rockbox/apps/recorder/pcm_record.c

2016 lines
53 KiB
C
Raw Normal View History

/***************************************************************************
* __________ __ ___.
* Open \______ \ ____ ____ | | _\_ |__ _______ ___
* Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
* Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
* Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
* \/ \/ \/ \/ \/
* $Id$
*
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
* Copyright (C) 2005 Linus Nielsen Feltzing
* Copyright (C) 2006 Antonius Hellmann
* 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 "config.h"
#include "system.h"
#include "kernel.h"
#include "string-extra.h"
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
#include "pcm_record.h"
#include "codecs.h"
#include "logf.h"
#include "thread.h"
#include "storage.h"
#include "general.h"
#include "codec_thread.h"
#include "audio.h"
#include "sound.h"
#include "metadata.h"
#include "appevents.h"
#ifdef HAVE_SPDIF_IN
#include "spdif.h"
#endif
#include "audio_thread.h"
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
/* Macros to enable logf for queues
logging on SYS_TIMEOUT can be disabled */
#ifdef SIMULATOR
/* Define this for logf output of all queuing except SYS_TIMEOUT */
#define PCMREC_LOGQUEUES
/* Define this to logf SYS_TIMEOUT messages */
/*#define PCMREC_LOGQUEUES_SYS_TIMEOUT*/
#endif /* SIMULATOR */
#ifdef PCMREC_LOGQUEUES
#define LOGFQUEUE logf
#else
#define LOGFQUEUE(...)
#endif
#ifdef PCMREC_LOGQUEUES_SYS_TIMEOUT
#define LOGFQUEUE_SYS_TIMEOUT logf
#else
#define LOGFQUEUE_SYS_TIMEOUT(...)
#endif
/** Target-related configuration **/
/**
* PCM_NUM_CHUNKS: Number of PCM chunks
* PCM_CHUNK_SAMP: Number of samples in a PCM chunk
* PCM_BOOST_SECONDS: PCM level at which to boost CPU
* PANIC_SECONDS: Flood watermark time until full
* FLUSH_SECONDS: Flush watermark time until full
* STREAM_BUF_SIZE: Size of stream write buffer
* PRIO_SECONDS: Max flush time before prio boost
*
* Total PCM buffer size should be mem aligned
*
* Fractions should be left without parentheses so the multiplier is
* multiplied by the numerator first.
*/
#if MEMORYSIZE <= 2
#define PCM_NUM_CHUNKS 56
#define PCM_CHUNK_SAMP 1024
#define PCM_BOOST_SECONDS 1/2
#define PANIC_SECONDS 1/2
#define FLUSH_SECONDS 1
#define FLUSH_MON_INTERVAL 1/6
#define STREAM_BUF_SIZE 32768
#elif MEMORYSIZE <= 16
#define PANIC_SECONDS 5
#define FLUSH_SECONDS 7
#else /* MEMORYSIZE > 16 */
#define PANIC_SECONDS 8
#define FLUSH_SECONDS 10
#endif /* MEMORYSIZE */
/* Default values if not overridden above */
#ifndef PCM_NUM_CHUNKS
#define PCM_NUM_CHUNKS 256
#endif
#ifndef PCM_CHUNK_SAMP
#define PCM_CHUNK_SAMP 2048
#endif
#ifndef PCM_BOOST_SECONDS
#define PCM_BOOST_SECONDS 1
#endif
#ifndef FLUSH_MON_INTERVAL
#define FLUSH_MON_INTERVAL 1/4
#endif
#ifndef STREAM_BUF_SIZE
#define STREAM_BUF_SIZE 65536
#endif
#ifndef PRIO_SECONDS
#define PRIO_SECONDS 10
#endif
/* FAT limit for filesize. Recording will accept no further data from the
* codec if this limit is reached in order to preserve its own data
* integrity. A split should have made by the higher-ups long before this
* point.
*
* Leave a generous 64k margin for metadata being added to file. */
#define MAX_NUM_REC_BYTES ((size_t)0x7fff0000u)
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
/***************************************************************************/
extern struct codec_api ci; /* in codec_thread.c */
extern struct event_queue audio_queue; /* in audio_thread.c */
extern unsigned int audio_thread_id; /* in audio_thread.c */
/** General recording state **/
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
/* Recording action being performed */
static enum record_status
{
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
RECORD_STOPPED = 0,
RECORD_PRERECORDING = AUDIO_STATUS_PRERECORD,
RECORD_RECORDING = AUDIO_STATUS_RECORD,
RECORD_PAUSED = (AUDIO_STATUS_RECORD | AUDIO_STATUS_PAUSE),
} record_status = RECORD_STOPPED;
/* State of engine operations */
static enum record_state
{
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
REC_STATE_IDLE, /* Stopped or prerecording */
REC_STATE_MONITOR, /* Monitoring buffer status */
REC_STATE_FLUSH, /* Flushing buffer */
} record_state = REC_STATE_IDLE;
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 uint32_t errors; /* An error has occured (bitmask) */
static uint32_t warnings; /* Non-fatal warnings (bitmask) */
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 uint32_t rec_errors; /* Mirror of errors but private to
* avoid race with controlling
* thread. Engine uses this
* internally. */
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
/** Stats on encoded data for current file **/
static int rec_fd = -1; /* Currently open file descriptor */
static size_t num_rec_bytes; /* Number of bytes recorded */
static uint64_t num_rec_samples; /* Number of PCM samples recorded */
static uint64_t encbuf_rec_count; /* Count of slots written to buffer
for current file */
/** These apply to current settings **/
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 rec_source; /* Current rec_source setting */
static unsigned long sample_rate; /* Samplerate setting in HZ */
static int num_channels; /* Current number of channels */
static struct encoder_config enc_config; /* Current encoder configuration */
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 unsigned int pre_record_seconds; /* Pre-record time in seconds */
/****************************************************************************
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
Use 2 circular buffers:
pcm_buffer=DMA output buffer: chunks (8192 Bytes) of raw pcm audio data
enc_buffer=encoded audio buffer: storage for encoder output data
Flow:
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
1. When entering recording_screen DMA feeds the ringbuffer pcm_buffer
2. If enough pcm data are available the encoder codec does encoding of pcm
chunks (4-8192 Bytes) into ringbuffer enc_buffer in codec_thread
3. pcmrec_callback detects enc_buffer 'near full' and writes data to disk
Functions calls (basic encoder steps):
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
1.audio: codec_load(); load the encoder
2.encoder: enc_init_parameters(); set the encoder parameters (at load)
3.audio: enc_callback(); configure encoder recording settings
4.audio: codec_go(); start encoding the new stream
5.encoder: enc_encbuf_get_buffer(); obtain an output buffer of size n
6.encoder: enc_pcmbuf_read(); read n bytes of unprocessed pcm data
7.encoder: enc_encbuf_finish_buffer(); add the obtained buffer to output
8.encoder: enc_pcmbuf_advance(); advance pcm by n samples
9.encoder: while more PCM available, repeat 5. to 9.
