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>
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>
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>
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>
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>
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>
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>
When reading partial chunks, and when returning partial data, it is
essential to maintain this lossless chunk status across either whole or
partial chunk reads. Otherwise, the converter chain sees the lossless
flag constantly changing on lossless files, such as PCM or DSD, and
causes the DSD decimator and/or resampler to be torn down and reset
repeatedly, causing glitches in the audio.
The glitch was not, in fact, with the decimator itself, and was
occurring to a degree without it, as it would be restarting the
resampler repeatedly as well.
Fixes#367
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The latency is half of the FIFO, or half the filter size, and each byte
is 8 samples, so return the value accordingly.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This comment was in the original sample decimator code, I neglected to
include it in my port over to Cog. Doesn't really serve any functional
change, though. It would have clarified that I needed to reduce the gain
level much sooner, though.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Input thread now signals when it has stopped and is about to return, in
case the input thread returns before the BufferChain dealloc function
would be waiting for it to terminate. Somehow, even though the Semaphore
is being signaled at this point, the BufferChain still ends up waiting
the default of 2.5 seconds for the signal that apparently never comes,
delaying file stoppage. This prevents the wait action entirely. Must
have been some sort of race condition.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This new method should cause all stops to default to immediate stoppage,
and only stops that occur after an end of track signal should indicate
to play out the entire buffer.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This magically fixes the stupid header maps that were pulling the system
semaphore.h into Swift projects, when they shouldn't have been doing
that in the first place. This is the same reason that the FLAC library
has its assert.h renamed to FLAC_assert.h.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
These imports needed to be changed so that Swift bridging didn't import
the system's semaphore.h instead of CogAudio's Semaphore.h, which is a
completely different thing.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Oops, this compare blunder resulted in DSD decimation breaking every
1024 samples or so, owing to block sizes, and caused ticking sounds as a
result. It would also cause HDCD decoding to break completely.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Neither of these two changes is really important, but they do simplify
things, and the division on that one function makes the non-decimating
DSD support actually functional, as the caller expects a specific number
of samples, and that was otherwise octupling the input sample count.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Sample rate changes will now occur on exact sample boundaries, like they
are supposed to. Also, FreeSurround accounts for its output latency.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Apparently, all these new changes with FreeSurround have pushed the
default 512KB thread stack size to the limit. And I'm not even using
stack variables, really, except for maybe the autoreleasepools.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Move the Float32 converter to a different location, for any future plans
to support decoding audio files to common data for any other purpose.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This should also seal up any potential hole for problems if there's an
audio format change and no audio buffered.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The input chain could hang up indefinitely, and MAD decoder didn't
indicate end of file properly.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
readAudio now returns an AudioChunk object directly, and all inputs have
been changed to accomodate this. Also, input and converter processing
have been altered to better work with this.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Stop output when requested, except on natural completion of the last
track in the play queue. Also fix deadlocks with stopping and
restarting.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The output now uses AVSampleBufferAudioRenderer to play all formats, and
uses that to resample. It also supports Spatial Audio on macOS 12.0 or
newer. Note that there are some outstanding bugs with Spatial Audio
support. Namely that it appears to be limited to only 192 kHz at mono or
stereo, or 352800 Hz at surround configurations. This breaks DSD64
playback at stereo formats, as well as possibly other things. This is
entirely an Apple bug. I have reported it to Apple with reference code
FB10441301 for reference, in case anyone else wants to complain that it
isn't fixed.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This code turned out to be somewhat of a mistake to employ, so it's now
being removed, and shall not be re-added, as it doesn't really work.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Add play count data collection, including first seen times for every
file first added to the playlist. Data is indexed by album, artist, and
title, or by filename, whichever matches first. Add interfaces to
AppleScript automation definition as well.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
As it doesn't seem to work properly on Intel machines, anyway. It just
leads to pointless crashes, and doesn't seem to serve any purpose.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Restrict the use of workgroup joining and workgroup intervals to macOS
Monterey or newer, as it seems the way I use it, it's completely broken
on macOS Big Sur, which was the original minimum target for the API.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Apply changes to exit the thread if workgroup initialization or joining
fails, instead of attempting to continue executing the thread.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Add an extra step to the workgroup exit call, so that it only calls to
leave if the join token is valid, or at least initialized.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The changes include no longer leaving the workgroup for seeking or for
converter format changes, and also still leaving the workgroup on thread
termination if there was an error with intervals starting or finishing.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
An impulse cache reduces any glitching from format channel count changes
to near insignificant levels, resulting in a more pleasant experience
when there are different mixed formats playing, or even a file which
changes format mid-playback.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Ensure that dynamic info updates, even on static files, only update the
exact track they apply to, by atomically assigning the userInfo property
before opening the decoder, so that callbacks to the player indicate the
correct track and don't assume it's the one that's currently visibly
playing. Fixes start of track metadata notifications from overwriting
the previously playing track.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
DFT float happens to clobber one extra sample on forward translate, so
allocate one extra for every complex buffer.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Work back to a vDSP implementation, this time using overlap-save instead
of overlap-add, also accumulating the results as complex values, only
inversing them once at the end, and finally, replacing the FFT method
with the newer DFT API.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
When leaving the workgroup, clear the token, as the join call requires
the token to be uninitialized.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Errors should stop all attempts to further use the audio thread priority
code, so there won't be debug breakpoints called on older OSes.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
As the decimator has shown to be twice as loud as it should be, the
volume should be reduced by half when converting DSD to PCM with it.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Pure downsampling is slower, but may or may not be more accurate. Though
probably not worth it. It did help me realize a minor error, though.
