DSPs should not be performing a cascading reset when resetting just
their own buffers, for example, on init or shutdown of just that one
DSP filter.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Cleaned up project settings to current defaults, except for the macOS
deployment version, which is still 10.13. Cleaned up a lot of headers
and such to include with angle braces instead of double quotes. Enabled
build sandbox in a lot of places. Disabled subproject signing in several
places, for libraries and frameworks which will be stripped and signed
when they are copied into place in the final build.
Also, while trying to solve compilation issues, the visualization
controller was reverted to the Objective C implementation, which is
probably faster anyway. Stupid Swift/Objective-C language mixing issues.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Seeking should clear the buffers completely now, and will be nearly
instant, depending on how fast the input can decode.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This includes setting and unsetting the equalizer DSP chain objects on
track change and advancing on track playback end, and also bugs with
applying equalizer presets to the band configuration items when the
equalizer is disabled or when playback is stopped.
Fixes#420
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Seeking now mutes properly, and will not leave the audio muted across
other operations. Audio output changes should also mute and destroy the
buffers of the input chain, so that the audio resets properly.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Fixes output volume setting on seek or audio output restart on format
change. Also safeguards these setters so they don't go off if the nodes
aren't actually allocated.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Now buffer twice as much audio as would be requested for a single
visualization PCM/FFT chunk, which should hopefully prevent it from
flickering due to running out of audio because of too low latency.
Now it buffers up to two chunks at the current hard coded visualization
sample rate, which works out to about 186 milliseconds.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
We implement this function to return the current latency buffered,
regardless of how often this function may be called. In practice, it is
only called on track completion, to time the reporting of the next track
display. We also avoid using Rubber Band's latency function, as in most
cases, this function will be called from other threads, and also, it
currently only gets called after Rubber Band has been emptied out, so it
would otherwise calculate zero samples buffered. And thirdly, Rubber
Band's latency function doesn't account for the buffered samples already
removed from it and waiting to be fed out.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The code was polling the input chunk duration after emptying out the
chunk's samples, which resulted in an input duration account sitting at
exactly zero, so the end overrun flush would not be cut short properly,
resulting in gaps between tracks.
Correct the input sum to tabulate before emptying the input chunk, so
output remains properly gapless.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This code was being duplicated across three different playback functions
which basically did most of the same things.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Check all audio chain elements for allocation failures, and also dispose
of all of the previous handles in reverse order, including nulling the
final node handle so the output does not attempt to poll for audio while
the chain is being rebuilt.
Also set up output node to handle the new null finalNode state, and
return an empty chunk to the caller.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
We should not be processing a potential playback restart when the chain
is being torn down for shutdown.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This should be perfectly safe to use in all situations now. It may have
been unstable due to mishandling return values, or not supporting
requesting more sample data from the library without feeding in more
input first.
Also, still signaling the End of Stream flag on chunk reading should be
correct, as downstream processors only react to it when the buffer runs
empty.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The Downmixer wasn't updating its output format correctly, so it was
prone to outputting the wrong format for a while, which could confuse
the output device and produce garbage output.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The delay value should be scaled by the resampling ratio, similar to
how it already is when allocating the impulse buffer. This went
undetected, as it scribbled over other memory without causing immediate
crashes, but instead later heap corruption.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Check for paused processing state in various places, so that startup
playback works properly, and resume playback at seek offset works
properly and doesn't hang the player.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The samples available function returns a signed integer, so it can
apparently return negative on error, and the DSP was incorrectly casting
this to an unsigned type, and thus attempting to buffer an inordinate
number of samples and crashing.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The merge function should be able to tell when the caller has no audio
left to process, such as on end of stream.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
FreeSurround needs more buffering from its input, so increase buffering
of previous node to 100ms.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Fixes timestamps in several cases where they were being processed
incorrectly, which was causing some chunked audio files to mis-report
timestamps into the past or the future, which caused the seekbar to jump
around in an unpredictable way.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Timed wait for 500us is kind of stupid and makes the threads wake up way
too much, and use way more CPU time. Reduce this, as the semaphores are
signaled appropriately, and the waiter should not wake up constantly.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Visualization now buffers in the audio output pipeline, and uses a
container system to delay multiple buffer chains from emitting
visualization data over top of each other. This should stabilize
display output significantly, while introducing minimal lag before
DSP configuration changes take effect.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The latency should not be incremented when writing sample data to the
buffer, but rather be posted by the output.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Upstream functions which return empty chunks on error do not return nil,
so the caller should check for an empty duration instead.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Impulses should be gain scaled roughly based on the sample ratio
relative to the original impulses. Lower target sample rate means less
impulses means gain goes up, higher target sample rate means more
impulses so gain goes down. Somewhat simple, seems to work.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
We were forcing a resampling ratio to match the HRTF filter supplied
with the app, now we resample the HRTF to match the input audio, which
will be resampled to match the output device settings.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This prepares the filter to be the same as the rest of the filters, in
that they support flexible sample rates to match the output device.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
And disable it by default in new installations, otherwise leave the
setting alone. The disablement setting is shared with the engine
setting, so the default should not really change anything, except for
new installs.