10.audio: codec_finish_stream(); finish the output for current stream
Function calls (basic stream flushing steps through enc_callback()):
1.audio: flush_stream_start(); stream flush destination is opening
2.audio: flush_stream_data(); flush encoded audio to stream
3.audio: while encoded data available, repeat 2.
4.audio: flush_stream_end(); stream flush destination is closing
****************************************************************************/
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
/** Buffer parameters where incoming PCM data is placed **/
#define PCM_DEPTH_BYTES (sizeof (int16_t))
#define PCM_SAMP_SIZE (2*PCM_DEPTH_BYTES)
#define PCM_CHUNK_SIZE (PCM_CHUNK_SAMP*PCM_SAMP_SIZE)
#define PCM_BUF_SIZE (PCM_NUM_CHUNKS*PCM_CHUNK_SIZE)
/* Convert byte sizes into buffer slot counts */
#define CHUNK_SIZE_COUNT(size) \
(((size) + ENC_HDR_SIZE - 1) / ENC_HDR_SIZE)
#define CHUNK_FILE_COUNT(size) \
({ typeof (size) __size = (size); \
CHUNK_SIZE_COUNT(MIN(__size, MAX_PATH) + ENC_HDR_SIZE); })
#define CHUNK_FILE_COUNT_PATH(path) \
CHUNK_FILE_COUNT(strlen(path) + 1)
#define CHUNK_DATA_COUNT(size) \
CHUNK_SIZE_COUNT((size) + sizeof (struct enc_chunk_data))
/* Min margin to write stream split headers without overwrap risk */
#define ENCBUF_MIN_SPLIT_MARGIN \
(2*(1 + CHUNK_FILE_COUNT(MAX_PATH)) - 1)
static void *rec_buffer; /* Root rec buffer pointer */
static size_t rec_buffer_size; /* Root rec buffer size */
static void *pcm_buffer; /* Circular buffer for PCM samples */
static volatile bool pcm_pause; /* Freeze DMA write position */
static volatile size_t pcm_widx; /* Current DMA write position */
static volatile size_t pcm_ridx; /* Current PCM read position */
static union enc_chunk_hdr *enc_buffer; /* Circular encoding buffer */
static size_t enc_widx; /* Encoder chunk write index */
static size_t enc_ridx; /* Encoder chunk read index */
static size_t enc_buflen; /* Length of buffer in slots */
static unsigned char *stream_buffer; /* Stream-to-disk write buffer */
static ssize_t stream_buf_used; /* Stream write buffer occupancy */
static struct enc_chunk_file *fname_buf;/* Buffer with next file to create */
static unsigned long enc_sample_rate; /* Samplerate used by encoder */
static bool pcm_buffer_empty; /* All PCM chunks processed? */
static typeof (memcpy) *pcm_copyfn; /* PCM memcpy or copy_buffer_mono */
static enc_callback_t enc_cb; /* Encoder's recording callback */
/** File flushing **/
static unsigned long encbuf_datarate; /* Rate of data per second */
#if (CONFIG_STORAGE & STORAGE_ATA)
static int spinup_time; /* Last spinup time */
#endif
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 high_watermark; /* Max limit for data flush */
#ifdef HAVE_PRIORITY_SCHEDULING
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 flood_watermark; /* Max limit for thread prio boost */
static bool prio_boosted;
#endif
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
/** Stream marking **/
enum mark_stream_action
{
MARK_STREAM_END = 0x1, /* Mark end current stream */
MARK_STREAM_START = 0x2, /* Mark start of new stream */
MARK_STREAM_SPLIT = 0x3, /* Insert split; orr of above values */
MARK_STREAM_PRE = 0x4, /* Do prerecord data tally */
MARK_STREAM_START_PRE = MARK_STREAM_PRE | MARK_STREAM_START,
};
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
/***************************************************************************/
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
/* Buffer pointer (p) to PCM sample memory address */
static inline void * pcmbuf_ptr(size_t p)
{
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 pcm_buffer + p;
}
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
/* Buffer pointer (p) plus value (v), wrapped if necessary */
static size_t pcmbuf_add(size_t p, size_t v)
{
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 res = p + v;
if (res >= PCM_BUF_SIZE)
res -= PCM_BUF_SIZE;
return res;
}
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 of data in PCM buffer */
size_t pcmbuf_used(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
size_t p1 = pcm_ridx;
size_t p2 = pcm_widx;
if (p1 > p2)
p2 += PCM_BUF_SIZE;
return p2 - p1;
}
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
/* Buffer pointer (p) to memory address of header */
static inline union enc_chunk_hdr * encbuf_ptr(size_t p)
{
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 enc_buffer + p;
}
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
/* Buffer pointer (p) plus value (v), wrapped if necessary */
static size_t encbuf_add(size_t p, size_t v)
{
size_t res = p + v;
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 (res >= enc_buflen)
res -= enc_buflen;
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 res;
}
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
/* Number of free buffer slots */
static size_t encbuf_free(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
size_t p1 = enc_ridx;
size_t p2 = enc_widx;
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 (p2 >= p1)
p1 += enc_buflen;
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 p1 - p2;
}
/* Number of used buffer slots */
static size_t encbuf_used(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
size_t p1 = enc_ridx;
size_t p2 = enc_widx;
if (p1 > p2)
p2 += enc_buflen;
return p2 - p1;
}
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
/* Is the encoder buffer empty? */
static bool encbuf_empty(void)
{
return enc_ridx == enc_widx;
}
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
/* Buffer pointer (p) plus size (v), written to enc_widx, new widx
* zero-initialized */
static void encbuf_widx_advance(size_t widx, size_t v)
{
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
widx = encbuf_add(widx, v);
encbuf_ptr(widx)->zero = 0;
enc_widx = widx;
}
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
/* Buffer pointer (p) plus size of chunk at (p), wrapped to (0) if
* necessary.
*
* pout points to variable to receive increment result
*
* Returns NULL if it was a wrap marker */
static void * encbuf_read_ptr_incr(size_t p, size_t *pout)
{
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
union enc_chunk_hdr *hdr = encbuf_ptr(p);
size_t v;
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 (hdr->type)
{
case CHUNK_T_DATA:
v = CHUNK_DATA_COUNT(hdr->size);
break;
case CHUNK_T_STREAM_START:
v = hdr->size;
break;
case CHUNK_T_STREAM_END:
default:
v = 1;
break;
case CHUNK_T_WRAP:
/* Wrap markers are not returned but caller may have to know that
the index was changed since it impacts available space */
*pout = 0;
return 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
*pout = encbuf_add(p, v);
return hdr;
}
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
/* Buffer pointer (p) of contiguous free space (v), wrapped to (0) if
* necessary.
*
* pout points to variable to receive possible-adjusted p
*
* Returns header at (p) or wrapped header at (0) if wrap was
* required in order to provide contiguous space. Header is zero-
* initialized.
