This commit changes the following APIs to add an additional parameter
called pAllocationCallbacks. This is a pointer to a
ma_allocation_callbacks structure which is used to perform the actual
allocation/free.
Allocation callbacks can be set in ma_context_config for contexts and
ma_decoder_config for decoders. If it's left unset it will use
MA_MALLOC(), MA_REALLOC() and MA_FREE().
With this change, the ma_device_id union should now be a consistent
size across all platforms. This is important if cross-platform
programs want to save the device ID and need the size to be consistent.
The following APIs are removed:
* ma_sine_wave_read_f32()
* ma_sine_wave_read_f32_ex()
Use ma_sine_wave_read_pcm_frames() as a replacement. This allows you to
generate sine wave samples in any format, controlled by a parameter.
Resampling is now done through the ma_resampler API.
Note that with this commit the old sinc resampler has been removed
because it never worked properly and is going to replaced with a better
solution in the future. If you were using ma_src_algorithm_sinc you
should consider using ma_src_algorithm_linear with the linear.lpfCount
config variable set to MA_MAX_RESAMPLER_LPF_FILTERS.
For safety, ma_convert_frames() has an additional parameter called
frameCountOut which is the capacity of the output buffer. In addition,
the frameCountIn parameter has been moved next to the pIn parameter.
ma_convert_frames_ex() has been changed to take a pointer to a
ma_data_converter_config object. This provides more flexibility as to
the input and output formats. In addition, the frameCountOut parameter
has been added which has the same meaning as ma_convert_frames().
Previously this would result in a glitch, however in the interest of
avoiding as many sources of glitching as possible, changing the format
and/or channel count of a biquad filter (and by extension, anything
derived from it, such as low-pass filters) will now result in an error.
This change reduces memory usage by removing half the number of delay
registers.
This commit also sets up some infrastructure for future work on a fixed
point implementation. Both the floating and fixed point implementations
will run on a 32-bit pipeline for now.
David Reid