rockbox/lib/rbcodec/codecs/wav_enc.c

247 lines
7.9 KiB
C
Raw Normal View History

/***************************************************************************
* __________ __ ___.
* Open \______ \ ____ ____ | | _\_ |__ _______ ___
* Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
* Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
* Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
* \/ \/ \/ \/ \/
* $Id$
*
* Copyright (C) 2006 Antonius Hellmann
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
* Copyright (C) 2006-2013 Michael Sevakis
*
* 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 software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
* KIND, either express or implied.
*
****************************************************************************/
#include <inttypes.h>
#include <stdio.h>
#include "codeclib.h"
CODEC_ENC_HEADER
struct riff_header
{
uint8_t riff_id[4]; /* 00h - "RIFF" */
uint32_t riff_size; /* 04h - sz following headers + data_size */
/* format header */
uint8_t format[4]; /* 08h - "WAVE" */
uint8_t format_id[4]; /* 0Ch - "fmt " */
uint32_t format_size; /* 10h - 16 for PCM (sz format data) */
/* format data */
uint16_t audio_format; /* 14h - 1=PCM */
uint16_t num_channels; /* 16h - 1=M, 2=S, etc. */
uint32_t sample_rate; /* 18h - HZ */
uint32_t byte_rate; /* 1Ch - num_channels*sample_rate*bits_per_sample/8 */
uint16_t block_align; /* 20h - num_channels*bits_per_samples/8 */
uint16_t bits_per_sample; /* 22h - 8=8 bits, 16=16 bits, etc. */
/* Not for audio_format=1 (PCM) */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* uint16_t extra_param_size; 24h - size of extra data */
/* uint8_t extra_params[extra_param_size]; */
/* data header */
uint8_t data_id[4]; /* 24h - "data" */
uint32_t data_size; /* 28h - num_samples*num_channels*bits_per_sample/8 */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* uint8_t data[data_size]; 2Ch - actual sound data */
} __attribute__((packed));
#define RIFF_FMT_HEADER_SIZE 12 /* format -> format_size */
#define RIFF_FMT_DATA_SIZE 16 /* audio_format -> bits_per_sample */
#define RIFF_DATA_HEADER_SIZE 8 /* data_id -> data_size */
#define PCM_DEPTH_BYTES 2
#define PCM_DEPTH_BITS 16
#define PCM_SAMP_PER_CHUNK 2048
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
static int num_channels;
static uint32_t sample_rate;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
static size_t frame_size;
static size_t data_size;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
static const struct riff_header riff_template_header =
{
/* "RIFF" header */
{ 'R', 'I', 'F', 'F' }, /* riff_id */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
0, /* riff_size (*) */
/* format header */
{ 'W', 'A', 'V', 'E' }, /* format */
{ 'f', 'm', 't', ' ' }, /* format_id */
htole32(16), /* format_size */
/* format data */
htole16(1), /* audio_format */
0, /* num_channels (*) */
0, /* sample_rate (*) */
0, /* byte_rate (*) */
0, /* block_align (*) */
htole16(PCM_DEPTH_BITS), /* bits_per_sample */
/* data header */
{ 'd', 'a', 't', 'a' }, /* data_id */
0 /* data_size (*) */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* (*) updated when finalizing stream */
};
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
static inline void frame_htole(uint32_t *p, size_t size)
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
#ifdef ROCKBOX_BIG_ENDIAN
/* Byte-swap samples, stereo or mono */
do
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
uint32_t t;
t = swap_odd_even32(*p); *p++ = t;
t = swap_odd_even32(*p); *p++ = t;
t = swap_odd_even32(*p); *p++ = t;
t = swap_odd_even32(*p); *p++ = t;
t = swap_odd_even32(*p); *p++ = t;
t = swap_odd_even32(*p); *p++ = t;
t = swap_odd_even32(*p); *p++ = t;
t = swap_odd_even32(*p); *p++ = t;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
while (size -= 8 * 2 * PCM_DEPTH_BYTES);
#endif /* ROCKBOX_BIG_ENDIAN */
(void)p; (void)size;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
static int on_stream_data(struct enc_chunk_data *data)
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
size_t size = data->hdr.size;
if (ci->enc_stream_write(data->data, size) != (ssize_t)size)
return -1;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
data_size += size;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
return 0;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
static int on_stream_start(void)
{
/* reset sample count */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
data_size = 0;
/* write template header */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
if (ci->enc_stream_write(&riff_template_header, sizeof (struct riff_header))
!= sizeof (struct riff_header))
return -1;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
return 0;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
static int on_stream_end(union enc_chunk_hdr *hdr)
{
/* update template header */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
struct riff_header riff;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
if (hdr->err)
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* Called for stream error; get correct data size */
ssize_t size = ci->enc_stream_lseek(0, SEEK_END);
if (size > (ssize_t)sizeof (riff))
data_size = size - sizeof (riff);
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
if (ci->enc_stream_lseek(0, SEEK_SET) ||
ci->enc_stream_read(&riff, sizeof (riff)) != sizeof (riff))
return -1;
/* "RIFF" header */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