*
* Marks the wrap point if a wrap is required to make the allocation. */
static void * encbuf_get_write_ptr(size_t p, size_t v, size_t *pout)
{
union enc_chunk_hdr *hdr = encbuf_ptr(p);
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 (p + v > enc_buflen)
{
hdr->type = CHUNK_T_WRAP; /* All other fields ignored */
p = 0;
hdr = encbuf_ptr(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
*pout = p;
hdr->zero = 0;
return hdr;
}
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
/* Post a flush request to audio thread, if none is currently queued */
static void encbuf_request_flush(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
if (!queue_peek_ex(&audio_queue, NULL, 0,
&(const long [2]){ Q_AUDIO_RECORD_FLUSH,
Q_AUDIO_RECORD_FLUSH }))
queue_post(&audio_queue, Q_AUDIO_RECORD_FLUSH, 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
/* Set the error bits in (e): no lock */
static inline void set_error_bits(uint32_t e)
{
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
errors |= e;
rec_errors |= e;
}
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
/* Clear the error bits in (e): no lock */
static inline void clear_error_bits(uint32_t e)
{
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
errors &= ~e;
}
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
/* Set the error bits in (e) */
static void raise_error_status(uint32_t e)
{
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
pcm_rec_lock();
set_error_bits(e);
pcm_rec_unlock();
}
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
/* Clear the error bits in (e) */
static void clear_error_status(uint32_t e)
{
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
pcm_rec_lock();
clear_error_bits(e);
pcm_rec_unlock();
}
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
/* Set the warning bits in (w): no lock */
static inline void set_warning_bits(uint32_t w)
{
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
warnings |= w;
}
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
/* Clear the warning bits in (w): no lock */
static inline void clear_warning_bits(uint32_t w)
{
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
warnings &= ~w;
}
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
/* Set the warning bits in (w) */
static void raise_warning_status(uint32_t w)
{
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
pcm_rec_lock();
set_warning_bits(w);
pcm_rec_unlock();
}
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
/* Clear the warning bits in (w) */
static void clear_warning_status(uint32_t w)
{
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
pcm_rec_lock();
clear_warning_bits(w);
pcm_rec_unlock();
}
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
/* Callback for when more data is ready - called by DMA ISR */
static void pcm_rec_have_more(void **start, 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
size_t next_idx = pcm_widx;
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 (!pcm_pause)
{
/* One empty chunk must remain after widx is advanced */
if (pcmbuf_used() <= PCM_BUF_SIZE - 2*PCM_CHUNK_SIZE)
next_idx = pcmbuf_add(next_idx, PCM_CHUNK_SIZE);
else
set_warning_bits(PCMREC_W_PCM_BUFFER_OVF);
}
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
*start = pcmbuf_ptr(next_idx);
*size = PCM_CHUNK_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
pcm_widx = next_idx;
}
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 enum pcm_dma_status pcm_rec_status_callback(enum pcm_dma_status status)
{
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 (status < PCM_DMAST_OK)
{
/* Some error condition */
if (status == PCM_DMAST_ERR_DMA)
{
set_error_bits(PCMREC_E_DMA);
return status;
}
else
{
/* Try again next transmission - frame is invalid */
set_warning_bits(PCMREC_W_DMA);
}
}
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 PCM_DMAST_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
/* Start DMA transfer */
static void pcm_start_recording(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
pcm_record_data(pcm_rec_have_more, pcm_rec_status_callback,
pcmbuf_ptr(pcm_widx), PCM_CHUNK_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
/* Initialize the various recording buffers */
static void init_rec_buffers(void)
{
/* Layout of recording buffer: |PCMBUF|STREAMBUF|FILENAME|ENCBUF| */
void *buf = rec_buffer;
size_t size = rec_buffer_size;
/* PCMBUF */
pcm_buffer = CACHEALIGN_UP(buf); /* Line align */
size -= pcm_buffer + PCM_BUF_SIZE - buf;
buf = pcm_buffer + PCM_BUF_SIZE;
/* STREAMBUF */
stream_buffer = buf; /* Also line-aligned */
buf += STREAM_BUF_SIZE;
size -= STREAM_BUF_SIZE;
/* FILENAME */
fname_buf = buf;
buf += CHUNK_FILE_COUNT(MAX_PATH)*ENC_HDR_SIZE;
size -= CHUNK_FILE_COUNT(MAX_PATH)*ENC_HDR_SIZE;
fname_buf->hdr.zero = 0;
/* ENCBUF */
enc_buffer = buf;
enc_buflen = size;
ALIGN_BUFFER(enc_buffer, enc_buflen, ENC_HDR_SIZE);
enc_buflen = CHUNK_SIZE_COUNT(enc_buflen);
}
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
/* Reset the circular buffers */
static void reset_fifos(bool hard)
{
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
/* PCM FIFO */
pcm_pause = true;
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 (hard)
pcm_widx = 0; /* Don't just empty but reset it */
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
pcm_ridx = pcm_widx;
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
/* Encoder FIFO */
encbuf_widx_advance(0, 0);
enc_ridx = 0;
/* No overflow-related warnings now */
clear_warning_status(PCMREC_W_PCM_BUFFER_OVF | PCMREC_W_ENC_BUFFER_OVF);
}
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
/* Initialize file statistics */
static void reset_rec_stats(void)
{
num_rec_bytes = 0;
num_rec_samples = 0;
encbuf_rec_count = 0;
clear_warning_status(PCMREC_W_FILE_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
/* Boost or unboost recording threads' priorities */
static void do_prio_boost(bool boost)
{
#ifdef HAVE_PRIORITY_SCHEDULING
prio_boosted = boost;
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
int prio = PRIORITY_RECORDING;
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 (boost)
prio -= 4;
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
codec_thread_set_priority(prio);
thread_set_priority(audio_thread_id, prio);
#endif
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
(void)boost;
}
/* Reset all relevant state */
static void init_state(void)
{
reset_fifos(true);
reset_rec_stats();
do_prio_boost(false);
cancel_cpu_boost();
record_state = REC_STATE_IDLE;
record_status = RECORD_STOPPED;
}
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
/* Set hardware samplerate and save it */
static void update_samplerate_config(unsigned long sampr)
{
/* PCM samplerate is either the same as the setting or the nearest
one hardware supports if using S/PDIF */
unsigned long pcm_sampr = sampr;
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 HAVE_SPDIF_IN
if (rec_source == AUDIO_SRC_SPDIF)
{
int index = round_value_to_list32(sampr, hw_freq_sampr,
HW_NUM_FREQ, false);
pcm_sampr = hw_freq_sampr[index];
}
#endif /* HAVE_SPDIF_IN */
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
pcm_set_frequency(pcm_sampr | SAMPR_TYPE_REC);
sample_rate = sampr;
}
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
/* Calculate the average data rate */
static unsigned long get_encbuf_datarate(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
/* If not yet calculable, start with uncompressed PCM byterate */
if (num_rec_samples && sample_rate && encbuf_rec_count)
{
return (encbuf_rec_count*sample_rate + num_rec_samples - 1)
/ num_rec_samples;
}
else
{
return CHUNK_SIZE_COUNT(sample_rate*num_channels*PCM_DEPTH_BYTES);
}
}
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
/* Returns true if the watermarks should be updated due to data rate
change */
static bool monitor_encbuf_datarate(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
