-ffast-math was added in ff14092d8d with a comment that it "helps SINC
resampler to auto-vectorize". The flag is an alias for "-fno-math-errno
-funsafe-math-optimizations -ffp-contract=fast -fno-honor-infinities
-fno-honor-nans". The last two of those cause compiler complaints
because while the flag was meant for this file, it's shared across the
codebase that includes statements that operate on infinities and NaNs.
GCC 13.3 with -fopt-info-vec reports 18 vectorizations for `-O3
-ffast-math`, 17 for `-O3 -fno-math-errno -funsafe-math-optimizations
-ffp-contract=fast` and also 17 for plain `-O3`. So using the subset of
-ffast-math without the offending flags buys nothing and loses 1
vectorization.
Both GCC and Clang provide the "fast-math" pragma directive, which I add
to this one file that benefits from it, under the condition that it's
supported. -ffast-math is removed from most of the makefiles.
* Less string copies
* Some general cleanups
* Add extra param to runloop_message_queue_push so we can pass size_t
of the message
* Consistent conventions for local variable usage for certain things
* Some slight fixes
* Update libretro.h
* Log calls to RETRO_ENVIRONMENT_GET_MICROPHONE_INTERFACE
* Finish proof-of-concept for mic support
- It works, but doesn't support floating-point audio yet
- It may need to be resampled, too
* Add macros that aren't available in SDL 2
* Comment out a variable definition for now
- For C89 compliance
* Add some comments for clarity
* Let ALSA tolerate a null new_rate
* Partial ALSA microphone support
- Not yet tested
- Mic is created and destroyed
- Mic can also be paused or unpaused
- Mic is paused or unpaused with the rest of the driver
- Microphone is not yet read
* Install error logging in the ALSA driver
- It defers to RARCH_ERR
* Free the ALSA microphone in alsa_free
* Fix an indent
* First draft of alsa_read_microphone
* Deinitialize SDL Audio in sdl_audio_free
* Save and restore the ALSA error logger
- You should always practice safe global state
* Add newlines to some RARCH_ERRs
* Add some logging
* Check for the mic being active via settings instead of via flags
* Adjusted a log entry to be less misleading
- A frequency of 0Hz looks weird to the uninformed
- In reality, it means the driver used the requested frequency
* Fix an incorrect format string
* Tidy up logging in alsa.c
* Rename audio_enable_microphone to audio_enable_input
* Rename microphone_device to audio_input_device
* Add audio_input_latency and audio_input_block_frames settings
* Add all mic-related settings to the options menu
* Adjust logging for alsa.c
- Log the ALSA library version
- Add errno details
* Refer to the microphone in logs by name
* Use %u instead of %d for some log items
* Add input_samples_buf
* Remove an inaccurate comment
* Change type of input_samples_buf
* Clean up audio_driver_flush_microphone_input
* Comment convert_float_to_s16
- It helped me understand what it's doing
- Turns out it'll work just fine on mono audio
* Don't use the resampler for mic input
* Fix crash in the ALSA driver when reading from a mic
* Update some logging messages
* ALSA support now works for mics
* Reuse some common functions in alsa.c
* Add alsa_thread_microphone_t
* Refactor alsa.c
- Introduce alsa_init_pcm to init any PCM that we're using
- Vastly simplifies the implementation of alsa_init and alsa_init_microphone
- Will be used for the read-based versions next
* Make ALSA logging a little more consistent
* Clean up the mic with alsa_free_microphone if alsa_init_microphone fails
* Remove an unused function
* Move some cleanup in alsa.c to a common function
* First crack at mic support for alsathread
- Refactor some duplicate code into functions
- Use functions introduced in alsa.c
- Create and destroy the mic
* Slight cleanups for clarity
* Implement alsa_thread_set/get_microphone_state
* More work on alsathread
- No more crashing, but the mic just returns silence
* Slight cleanups for clarity
* Add alsa_set_mic_enabled_internal
- For setting the state of a microphone while considering its current state
* Use alsa_set_mic_enabled_internal
* Log a little more info
* Log when the audio driver is started/stopped
* Move base microphone driver code into a new directory
- Add microphone_driver.c to Makefile.common
- Rename functions as needed
* Initialize and deinitialize the microphone driver
* Implement sdl_microphone.c
* Un-const an argument
- In case the driver context needs to do any locking
* Revise comments for microphone_driver.h
* Remove an unimplemented function
* Remove some functions from the mic driver
* Remove mic functions from audio_thread_wrapper
* Remove mic functions from sdl_audio
* Fix microphone_null
* Split the mic code for the alsa audio drivers into microphone drivers
* Fix an extra struct member
* Add a setting for the mic driver
* Add a command to reinitialize the microphone driver
* Rename mic-related settings
* Add DRIVER_MICROPHONE_MASK to DRIVERS_CMD_ALL
* Rename audio_enable_input to microphone_enable
* Remove some labels from qt_options
* Search for microphone_driver within find_driver_nonempty
* Clean up some mic driver code
* Pending mics now return silence
* Adjust some logging and comments
* Some cleanup in the microphone driver
* Invert a flag check
- Oops
* Fix a log message
* Fix the wrong flags being checked