riff.riff_size = htole32(RIFF_FMT_HEADER_SIZE + RIFF_FMT_DATA_SIZE
+ RIFF_DATA_HEADER_SIZE + data_size);
/* format data */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
riff.num_channels = htole16(num_channels);
riff.sample_rate = htole32(sample_rate);
riff.byte_rate = htole32(sample_rate*num_channels*PCM_DEPTH_BYTES);
riff.block_align = htole16(num_channels*PCM_DEPTH_BYTES);
/* data header */
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
riff.data_size = htole32(data_size);
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
if (ci->enc_stream_lseek(0, SEEK_SET) != 0)
return -2;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
if (ci->enc_stream_write(&riff, sizeof (riff)) != sizeof (riff))
return -3;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
return 0;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* this is the codec entry point */
enum codec_status codec_main(enum codec_entry_call_reason reason)
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
return CODEC_OK;
(void)reason;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* this is called for each file to process */
enum codec_status ICODE_ATTR codec_run(void)
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
enum { GETBUF_ENC, GETBUF_PCM } getbuf = GETBUF_ENC;
struct enc_chunk_data *data = NULL;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* main encoding loop */
while (1)
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
enum codec_command_action action = ci->get_command(NULL);
if (action != CODEC_ACTION_NULL)
break;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* First obtain output buffer; when available, get PCM data */
switch (getbuf)
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
case GETBUF_ENC:
if (!(data = ci->enc_encbuf_get_buffer(frame_size)))
continue;
getbuf = GETBUF_PCM;
case GETBUF_PCM:
if (!ci->enc_pcmbuf_read(data->data, PCM_SAMP_PER_CHUNK))
continue;
getbuf = GETBUF_ENC;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
data->hdr.size = frame_size;
data->pcm_count = PCM_SAMP_PER_CHUNK;
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
frame_htole((uint32_t *)data->data, frame_size);
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
ci->enc_pcmbuf_advance(PCM_SAMP_PER_CHUNK);
ci->enc_encbuf_finish_buffer();
}
return CODEC_OK;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
/* this is called by recording system */
int ICODE_ATTR enc_callback(enum enc_callback_reason reason,
void *params)
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
if (LIKELY(reason == ENC_CB_STREAM))
{
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
switch (((union enc_chunk_hdr *)params)->type)
{
case CHUNK_T_DATA:
return on_stream_data(params);
case CHUNK_T_STREAM_START:
return on_stream_start();
case CHUNK_T_STREAM_END:
return on_stream_end(params);
}
}
else if (reason == ENC_CB_INPUTS)
{
struct enc_inputs *inputs = params;
sample_rate = inputs->sample_rate;
num_channels = inputs->num_channels;
frame_size = PCM_SAMP_PER_CHUNK*PCM_DEPTH_BYTES*num_channels;
}
Update software recording engine to latest codec interface. Basically, just give it a good rewrite. Software codec recording can be implemented in a more straightforward and simple manner and made more robust through the better codec control now available. Encoded audio buffer uses a packed format instead of fixed-size chunks and uses smaller data headers leading to more efficient usage. The greatest benefit is with a VBR format like wavpack which needs to request a maximum size but only actually ends up committing part of that request. No guard buffers are used for either PCM or encoded audio. PCM is read into the codec's provided buffer and mono conversion done at that time in the core if required. Any highly-specialized sample conversion is still done within the codec itself, such as 32-bit (wavpack) or interleaved mono (mp3). There is no longer a separate filename array. All metadata goes onto the main encoded audio buffer, eliminating any predermined file limit on the buffer as well as not wasting the space for unused path queue slots. The core and codec interface is less awkward and a bit more sensible. Some less useful interface features were removed. Threads are kept on narrow code paths ie. the audio thread never calls encoding functions and the codec thread never calls file functions as before. Codecs no longer call file functions directly. Writes are buffered in the core and data written to storage in larger chunks to speed up flushing of data. In fact, codecs are no longer aware of the stream being a file at all and have no access to the fd. SPDIF frequency detection no longer requires a restart of recording or plugging the source before entering the screen. It will poll for changes and update when stopped or prerecording (which does discard now-invalid prerecorded data). I've seen to it that writing a proper header on full disk works when the format makes it reasonably practical to do so. Other cases may have incorrect data sizes but sample info will be in tact. File left that way may play anyway. mp3_enc.codec acquires the ability to write 'Info' headers with LAME tags to make it gapless (bonus). Change-Id: I670685166d5eb32ef58ef317f50b8af766ceb653 Reviewed-on: http://gerrit.rockbox.org/493 Reviewed-by: Michael Sevakis <jethead71@rockbox.org> Tested-by: Michael Sevakis <jethead71@rockbox.org>
2013-06-22 20:41:16 +00:00
return 0;
}