unsigned long rate = get_encbuf_datarate();
long diff = rate - encbuf_datarate;
/* Off by more than 1/2 FLUSH_MON_INTERVAL? */
return 2*(unsigned long)abs(diff) > encbuf_datarate*FLUSH_MON_INTERVAL;
}
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
/* Get adjusted spinup time */
static int get_spinup_time(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
int spin = storage_spinup_time();
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 (CONFIG_STORAGE & STORAGE_ATA)
/* Write at FLUSH_SECONDS + st remaining in enc_buffer - range fs+2s to
fs+10s total - default to 3.5s spinup. */
if (spin == 0)
spin = 35*HZ/10; /* default - cozy */
else if (spin < 2*HZ)
spin = 2*HZ; /* ludicrous - ramdisk? */
else if (spin > 10*HZ)
spin = 10*HZ; /* do you have a functioning HD? */
#endif /* (CONFIG_STORAGE & STORAGE_ATA) */
return spin;
}
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
/* Returns true if the watermarks should be updated due to spinup time
change */
static inline bool monitor_spinup_time(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
#if (CONFIG_STORAGE & STORAGE_ATA)
return get_spinup_time() != spinup_time;
#else
return false;
#endif
}
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
/* Update buffer watermarks with spinup time compensation */
static void refresh_watermarks(void)
{
int spin = get_spinup_time();
#if (CONFIG_STORAGE & STORAGE_ATA)
logf("ata spinup: %d", spin);
spinup_time = spin;
#endif
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
unsigned long rate = get_encbuf_datarate();
logf("byterate: %lu", rate * ENC_HDR_SIZE);
encbuf_datarate = 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
/* Try to start writing with FLUSH_SECONDS remaining after disk spinup */
high_watermark = (uint64_t)rate*(FLUSH_SECONDS*HZ + spin) / HZ;
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 (high_watermark > enc_buflen)
high_watermark = enc_buflen;
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
high_watermark = enc_buflen - high_watermark;
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
logf("high wm: %lu", (unsigned long)high_watermark);
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 HAVE_PRIORITY_SCHEDULING
/* Boost thread priority if enough ground is lost since flushing started
or is taking an unreasonably long time */
flood_watermark = rate*PANIC_SECONDS;
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 (flood_watermark > enc_buflen)
flood_watermark = enc_buflen;
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
flood_watermark = enc_buflen - flood_watermark;
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
logf("flood wm: %lu", (unsigned long)flood_watermark);
#endif /* HAVE_PRIORITY_SCHEDULING */
}
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
/* Tell encoder the stream parameters and get information back */
static bool configure_encoder_stream(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
struct enc_inputs inputs;
inputs.sample_rate = sample_rate;
inputs.num_channels = num_channels;
inputs.config = &enc_config;
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
/* encoder can change these - init with defaults */
inputs.enc_sample_rate = 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
if (enc_cb(ENC_CB_INPUTS, &inputs) < 0)
{
raise_error_status(PCMREC_E_ENC_SETUP);
return false;
}
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
enc_sample_rate = inputs.enc_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
if (enc_sample_rate != 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
/* Codec doesn't want to/can't use the setting and has chosen a
different sample rate */
raise_warning_status(PCMREC_W_SAMPR_MISMATCH);
logf("enc sampr:%lu", enc_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
else
{
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
clear_warning_status(PCMREC_W_SAMPR_MISMATCH);
}
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
refresh_watermarks();
return true;
}
#ifdef HAVE_SPDIF_IN
/* Return the S/PDIF sample rate closest to a value in the master list */
static unsigned long get_spdif_samplerate(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
unsigned long sr = spdif_measure_frequency();
int index = round_value_to_list32(sr, audio_master_sampr_list,
SAMPR_NUM_FREQ, false);
return audio_master_sampr_list[index];
}
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
/* Check the S/PDIF rate and compare to current setting. Apply the new
* rate if it changed. */
static void check_spdif_samplerate(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
unsigned long sampr = get_spdif_samplerate();
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 (sampr == sample_rate)
return;
codec_stop();
pcm_stop_recording();
reset_fifos(true);
reset_rec_stats();
update_samplerate_config(sampr);
pcm_apply_settings();
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 (!configure_encoder_stream() || rec_errors)
return;
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
pcm_start_recording();
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 (record_status == RECORD_PRERECORDING)
{
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
codec_go();
pcm_pause = false;
}
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
}
#endif /* HAVE_SPDIF_IN */
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
/* Discard the stream buffer contents */
static inline void stream_discard_buf(void)
{
stream_buf_used = 0;
}
/* Flush stream buffer to disk */
static bool stream_flush_buf(void)
{
if (stream_buf_used == 0)
return true;
ssize_t rc = write(rec_fd, stream_buffer, stream_buf_used);
if (LIKELY(rc == stream_buf_used))
{
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
stream_discard_buf();
return true;
}
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 (rc > 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
/* Some was written; keep in sync */
stream_buf_used -= rc;
memmove(stream_buffer, stream_buffer + rc, stream_buf_used);
}
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 false;
}
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
/* Close the output file */
static void close_rec_file(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
if (rec_fd < 0)
return;
bool ok = stream_flush_buf();
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 (close(rec_fd) != 0 || !ok)
raise_error_status(PCMREC_E_IO);
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
rec_fd = -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
/* Creates or opens the current path */
static bool open_rec_file(bool create)
{
if (rec_fd >= 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
/* Any previous file should have been closed */
logf("open file: file already open");
close_rec_file();
}
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
stream_discard_buf();
int oflags = create ? O_CREAT|O_TRUNC : 0;
rec_fd = open(fname_buf->path, O_RDWR|oflags, 0666);
if (rec_fd < 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
raise_error_status(PCMREC_E_IO);
return false;
}
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 true;
}
/* Copy with mono conversion - output 1/2 size of input */
static void * ICODE_ATTR
copy_buffer_mono_lr(void *dst, const void *src, size_t src_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
int16_t *d = dst;
int16_t const *s = src;
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
/* mono = (L + R) / 2 */
do
*d++ = ((int32_t){ *s++ } + *s++ + 1) >> 1;
while (src_size -= PCM_SAMP_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 dst;
}
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
/* Copy with mono conversion - output 1/2 size of input */
static void * ICODE_ATTR
copy_buffer_mono_l(void *dst, const void *src, size_t src_size)
{
int16_t *d = dst;
int16_t const *s = (int16_t *)src - 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