Also, the time/pitch shifting dialog disables itself and displays an
obvious notice button, which opens the Rubber Band settings.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
It's more like the output monitor thread, since it only monitors output,
rather than actually handing the output callbacks.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Sample block merging code should not be duplicated across the DSPs that
require it, but instead should be a common function. Also added some
optimizations to the Float32 converter function, to bypass conversion if
the audio format needs no conversion.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Attempt to completely fill the input buffer of the Rubber Band library
between each call to the process function, instead of processing in
as small an increment as the source node provides. May reduce processing
power required.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Audio Chunks now have full timestamp accounting, including DSP playback
speed ratio for the one DSP that can change play ratio, Rubber Band.
Inputs which support looping and actually reporting the absolute play
position now do so.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
FreeSurround, like the Equalizer, which attempt to coalesce Audio Chunks
into larger blocks of 4096 samples, must check if the audio format has
changed between blocks, and stop stacking chunks together when a new
format is detected. They will continue processing with less sample data
than expected, as necessary.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The last of the built-in processors is now in the threaded processing
chain, and all DSPs are marked high priority and with short buffers.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Change one remaining semaphore wait to 500us, and change the buffering
so that it can always overflow the requested duration by one chunk, so
that at least one chunk will always fit in the buffer. This also allows
the DSP nodes to flush at the end of the stream without losing their
output.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The end of stream flushing should only request remaining samples once,
as should the rest of the process. The problem with the Rubber Band code
in this case is that it will wrap the remaining samples pointer after it
has been flushed, and emit a really huge number.
Also, add code to try to equalize the samples output with the samples
input, relative to the tempo stretching, as Rubber Band seems to flush
entirely too much data at end of stream, which can create noticeable
gaps in the output. This solves that as well.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This should be guarded, so that no other thread tries to free the DSP
while it is potentially writing to the Rubber Band instance.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This comment was copied by accident when duplicating the original
Converter Node class for the new DSP base.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
DSP threads, such as the Rubber Band processing, and planned moves of
other processing to buffer threads, such as the Equalizer, FreeSurround,
HRTF, and Downmixing for output, because they all have small output
buffers. Since these buffers drain and fill fast, they should be
processed at a high priority. Hopefully, App Store doesn't complain
about the use of these APIs.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This class can more flexibly process and emit varying chunk sizes than
the previous code could, solving the problem of wide tempo changes.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Now there's a configuration dialog for tweaking the settings
in semi-real time. Everything that can be changed without
restarting is changed without restarting, otherwise the audio
pipeline is reset, which happens quickly enough anyway.
Awaiting translation to Spanish, other languages have been
removed pending their maintainers fixing most of their
problems, which includes me being lazy and AI translating
bits so I could rush updates.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This should improve performance slightly. It's
still recommended to switch off SceneKit to
save CPU usage, or switch of vis entirely.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
When the input buffer has less samples in it than the LPC order,
it would crash reaching past the ends of the buffer. Now, it will
pad past the correct end of the audio with silence, while still
extrapolating a prime input minimum of the LPC order. Should fix
the last of the outstanding crashes.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
I will implement the more complex setup of providing options for
most of the configuration that Rubber Band provides, at a later
date, when I feel like creating a complex configuration dialog
for it, and asking for help translating every option and setting.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
It should be deriving its channel count from the file format,
since it's applied before any other filters.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Implements a simple speed control using a resampler
designed for real time changes. A rubberband speed
control will be implemented at a later date.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Volume scaling would potentially crash when handling
unaligned blocks of samples, and also handled them
completely wrong. It should be counting up single
samples until the buffer is aligned to a multiple of 16
bytes, and it should not exceed the intended count.
BUG: It was not only counting the unaligned samples
backwards, it was ignoring the real sample count.
Fixes#380
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Shuffle around @autoreleasepool blocks, and also add one
to the audio processing code in the playback callback, so
audio memory is released during playback instead of
accumulating.
Fixes#379
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This appears to maybe be necessary as the prior join call doesn't seem to
be doing what it should.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This is checked inside the audio thread, it isn't needed in the watcher
thread. Remove the second check.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Adjust the buffering so if latency is too low, we fill the rest of
the output with silence instead of peeking at the oldest part
of the buffer. Also increase latency by half a buffer size so
that the requested sample is in the center of the buffer, which
improves the 4096 sample situation with the current low
latency output.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Buffer up to 20 seconds per stage, and buffer only up
to 2 seconds before starting the next stage.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
A stopped instance of OutputCoreAudio should not continue to feed the
visualization system with stale audio, potentially while another instance
is already starting up and feeding its own audio output.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This reverts usage of the AVFoundation output to use
the previous lower latency CoreAudio output, and
paves the way for a change I am cooking up soon.
Fixes several issues with playback and seeking latency.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Do this by serializing the background thread actions against
the AudioPlayer object, so we don't start playback multiple
times at once.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Perform playback start and seeking operations in the background, instead
of on the main thread, which should help prevent them from stalling the
user interface.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This output may prove to have lower latency, but the results are too
glitchy to really be usable. Not even visualization latency is handled
correctly.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
And default it to disabled. As was pointed out to me by a user, DSD is
apparently mastered to a level of -6 dB, so double its level on output
by default.
Also reorder all preferences dialog controls so they are instantiated in
display order, which should help screen readers, maybe.
Fixes#368
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Resampler flush may indefinitely produce 1 sample if there is a rounding
error with the buffering calculations. Work around this.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
When the clipped sample rate changes, the resampler needs to be
restarted. This was previously failing because the target sample rate
wasn't changing.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