The decimator's volume is twice as loud as it should be.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Set baseline real-time priority for audio threads even on old macOS,
since that API is available there. Only set it once, and do not attempt
again if it fails, only once per thread.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
I'm not sure about macOS Ventura, but stable releases of macOS, at
least on Intel, require that threads joining Audio Interval
workgroups already be set to run as real-time before joining. Not
doing this results in an uncaught exception and a crash.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Now it allocates audio workgroups per thread, using work slices like the
Apple documentation describes for asynchronous threads.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
On Big Sur or newer, it is possible to join the audio threads to the
same OS workgroup as the audio output device, improving response.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Replace overlap-add vDSP/Accelerate implementation with a faster PFFFT
overlap-save implementation, using fewer FFT steps as well.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Rewrite some of the output and a lot of the downmixer to use Accelerate
framework instead of dumb for loops.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
When resampling the impulse according to the playback rate, it becomes
necessary to normalize the resulting impulse by the inverse of the
sample ratio, as resampling adds more or less loudness by virtue of
interpolating samples.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
The resampler wasn't being given enough room to flush its final output,
so a function was added to determine the current output latency, and
more sample data is requested, allowing the full output flush to occur.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
These two changes fix playback issues with either starting in the middle
of the playlist on a really short file terminating immediately instead
of queueing more files (InputNode.m), and issues with starting playback
at all on the end of a playlist on a short file. (OutputCoreAudio.m)
Fixes#246
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
vDSP functions expect their input and output pointers to be aligned to
an even four values. Correct this by aligning all pointers. The
allocated buffers used for one parameter should already be aligned
somewhat, but align the incremented positions used on some of them so
that the vDSP functions don't misbehave. Also align the volume scaler
input by doing scalar math until the pointer is aligned prior to calling
vDSP_vsmul. Also, change 16-bit and 32-bit scale to use vsdiv instead of
vsmul with a really small number already divided into one.
Fixes the test vectors that were sent in extrapolating incorrectly due
to their final blocks having uneven sample counts, resulting in
unaligned pointers.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
A bad sample scanner and cleaner will point out in the log whenever a
bad sample, such as infinity, or Not a Number, or even huge values over
±2.0, in case some piece of code, or a decoder, or even a bad file, has
taken over the output.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
In the rare event that we're somehow playing decimated DSD at full
sample rate instead of resampling, only the start needs to be skipped,
and the end needs the input to the decimator padded to flush it, but
nothing needs to be truncated from the end of the output in that case.
Still, mostly pointless, since next to nobody will be outputting 384 kHz
from their Macs, in any case, much less unprocessed DSD.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
We should be extrapolating right over top of the DSD decimator latency,
rather than in front of it. Yeah, that'll do.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Metadata versus properties merging, correctly merge over empty fields if
they are assigned with empty strings or zeroed numbers, instead of only
merging over completely missing fields.
Fixes emailed bug, CUE Sheet metadata reading, primarily.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
For big endian sample formats, endianness can be swapped using Clang
specific byte swap functions, which are present in all supported
versions of Xcode.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
Simple upmixing algorithms now use Accelerate framework functions
instead of complex loops, and the HRIR filter now supports forcing
stereo output to any channel output configuration that has at least
front stereo speakers.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
When decoder is redirected to the internal silence decoder, show an icon
on the playlist indicating a playback error.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
This resulted in horrible things, the generic N to N upmixer was leaving
unmapped channels as uninitialized memory. This fixes horrible things
happening for people with interfaces with more channels than the source
file, frequently when the source file is stereo, or if the file is mono
and a center channel is present.
Signed-off-by: Christopher Snowhill <kode54@gmail.com>