/* mono = L */
do
*d++ = *(s += 2);
while (src_size -= PCM_SAMP_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 dst;
}
/* Copy with mono conversion - output 1/2 size of input */
static void * ICODE_ATTR
copy_buffer_mono_r(void *dst, const void *src, size_t src_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
int16_t *d = dst;
int16_t const *s = (int16_t *)src - 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
/* mono = R */
do
*d++ = *(s += 2);
while (src_size -= PCM_SAMP_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 dst;
}
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
/** pcm_rec_* group **/
/* Clear all errors and warnings */
void pcm_rec_error_clear(void)
{
clear_error_status(PCMREC_E_ALL);
clear_warning_status(PCMREC_W_ALL);
}
/* Check mode, errors and warnings */
unsigned int pcm_rec_status(void)
{
unsigned int ret = record_status;
if (errors)
ret |= AUDIO_STATUS_ERROR;
if (warnings)
ret |= AUDIO_STATUS_WARNING;
return ret;
}
/* Return warnings that have occured since recording started */
uint32_t pcm_rec_get_warnings(void)
{
return warnings;
}
#ifdef HAVE_SPDIF_IN
/* Return the currently-configured sample rate */
unsigned long pcm_rec_sample_rate(void)
{
return sample_rate;
}
#endif
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
/** audio_* group **/
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
/* Initializes recording - call before calling any other recording function */
void audio_init_recording(void)
{
LOGFQUEUE("audio >| pcmrec Q_AUDIO_INIT_RECORDING");
audio_queue_send(Q_AUDIO_INIT_RECORDING, 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
/* Closes recording - call audio_stop_recording first or risk data loss */
void audio_close_recording(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
LOGFQUEUE("audio >| pcmrec Q_AUDIO_CLOSE_RECORDING");
audio_queue_send(Q_AUDIO_CLOSE_RECORDING, 0);
}
/* Sets recording parameters */
void audio_set_recording_options(struct audio_recording_options *options)
{
LOGFQUEUE("audio >| pcmrec Q_AUDIO_RECORDING_OPTIONS");
audio_queue_send(Q_AUDIO_RECORDING_OPTIONS, (intptr_t)options);
}
/* Start recording if not recording or else split */
void audio_record(const char *filename)
{
LOGFQUEUE("audio >| pcmrec Q_AUDIO_RECORD: %s", filename);
audio_queue_send(Q_AUDIO_RECORD, (intptr_t)filename);
}
/* audio_record alias for API compatibility with HW codec */
void audio_new_file(const char *filename)
__attribute__((alias("audio_record")));
/* Stop current recording if recording */
void audio_stop_recording(void)
{
LOGFQUEUE("audio > pcmrec Q_AUDIO_RECORD_STOP");
audio_queue_post(Q_AUDIO_RECORD_STOP, 0);
}
/* Pause current recording */
void audio_pause_recording(void)
{
LOGFQUEUE("audio > pcmrec Q_AUDIO_RECORD_PAUSE");
audio_queue_post(Q_AUDIO_RECORD_PAUSE, 0);
}
/* Resume current recording if paused */
void audio_resume_recording(void)
{
LOGFQUEUE("audio > pcmrec Q_AUDIO_RECORD_RESUME");
audio_queue_post(Q_AUDIO_RECORD_RESUME, 0);
}
/* Set the input source gain. For mono sources, only left gain is used */
void audio_set_recording_gain(int left, int right, int type)
{
#if 0
logf("pcmrec: t=%d l=%d r=%d", type, left, right);
#endif
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
audiohw_set_recvol(left, right, type);
}
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
/** Information about current state **/
/* Return sample clock in HZ */
static unsigned long get_samples_time(void)
{
if (enc_sample_rate == 0)
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
return (unsigned long)(HZ*num_rec_samples / enc_sample_rate);
}
/* Return current prerecorded time in ticks (playback equivalent time) */
unsigned long audio_prerecorded_time(void)
{
if (record_status != RECORD_PRERECORDING)
return 0;
unsigned long t = get_samples_time();
return MIN(t, pre_record_seconds*HZ);
}
/* Return current recorded time in ticks (playback equivalent time) */
unsigned long audio_recorded_time(void)
{
if (record_state == REC_STATE_IDLE)
return 0;
return get_samples_time();
}
/* Return number of bytes encoded to output */
unsigned long audio_num_recorded_bytes(void)
{
if (record_state == REC_STATE_IDLE)
return 0;
return num_rec_bytes;
}
/** Data Flushing **/
/* Stream start chunk with path was encountered */
static void flush_stream_start(struct enc_chunk_file *file)
{
/* Save filename; don't open file here which avoids creating files
with no audio content. Splitting while paused can create those
in large numbers. */
fname_buf->hdr = file->hdr;
/* Correct size if this was wrap-padded */
fname_buf->hdr.size = CHUNK_FILE_COUNT(
strlcpy(fname_buf->path, file->path, MAX_PATH) + 1);
}
/* Data chunk was encountered */
static bool flush_stream_data(struct enc_chunk_data *data)
{
if (fname_buf->hdr.zero)
{
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 data chunk; create the file */
if (open_rec_file(true))
{
/* Inherit some flags from initial data chunk */
fname_buf->hdr.err = data->hdr.err;
fname_buf->hdr.pre = data->hdr.pre;
fname_buf->hdr.aux0 = data->hdr.aux0;
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 (enc_cb(ENC_CB_STREAM, fname_buf) < 0)
raise_error_status(PCMREC_E_ENCODER_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
fname_buf->hdr.zero = 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
if (rec_errors)
return false;
}
if (rec_fd < 0)
return true; /* Just keep discarding */
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 (enc_cb(ENC_CB_STREAM, data) < 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
raise_error_status(PCMREC_E_ENCODER_STREAM);
return false;
}
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 true;
}
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
/* Stream end chunk was encountered */
static bool flush_stream_end(union enc_chunk_hdr *hdr)
{
if (rec_fd < 0)
return true;
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 (enc_cb(ENC_CB_STREAM, hdr) < 0)
{
raise_error_status(PCMREC_E_ENCODER_STREAM);
return false;
}
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
close_rec_file();
return true;
}
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
/* Discard remainder of stream in encoder buffer */
static void discard_stream(void)
{
/* Discard everything up until the next non-data chunk */
while (!encbuf_empty())
{
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 ridx;
union enc_chunk_hdr *hdr = encbuf_read_ptr_incr(enc_ridx, &ridx);
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 && hdr->type != CHUNK_T_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
if (hdr->type != CHUNK_T_STREAM_START)
enc_ridx = ridx;
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
enc_ridx = ridx;
}
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
/* Try to finish header by closing and reopening the file. A seek or
other operation will likely fail because buffers will need to be
flushed (here and in file code). That will likely fail but a close
will just close the fd and discard everything. We reopen with what
actually made it to disk. Modifying existing file contents will
more than likely succeed even on a full disk. The result might not
be entirely correct as far as the headers' sizes and counts unless
the codec can correct that but the sample format information
should be. */
if (rec_fd >= 0 && open_rec_file(false))
{
/* Synthesize a special end chunk here */
union enc_chunk_hdr end;
end.zero = 0;
end.err = 1; /* Codec should try to correct anything that's off */
end.type = CHUNK_T_STREAM_END;
if (!flush_stream_end(&end))
close_rec_file();
}
}
/* Flush a chunk to disk
*
* Transitions state from REC_STATE_MONITOR to REC_STATE_FLUSH when buffer
* is filling. 'margin' is fullness threshold that transitions to flush state.
*
* Call with REC_STATE_IDLE to indicate a forced flush which flushes buffer
* to less than 'margin'.