* Slight refactor of wasapi_init_device
- Add a data_flow parameter
- Declare it in a header
- In preparation for WASAPI mic support
* Add some WASAPI macros for _IAudioCaptureClient
* Move some common WASAPI functions to audio/common/wasapi.c
- They'll be used by the mic and the audio drivers
* Add wasapi_log_hr
* Generalize mmdevice_list_new to look for capture devices, too
* Fix a function declaration
* Move driver-specific device_list_new functions into their respective files
* Clean up some declarations
* First draft of wasapi microphone driver
* Add wasapi_microphone_device_list_free
* Change function parameter names to be consistent with microphone_driver
* Partially implement wasapi_microphone_read
- Mostly copied from the audio driver so far
- It doesn't compile yet
- But it'll be beautiful when I'm done with it
* Refactor the mic driver's functions
- Rename get_mic_active to mic_alive
- Split set_mic_active into start_mic and stop_mic
- Refactor the SDL mic driver accordingly
* Edit some WASAPI functions for logging and clarity
* Implement more of the WASAPI mic driver
* Rename write_event to read_event
* Pass the WASAPI driver context to the various read functions
* Mostly implement the read function for the WASAPI mic driver
* Fix a crash in microphone_driver
- Forgot to move the position of the name of null_driver
* Reduce some logging in wasapi common functions
- Only log the chosen audio client format, not all attempted ones
* Add some macro wrappers for IAudioClient methods
* Update mic driver configuration
- Make the mic driver configurable in the menu
- Add config items for WASAPI-related options similar to the audio driver
* Fix a menu entry scrolling through audio devices instead of mic devices
* Add some utility functions
* Expose the new utility functions in wasapi.h
* Add extra logging in the WASAPI common functions
* Add sharemode_name
* Use _IAudioClient_Initialize macro in some places
* Pass channels to wasapi_init_client
- Remember, mics are in mono
* Use _IAudioClient_Initialize macro some more
* Forgot to pass channels in some places
* Add some utility functions
* Forgot an #include
* Add wasapi_select_device_format
* Simplify the format selection logic in wasapi_init_client_sh
* Unset the microphone in wasapi_microphone_close_mic
- Ought to prevent a potential segfault
* Simplify some logging
* Fix incorrect value being passed to _IAudioCaptureClient_ReleaseBuffer
* Remove some unneeded logging
* Add some values to hresult_name
* Polish up wasapi_select_device_format
- Test for formats manually when Windows can't
- Add some debug logging
- Check for channels
* Compute the fields of WAVEFORMATEXTENSIBLE correctly
- As per the doc's stated requirements
* Simplify logic for WASAPI client creation
* Fix a potential hang in wasapi_microphone_read_shared_buffered
* Stop the microphone if the driver is stopped
* Don't name the microphone event
* Ensure that wasapi_init_client reports the correct format and rate
* Implement exclusive microphone read access for WASAPI
* Add _IAudioCaptureClient_GetNextPacketSize macro
* Organize cases in hresult_name
* Clear some extra fields if wasapi_set_format is setting a Pcm format
* Adjust some logs
* Adjust some logs
* Remove unneeded local vars
* Add a log
* Update wasapi.c
* Update wasapi.c
* Fix shared-mode mic support in WASAPI producing broken input
- Turns out it had nothing to do with shared mode
* Reuse a common function
- Remove wasapi_microphone_read_shared_buffered
- Rename wasapi_microphone_read_exclusive to wasapi_microphone_read_buffered
* Remove some code I was using for test purposes
* Clarify some language
* Double the default shared-mode mic buffer length
* Split getting a device's name into a separate function, then use it
* Fix the ALSA mic drivers
- To comply with changes I previously made to the mic driver interface
* Remove unused synchronization primitives from the SDL microphone driver
* Add sdl_microphone_mic_use_float
* Document audio_driver_state_flags
- I needed to understand these to see if similar flags were required for the mic driver
* Remove an unused function in wasapi.c
* Add and document flags in microphone_driver.h
* Remove driver-specific mic start/stop functions
- The mic driver itself doesn't do much processing
- That honor goes to individual mics
* Remove some unused fields in microphone_driver.h
* Add CMD_EVENT_MICROPHONE_STOP/START
* Remove unused functions from microphone_null
* Change how the mic driver state is referenced in some places
* Simplify the SDL microphone driver
- The driver backend no longer keeps a reference to the mic (the frontend does that)
- Remove functions that are no longer needed
- Don't track paused state, just query the mic itself
* Simplify the WASAPI microphone driver
- Don't track the driver running state or the microphone handle, the frontend does that now
- Remove support for unbuffered input (hunterk suggested that it wasn't necessary)
* Make microphone_wasapi_sh_buffer_length a uint, not an int
- It won't be negative anymore
- 0 now represents the default value
* Make the microphone frontend more robust
- Improve documentation for how various functions should be implemented
- Closes all microphones before freeing the driver (so backends don't have to)
- Tracks the enabled state of each microphone, so backends don't have to (but they still can)
