3.24M	{
338	3.24M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
339	3.24M	(sample_buffer)[(ichannels)+ch] = inp*FLOAT_SCALE;
340	3.24M	}
341	2.22k	}
342	315	break;
343	378	}
344	378	}
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	291	{
350	291	uint8_t ch, ch1;
351	291	uint16_t i;
352
353	291	switch (CONV(channels,hDecoder->downMatrix))
354	291	{
355	0	case CONV(1,0):
356	0	case CONV(1,1):
357	0	for(i = 0; i < frame_len; i++)
358	0	{
359	0	real_t inp = input[hDecoder->internal_channel[0]][i];
360	0	(sample_buffer)[i] = (double)inpFLOAT_SCALE;
361	0	}
362	0	break;
363	87	case CONV(2,0):
364	87	if (hDecoder->upMatrix)
365	32	{
366	32	ch = hDecoder->internal_channel[0];
367	44.8k	for(i = 0; i < frame_len; i++)
368	44.8k	{
369	44.8k	real_t inp0 = input[ch][i];
370	44.8k	(sample_buffer)[(i2)+0] = (double)inp0*FLOAT_SCALE;
371	44.8k	(sample_buffer)[(i2)+1] = (double)inp0*FLOAT_SCALE;
372	44.8k	}
373	55	} else {
374	55	ch = hDecoder->internal_channel[0];
375	55	ch1 = hDecoder->internal_channel[1];
376	69.5k	for(i = 0; i < frame_len; i++)
377	69.5k	{
378	69.5k	real_t inp0 = input[ch ][i];
379	69.5k	real_t inp1 = input[ch1][i];
380	69.5k	(sample_buffer)[(i2)+0] = (double)inp0*FLOAT_SCALE;
381	69.5k	(sample_buffer)[(i2)+1] = (double)inp1*FLOAT_SCALE;
382	69.5k	}
383	55	}
384	87	break;
385	204	default:
386	2.20k	for (ch = 0; ch < channels; ch++)
387	1.99k	{
388	3.29M	for(i = 0; i < frame_len; i++)
389	3.29M	{
390	3.29M	real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel);
391	3.29M	(sample_buffer)[(ichannels)+ch] = (double)inp*FLOAT_SCALE;
392	3.29M	}
393	1.99k	}
394	204	break;
395	291	}
396	291	}
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	3.03k	{
402	3.03k	int16_t short_sample_buffer = (int16_t)sample_buffer;
403	3.03k	int32_t int_sample_buffer = (int32_t)sample_buffer;
404	3.03k	float32_t float_sample_buffer = (float32_t)sample_buffer;
405	3.03k	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	3.03k	switch (format)
413	3.03k	{
414	979	case FAAD_FMT_16BIT:
415	979	to_PCM_16bit(hDecoder, input, channels, frame_len, &short_sample_buffer);
416	979	break;
417	709	case FAAD_FMT_24BIT:
418	709	to_PCM_24bit(hDecoder, input, channels, frame_len, &int_sample_buffer);
419	709	break;
420	674	case FAAD_FMT_32BIT:
421	674	to_PCM_32bit(hDecoder, input, channels, frame_len, &int_sample_buffer);
422	674	break;
423	378	case FAAD_FMT_FLOAT:
424	378	to_PCM_float(hDecoder, input, channels, frame_len, &float_sample_buffer);
425	378	break;
426	291	case FAAD_FMT_DOUBLE:
427	291	to_PCM_double(hDecoder, input, channels, frame_len, &double_sample_buffer);
428	291	break;
429	3.03k	}
430
431		#ifdef PROFILE
432		count = faad_get_ts() - count;
433		hDecoder->output_cycles += count;
434		#endif
435
436	3.03k	return sample_buffer;
437	3.03k	}
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		tmp += (1 << (REAL_BITS-1));
492		if (tmp >= REAL_CONST(32767))
493		{
494		tmp = REAL_CONST(32767);
495		}
496		} else {
497		tmp += -(1 << (REAL_BITS-1));
498		if (tmp <= REAL_CONST(-32768))
499		{
500		tmp = REAL_CONST(-32768);
501		}
502		}
503		tmp >>= REAL_BITS;
504		short_sample_buffer[(i*channels)+ch] = (int16_t)tmp;
505		}
506		break;
507		case FAAD_FMT_24BIT:
508		for(i = 0; i < frame_len; i++)
509		{
510		int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
511		hDecoder->internal_channel);
512		if (tmp >= 0)
513		{
514		tmp += (1 << (REAL_BITS-9));
515		tmp >>= (REAL_BITS-8);
516		if (tmp >= 8388607)
517		{
518		tmp = 8388607;
519		}
520		} else {
521		tmp += -(1 << (REAL_BITS-9));
522		tmp >>= (REAL_BITS-8);
523		if (tmp <= -8388608)
524		{
525		tmp = -8388608;
526		}
527		}
528		int_sample_buffer[(i*channels)+ch] = (int32_t)tmp;
529		}
530		break;
531		case FAAD_FMT_32BIT:
532		exp = 16 - REAL_BITS;
533		half = 1 << (exp - 1);
534		sat_shift_mask = SAT_SHIFT_MASK(exp);
535		for(i = 0; i < frame_len; i++)
536		{
537		int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
538		hDecoder->internal_channel);
539		if (tmp >= 0)
540		{
541		tmp += half;
542		} else {
543		tmp += -half;
544		}
545		tmp = SAT_SHIFT(tmp, exp, sat_shift_mask);
546		int_sample_buffer[(i*channels)+ch] = tmp;
547		}
548		break;
549		case FAAD_FMT_FIXED:
550		for(i = 0; i < frame_len; i++)
551		{
552		real_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->upMatrix,
553		hDecoder->internal_channel);
554		int_sample_buffer[(i*channels)+ch] = (int32_t)tmp;
555		}
556		break;
557		}
558		}
559
560		return sample_buffer;
561		}
562
563		#endif