*/
static enum record_state flush_chunk(enum record_state state, size_t margin)
{
#ifdef HAVE_PRIORITY_SCHEDULING
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 unsigned long prio_tick; /* Timeout for auto boost */
#endif
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 used = encbuf_used();
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 (state)
{
case REC_STATE_MONITOR:
if (monitor_encbuf_datarate() || monitor_spinup_time())
refresh_watermarks();
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 (used < margin)
return REC_STATE_MONITOR;
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
state = REC_STATE_FLUSH;
trigger_cpu_boost();
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 HAVE_PRIORITY_SCHEDULING
prio_tick = current_tick + PRIO_SECONDS*HZ;
#if (CONFIG_STORAGE & STORAGE_ATA)
prio_tick += spinup_time;
#endif
#endif /* HAVE_PRIORITY_SCHEDULING */
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
/* Fall-through */
case REC_STATE_IDLE: /* As a hint for "forced" */
if (used < margin)
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
/* Fall-through */
case REC_STATE_FLUSH:
#ifdef HAVE_PRIORITY_SCHEDULING
if (!prio_boosted && state != REC_STATE_IDLE &&
(used >= flood_watermark || TIME_AFTER(current_tick, prio_tick)))
do_prio_boost(true);
#endif /* HAVE_PRIORITY_SCHEDULING */
while (used)
{
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
union enc_chunk_hdr *hdr = encbuf_ptr(enc_ridx);
size_t count = 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
switch (hdr->type)
{
case CHUNK_T_DATA:
if (flush_stream_data(ENC_DATA_HDR(hdr)))
count = CHUNK_DATA_COUNT(hdr->size);
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
case CHUNK_T_STREAM_START:
/* Doesn't do stream writes */
flush_stream_start(ENC_FILE_HDR(hdr));
count = hdr->size;
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
case CHUNK_T_STREAM_END:
if (flush_stream_end(hdr))
count = 1;
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
case CHUNK_T_WRAP:
enc_ridx = 0;
used = encbuf_used();
continue;
}
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 (count)
enc_ridx = encbuf_add(enc_ridx, count);
else
discard_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
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
if (!encbuf_empty())
return state;
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
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
if (encbuf_empty())
{
do_prio_boost(false);
cancel_cpu_boost();
}
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 REC_STATE_MONITOR;
}
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
/* Monitor buffer and finish stream, freeing-up space at the same time */
static void finish_stream(bool stopping)
{
size_t threshold = stopping ? 1 : enc_buflen - ENCBUF_MIN_SPLIT_MARGIN;
enum record_state state = REC_STATE_MONITOR;
size_t need = 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
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
switch (state)
{
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 REC_STATE_IDLE:
state = flush_chunk(state, threshold);
continue;
default:
if (!need)
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
if (!stopping || pcm_buffer_empty)
{
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
need = codec_finish_stream();
if (need)
{
need = 2*CHUNK_DATA_COUNT(need) - 1;
if (need >= enc_buflen)
{
need = 0;
codec_stop();
threshold = 1;
}
else if (threshold > enc_buflen - need)
{
threshold = enc_buflen - need;
}
}
}
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 (!need || encbuf_used() >= threshold)
state = REC_STATE_IDLE; /* Start flush */
else
sleep(HZ/10); /* Don't flood with pings */
continue;
}
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
break;
}
}
/* Start a new stream, transistion to a new one or end the current one */
static void mark_stream(const char *path, enum mark_stream_action action)
{
if (action & MARK_STREAM_END)
{
size_t widx;
union enc_chunk_hdr *hdr = encbuf_get_write_ptr(enc_widx, 1, &widx);
hdr->type = CHUNK_T_STREAM_END;
encbuf_widx_advance(widx, 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
if (action & MARK_STREAM_START)
{
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 count = CHUNK_FILE_COUNT_PATH(path);
struct enc_chunk_file *file;
size_t widx;
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 (action & MARK_STREAM_PRE)
{
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
/* Prerecord: START marker goes first or before existing data */
if (enc_ridx < count)
{
/* Adjust to occupy end of buffer and pad accordingly */
count += enc_ridx;
enc_ridx += enc_buflen;
}
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
enc_ridx -= count;
/* Won't adjust p since enc_ridx is already set as non-wrapping */
file = encbuf_get_write_ptr(enc_ridx, count, &widx);
}
else
{
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
/* The usual: START marker goes first or after existing data */
file = encbuf_get_write_ptr(enc_widx, count, &widx);
encbuf_widx_advance(widx, 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
file->hdr.type = CHUNK_T_STREAM_START;
file->hdr.size = count;
strlcpy(file->path, path, MAX_PATH);
}
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
}
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
/* Tally-up and keep the required amount of prerecord data.
* Updates record stats accordingly. */
static void tally_prerecord_data(void)
{
unsigned long count = 0;
size_t bytes = 0;
unsigned long samples = 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
/* Find out how much is there */
for (size_t idx = enc_ridx; idx != enc_widx;)
{
struct enc_chunk_data *data = encbuf_read_ptr_incr(idx, &idx);
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 (!data)
continue;
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
count += CHUNK_DATA_COUNT(data->hdr.size);
bytes += data->hdr.size;
samples += data->pcm_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
/* Have too much? Discard oldest data. */
unsigned long pre_samples = enc_sample_rate*pre_record_seconds;
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 (samples > pre_samples)
{
struct enc_chunk_data *data =
encbuf_read_ptr_incr(enc_ridx, &enc_ridx);
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 (!data)
continue;
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
count -= CHUNK_DATA_COUNT(data->hdr.size);
bytes -= data->hdr.size;
samples -= data->pcm_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
encbuf_rec_count = count;
num_rec_bytes = bytes;
num_rec_samples = samples;
}
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
/** Event handlers for recording thread **/
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
/* Q_AUDIO_INIT_RECORDING */
static void on_init_recording(void)
{
send_event(RECORDING_EVENT_START, 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
rec_buffer = audio_get_buffer(true, &rec_buffer_size);
init_rec_buffers();
init_state();
pcm_init_recording();
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
}
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
/* Q_AUDIO_CLOSE_RECORDING */
static void on_close_recording(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
/* Simply shut down the recording system. Whatever wasn't saved is
lost. */
codec_unload();
pcm_close_recording();
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
close_rec_file();
init_state();
rec_errors = 0;
pcm_rec_error_clear();
/* Reset PCM to defaults */
pcm_set_frequency(HW_SAMPR_RESET | SAMPR_TYPE_REC);
audio_set_output_source(AUDIO_SRC_PLAYBACK);
pcm_apply_settings();
send_event(RECORDING_EVENT_STOP, 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
}
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
/* Q_AUDIO_RECORDING_OPTIONS */
static void on_recording_options(struct audio_recording_options *options)
{
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 (!options)
{
logf("options: option NULL!");
return;
}
if (record_state != REC_STATE_IDLE)
{
/* This would ruin things */
logf("options: still recording!");
return;
}
/* Stop everything else that might be running */
pcm_stop_recording();
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
int afmt = rec_format_afmt[options->enc_config.rec_format];
bool enc_load = true;
if (codec_loaded() != AFMT_UNKNOWN)
{
if (get_audio_base_codec_type(enc_config.afmt) !=
get_audio_base_codec_type(afmt))
{
/* New format, new encoder; unload this one */
codec_unload();
}
else
{
/* Keep current encoder */
codec_stop();