* Stop the mic driver in core_unload_game
* Ensure mic support is compatible with the revised menu code
* Move alsa.h into audio/common
* Remove RETRO_ENVIRONMENT_GET_MICROPHONE_ENABLED
- It was never really needed
* Refactor the ALSA microphone driver
- Move common ALSA functions to audio/common/alsa.c
- Replace alsa_set_mic_enabled_internal with alsa_start/stop_pcm
- Don't track the microphone handle in the ALSA driver context
- Remove unneeded fields
* Move some common alsathread code into audio/common/alsathread.c
* Change return type of mic_driver_open_mic_internal to bool
* First crack at resampling mic input
* Remove an extraneous check
- I think something distracted me when I was writing this line
* Add stereo/mono conversion functions
* Make alsa_start_pcm and alsa_stop_pcm more robust
- They now return success if the stream is already running and stopped, respectively
* Revise some mic-related comments in libretro.h
* First crack at resampling mic input
* Simplify an expression
* Simplify an expression
* Fix a log tag
* Allow mic resampler to be configured separately from audio resampler
* Add some comments
* Set the source ratio to something sensible
* Stop deadlock in `alsathread` mic driver
* Allow mics to be initialized even when core is loaded from CLI
- When loading content from CLI, the drivers are initialized a little differently
- That threw off the mic initialization code
* Rename the functions in retro_microphone_interface
* Revise some mic-related comments in libretro.h
* Update retro_microphone_interface
- Add get_mic_rate
- Add a parameter to open_mic
- The modifications don't do anything yet
* Use parameter objects in the microphone handle
* Replace get_mic_rate with get_params
* Add a microphone interface version
* Remove part of a comment
* Set the effective params in mic_driver_microphone_handle_init
* Drop a stray newline
* Change where the mic interface is zeroed
- I was accidentally throwing out the version that the core was asking for
* Reduce logspam for wasapi_set_nonblock_state
- Now it only logs when the sync mode is changed
* Change DEFAULT_WASAPI_SH_BUFFER_LENGTH to 0
- -16 is no longer a valid value
* Set the new_rate in wasapi_init
* Change description of microphone sample rate in the settings
* First attempt at resampling configured mic input
* Forgot a section
* Fix some input samples being skipped
* Rename a variable for clarity
* Add microphone.outgoing_samples
* Update the mic driver
- Processed samples are now buffered
- The resampler is skipped if the ratio is (very close to) 1
* Remove part of a comment
* Update some comments in audio_resampler.h
* Slightly refactor the SDL microphone driver
- Move SDL_AudioSpec to a field of sdl_microphone_handle_t
- Allow SDL to change the requested format and sample rate
- Request floating-point input
- Implement sdl_microphone_mic_use_float
* Fix a non-C89-compliant declaration
* Add new files to griffin.c
* Remove a C++-style comment
* Add two more files to griffin.c
* Remove some unneeded declarations in microphone_driver.h
* Remove a stray comma in configuration.c
- For C89 compliance
* Fix compilation on some platforms
* Change some function signatures
* Make the ALSA drivers always set the audio rate
* Fix the alsathread mic driver
* Make state_manager_frame_is_reversed return false if HAVE_REWIND isn't defined
* Mute the microphone if the core is running in fast-forward, slow-mo, or rewind
* Clarify a comment
* Clarify a comment
* Add a comment
* Don't allocate memory for slowmo samples in the mic driver
- We're not supporting slowmo for mics, so it's not needed
* Fix a {
* Add my name to AUTHORS.h
* Add driver_lifetime_flags
- For drivers that have special setup/teardown needs
* Ensure that resetting the mic driver maintains active mic handles
- Prevents fullscreen toggle from stopping all mic input
* Update CHANGES.md
* Move some default microphone settings to a new part of the config file
* Ensure that RetroArch can use the audio format that Windows suggests
* Remove references to mic support in the SDL audio driver
* Remove unused WASAPI functions
* Return failure if RetroArch couldn't select a WASAPI format
* Ensure that Windows uses the WASAPI mic driver by default
* Treat disabled mic support as a warning, not an error
* Clarify some WASAPI-related microphone settings
* Remove some unused variables
* Add or revise microphone-related comments
* Rearrange doc comments for microphone types in libretro.h
* Remove a space
* Remove some unused flags
* Remove ALSA error logger
- It was never used anyway
* Remove unneeded microphone-related arguments
* Document a parameter
* Remove a logging call
* Add a constant for the microphone's shared buffer length for WASAPI
* Fix stylistic inconsistencies
* Make mic_driver_get_sample_size a macro instead of a function
* Move the microphone implementation to the audio directory
* Make microphone support optional (but enabled by default)
* Fix the griffin build
* Add intrinsic NEON versions for float_to_s16/s16_to_float courtesy
of davidgfnet
* Define -DDONT_WANT_ARM_OPTIMIZATIONS for resampler sinc - this should
default to intrinsic versions
* Default to ARM NEON intrinsic codepath and make the ASM codepaths
optional by defining HAVE_ARM_NEON_ASM_OPTIMIZATIONS
* (Pkg/apple/Android) Take out ASM files being compiled in
pstef