enc_load = false;
}
}
init_state();
/* Read recording options, remember the ones used elsewhere */
unsigned frequency = options->rec_frequency;
rec_source = options->rec_source;
num_channels = options->rec_channels == 1 ? 1 : 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
unsigned mono_mode = options->rec_mono_mode;
pre_record_seconds = options->rec_prerecord_time;
enc_config = options->enc_config;
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
enc_config.afmt = afmt;
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
queue_reply(&audio_queue, 0); /* Let caller go */
/* Pick appropriate PCM copy routine */
pcm_copyfn = memcpy;
if (num_channels == 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
static typeof (memcpy) * const copy_buffer_mono[] =
{
copy_buffer_mono_lr,
copy_buffer_mono_l,
copy_buffer_mono_r
};
if (mono_mode >= ARRAYLEN(copy_buffer_mono))
mono_mode = 0;
pcm_copyfn = copy_buffer_mono[mono_mode];
}
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
/* Get the hardware samplerate to be used */
unsigned long sampr;
#ifdef HAVE_SPDIF_IN
if (rec_source == AUDIO_SRC_SPDIF)
sampr = get_spdif_samplerate(); /* Determined by source */
else
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
#endif /* HAVE_SPDIF_IN */
sampr = rec_freq_sampr[frequency];
update_samplerate_config(sampr);
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
/* Set monitoring */
audio_set_output_source(rec_source);
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
/* Apply hardware setting to start monitoring now */
pcm_apply_settings();
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 (!enc_load || codec_load(-1, afmt | CODEC_TYPE_ENCODER))
{
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
enc_cb = codec_get_enc_callback();
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 (!enc_cb || !configure_encoder_stream())
{
codec_unload();
return;
}
if (pre_record_seconds != 0)
{
record_status = RECORD_PRERECORDING;
codec_go();
pcm_pause = false;
}
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
pcm_start_recording();
}
else
{
logf("set rec opt: enc load failed");
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
raise_error_status(PCMREC_E_LOAD_ENCODER);
}
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
}
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
/* Q_AUDIO_RECORD - start recording (not gapless)
or split stream (gapless) */
static void on_record(const char *filename)
{
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 (rec_errors)
{
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
logf("on_record: errors not cleared");
return;
}
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 (!filename)
{
logf("on_record: No filename");
return;
}
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 (codec_loaded() == AFMT_UNKNOWN)
{
logf("on_record: Recording options not set");
return;
}
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
logf("on_record: new file '%s'", filename);
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
/* Copy path and let caller go */
char path[MAX_PATH];
strlcpy(path, filename, MAX_PATH);
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
queue_reply(&audio_queue, 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
enum mark_stream_action mark_action;
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 (record_state == REC_STATE_IDLE)
{
mark_action = MARK_STREAM_START;
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 (pre_record_seconds)
{
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
codec_pause();
tally_prerecord_data();
mark_action = MARK_STREAM_START_PRE;
}
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
clear_warning_status(PCMREC_W_ALL &
~(PCMREC_W_SAMPR_MISMATCH|PCMREC_W_DMA));
record_state = REC_STATE_MONITOR;
record_status = RECORD_RECORDING;
}
else
{
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
/* Already recording, just split the stream */
logf("inserting split");
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
mark_action = MARK_STREAM_SPLIT;
finish_stream(false);
reset_rec_stats();
}
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 (rec_errors)
{
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
pcm_pause = true;
codec_stop();
reset_fifos(false);
return;
}
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
mark_stream(path, mark_action);
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
codec_go();
pcm_pause = record_status != RECORD_RECORDING;
}
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
/* Q_AUDIO_RECORD_STOP */
static void on_record_stop(void)
{
if (record_state == REC_STATE_IDLE)
return;
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
/* Drain encoder and PCM buffers */
pcm_pause = true;
finish_stream(true);
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
/* End stream at last data and flush end marker */
mark_stream(NULL, MARK_STREAM_END);
while (flush_chunk(REC_STATE_IDLE, 1) == REC_STATE_IDLE);
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
reset_fifos(false);
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
bool prerecord = pre_record_seconds != 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
if (rec_errors)
{
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
codec_stop();
prerecord = false;
}
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
close_rec_file();
rec_errors = 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
record_state = REC_STATE_IDLE;
record_status = prerecord ? RECORD_PRERECORDING : RECORD_STOPPED;
reset_rec_stats();
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 (prerecord)
{
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
codec_go();
pcm_pause = false;
}
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
}
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
/* Q_AUDIO_RECORD_PAUSE */
static void on_record_pause(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
if (record_status != RECORD_RECORDING)
return;
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
pcm_pause = true;
record_status = RECORD_PAUSED;
}
/* Q_AUDIO_RECORD_RESUME */
static void on_record_resume(void)
{
if (record_status != RECORD_PAUSED)
return;
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
record_status = RECORD_RECORDING;
pcm_pause = !!rec_errors;
}
/* Called by audio thread when recording is initialized */
void audio_recording_handler(struct queue_event *ev)
{
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 HAVE_PRIORITY_SCHEDULING
/* Get current priorities since they get changed */
int old_prio = thread_get_priority(audio_thread_id);
int old_cod_prio = codec_thread_get_priority();
#endif
LOGFQUEUE("record < Q_AUDIO_INIT_RECORDING");
on_init_recording();
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
int watermark = high_watermark;
switch (ev->id)
{
case Q_AUDIO_CLOSE_RECORDING:
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
LOGFQUEUE("record < Q_AUDIO_CLOSE_RECORDING");
goto recording_done;
case Q_AUDIO_RECORDING_OPTIONS:
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
LOGFQUEUE("record < Q_AUDIO_RECORDING_OPTIONS");
on_recording_options((struct audio_recording_options *)ev->data);
break;
case Q_AUDIO_RECORD:
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
LOGFQUEUE("record < Q_AUDIO_RECORD: %s", (const char *)ev->data);
on_record((const char *)ev->data);
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
case Q_AUDIO_RECORD_STOP:
LOGFQUEUE("record < Q_AUDIO_RECORD_STOP");
on_record_stop();
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
case Q_AUDIO_RECORD_PAUSE:
LOGFQUEUE("record < Q_AUDIO_RECORD_PAUSE");
on_record_pause();
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
case Q_AUDIO_RECORD_RESUME:
LOGFQUEUE("record < Q_AUDIO_RECORD_RESUME");
on_record_resume();
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
case Q_AUDIO_RECORD_FLUSH:
watermark = 1;
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
case SYS_USB_CONNECTED:
LOGFQUEUE("record < SYS_USB_CONNECTED");
if (record_state != REC_STATE_IDLE)
{
LOGFQUEUE(" still recording");
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
goto recording_done;
} /* switch */
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
int timeout;
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 (record_state)
{
case REC_STATE_FLUSH:
case REC_STATE_MONITOR:
do
record_state = flush_chunk(record_state, watermark);
while (record_state == REC_STATE_FLUSH &&
queue_empty(&audio_queue));
timeout = record_state == REC_STATE_FLUSH ?
HZ*0 : HZ*FLUSH_MON_INTERVAL;
break;
case REC_STATE_IDLE:
#ifdef HAVE_SPDIF_IN
if (rec_source == AUDIO_SRC_SPDIF)
{
check_spdif_samplerate();
timeout = HZ/2;
break;
}
#endif /* HAVE_SPDIF_IN */
default:
timeout = TIMEOUT_BLOCK;
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
queue_wait_w_tmo(&audio_queue, ev, timeout);
} /* while */
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
recording_done:
on_close_recording();
#ifdef HAVE_PRIORITY_SCHEDULING
/* Restore normal thread priorities */
thread_set_priority(audio_thread_id, old_prio);
codec_thread_set_priority(old_cod_prio);
#endif
}
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
/** Encoder callbacks **/
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
/* Read a block of unprocessed PCM data, with mono conversion if
* num_channels == 1
*
* NOTE: Request must be less than the PCM buffer length in samples in order
* to progress.
* (ie. count <= PCM_NUM_CHUNKS*PCM_CHUNK_SAMP)
*/
static int enc_pcmbuf_read(void *buffer, int count)
{
size_t avail = pcmbuf_used();
size_t size = count*PCM_SAMP_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 (count > 0 && avail >= size)
{
size_t endidx = pcm_ridx + 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 (endidx > PCM_BUF_SIZE)
{
size_t wrap = endidx - PCM_BUF_SIZE;
size_t offset = size -= wrap;
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 (num_channels == 1)
offset /= 2; /* src offset -> dst offset */
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
pcm_copyfn(buffer + offset, pcmbuf_ptr(0), wrap);
}
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
pcm_copyfn(buffer, pcmbuf_ptr(pcm_ridx), 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 (avail >= sample_rate*PCM_SAMP_SIZE*PCM_BOOST_SECONDS ||
avail >= PCM_BUF_SIZE*1/2)
{
/* Filling up - boost threshold data available or more or 1/2 full
or more - boost codec */
trigger_cpu_boost();
}
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
pcm_buffer_empty = false;
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 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
/* Not enough data available - encoder should idle */
pcm_buffer_empty = true;
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
cancel_cpu_boost();
/* Sleep a little bit */
sleep(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
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
/* Advance PCM buffer by count samples */
static int enc_pcmbuf_advance(int 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
if (count <= 0)
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
size_t avail = pcmbuf_used();
size_t size = count*PCM_SAMP_SIZE;
if (avail < 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
size = avail;
count = size / PCM_SAMP_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
pcm_ridx = pcmbuf_add(pcm_ridx, 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 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
/* Return encoder chunk at current write position, wrapping to 0 if
* requested size demands it.
*
* NOTE: No request should be more than 1/2 the buffer length, all elements
* included, or progress will not be guaranteed.
* (ie. CHUNK_DATA_COUNT(need) <= enc_buflen / 2)
*/
static struct enc_chunk_data * enc_encbuf_get_buffer(size_t need)
{
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
/* Convert to buffer slot count, including the header */
need = CHUNK_DATA_COUNT(need);
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 record_state state = record_state;
size_t avail = encbuf_free();
/* Must have the split margin as well but it does not have to be
continuous with the request */
while (avail <= need + ENCBUF_MIN_SPLIT_MARGIN ||
(enc_widx + need > enc_buflen &&
enc_ridx <= need + ENCBUF_MIN_SPLIT_MARGIN))
{
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 (UNLIKELY(state == REC_STATE_IDLE))
{
/* Prerecording - delete some old data */
size_t ridx;
struct enc_chunk_data *data =
encbuf_read_ptr_incr(enc_ridx, &ridx);
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 (data)
{
encbuf_rec_count -= CHUNK_DATA_COUNT(data->hdr.size);
num_rec_bytes -= data->hdr.size;
num_rec_samples -= data->pcm_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
enc_ridx = ridx;
avail = encbuf_free();
continue;
}
else if (avail == enc_buflen)
{
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
/* Empty but request larger than any possible space */
raise_warning_status(PCMREC_W_ENC_BUFFER_OVF);
}
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
else if (state != REC_STATE_FLUSH && encbuf_used() < high_watermark)
{
/* Not yet even at high watermark but what's needed won't fit */
encbuf_request_flush();
}
sleep(0);
return 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
struct enc_chunk_data *data =
encbuf_get_write_ptr(enc_widx, need, &enc_widx);
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 (state == REC_STATE_IDLE)
data->hdr.pre = 1;
return data;
}
/* Releases the current buffer into the available chunks */
static void enc_encbuf_finish_buffer(void)
{
struct enc_chunk_data *data = ENC_DATA_HDR(encbuf_ptr(enc_widx));
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 (data->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
/* Encoder set error flag */
raise_error_status(PCMREC_E_ENCODER);
return;
}
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 data_size = data->hdr.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 (data_size == 0)
return; /* Claims nothing was written */
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 count = CHUNK_DATA_COUNT(data_size);
size_t avail = encbuf_free();
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 (avail <= count || enc_widx + count > enc_buflen)
{
/* Claims it wrote too much? */
raise_warning_status(PCMREC_W_ENC_BUFFER_OVF);
return;
}
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 (num_rec_bytes + data_size > MAX_NUM_REC_BYTES)
{
/* Would exceed filesize limit; should have split sooner.
This chunk will be dropped. :'( */
raise_warning_status(PCMREC_W_FILE_SIZE);
return;
}
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
encbuf_widx_advance(enc_widx, 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
encbuf_rec_count += count;
num_rec_bytes += data_size;
num_rec_samples += data->pcm_count;
}
/* Read from the output stream */
static ssize_t enc_stream_read(void *buf, size_t count)
{
if (!stream_flush_buf())
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 read(rec_fd, buf, 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
/* Seek the output steam */
static off_t enc_stream_lseek(off_t offset, int whence)
{
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 (!stream_flush_buf())
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 lseek(rec_fd, offset, whence);
}
/* Write to the output stream */
static ssize_t enc_stream_write(const void *buf, size_t count)
{
if (UNLIKELY(count >= STREAM_BUF_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
/* Too big to buffer */
if (stream_flush_buf())
return write(rec_fd, buf, 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
if (!count)
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
if (stream_buf_used + count > STREAM_BUF_SIZE)
{
if (!stream_flush_buf() && stream_buf_used + count > STREAM_BUF_SIZE)
count = STREAM_BUF_SIZE - stream_buf_used;
}
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
memcpy(stream_buffer + stream_buf_used, buf, count);
stream_buf_used += count;
return count;
}
/* One-time init at startup */
void INIT_ATTR recording_init(void)
{
/* Init API */
ci.enc_pcmbuf_read = enc_pcmbuf_read;
ci.enc_pcmbuf_advance = enc_pcmbuf_advance;
ci.enc_encbuf_get_buffer = enc_encbuf_get_buffer;
ci.enc_encbuf_finish_buffer = enc_encbuf_finish_buffer;
ci.enc_stream_read = enc_stream_read;
ci.enc_stream_lseek = enc_stream_lseek;
ci.enc_stream_write = enc_stream_write;
}