And adjust some GML properties. Since a808cfa, splitters grow
opportunistically. Setting them to fixed sizes now quite literally
fixes them in place. Fixes immovable splitters missed in the
aforementioned commit.
The new layout system conveniently calculates these for us now.
In the case of Mandelbrot where it needs to be overriden, make
sure to disable obey min widget size first. In EmojiInputDialog's
case, the window needs to be resized instead to center correctly.
Each of these strings would previously rely on StringView's char const*
constructor overload, which would call __builtin_strlen on the string.
Since we now have operator ""sv, we can replace these with much simpler
versions. This opens the door to being able to remove
StringView(char const*).
No functional changes.
This has been overkill from the start, and it has been bugging me for a
long time. With this change, we're probably a bit slower on most
platforms but save huge amounts of space with all in-memory sample
datastructures.
The file is now renamed to Queue.h, and the Resampler APIs with
LegacyBuffer are also removed. These changes look large because nobody
actually needs Buffer.h (or Queue.h). It was mostly transitive
dependencies on the massive list of includes in that header, which are
now almost all gone. Instead, we include common things like Sample.h
directly, which should give faster compile times as very few files
actually need Queue.h.
Previously, we were sending Buffers to the server whenever we had new
audio data for it. This meant that for every audio enqueue action, we
needed to create a new shared memory anonymous buffer, send that
buffer's file descriptor over IPC (+recfd on the other side) and then
map the buffer into the audio server's memory to be able to play it.
This was fine for sending large chunks of audio data, like when playing
existing audio files. However, in the future we want to move to
real-time audio in some applications like Piano. This means that the
size of buffers that are sent need to be very small, as just the size of
a buffer itself is part of the audio latency. If we were to try
real-time audio with the existing system, we would run into problems
really quickly. Dealing with a continuous stream of new anonymous files
like the current audio system is rather expensive, as we need Kernel
help in multiple places. Additionally, every enqueue incurs an IPC call,
which are not optimized for >1000 calls/second (which would be needed
for real-time audio with buffer sizes of ~40 samples). So a fundamental
change in how we handle audio sending in userspace is necessary.
This commit moves the audio sending system onto a shared single producer
circular queue (SSPCQ) (introduced with one of the previous commits).
This queue is intended to live in shared memory and be accessed by
multiple processes at the same time. It was specifically written to
support the audio sending case, so e.g. it only supports a single
producer (the audio client). Now, audio sending follows these general
steps:
- The audio client connects to the audio server.
- The audio client creates a SSPCQ in shared memory.
- The audio client sends the SSPCQ's file descriptor to the audio server
with the set_buffer() IPC call.
- The audio server receives the SSPCQ and maps it.
- The audio client signals start of playback with start_playback().
- At the same time:
- The audio client writes its audio data into the shared-memory queue.
- The audio server reads audio data from the shared-memory queue(s).
Both sides have additional before-queue/after-queue buffers, depending
on the exact application.
- Pausing playback is just an IPC call, nothing happens to the buffer
except that the server stops reading from it until playback is
resumed.
- Muting has nothing to do with whether audio data is read or not.
- When the connection closes, the queues are unmapped on both sides.
This should already improve audio playback performance in a bunch of
places.
Implementation & commit notes:
- Audio loaders don't create LegacyBuffers anymore. LegacyBuffer is kept
for WavLoader, see previous commit message.
- Most intra-process audio data passing is done with FixedArray<Sample>
or Vector<Sample>.
- Improvements to most audio-enqueuing applications. (If necessary I can
try to extract some of the aplay improvements.)
- New APIs on LibAudio/ClientConnection which allows non-realtime
applications to enqueue audio in big chunks like before.
- Removal of status APIs from the audio server connection for
information that can be directly obtained from the shared queue.
- Split the pause playback API into two APIs with more intuitive names.
I know this is a large commit, and you can kinda tell from the commit
message. It's basically impossible to break this up without hacks, so
please forgive me. These are some of the best changes to the audio
subsystem and I hope that that makes up for this :yaktangle: commit.
:yakring:
With the following change in how we send audio, the old Buffer type is
not really needed anymore. However, moving WavLoader to the new system
is a bit more involved and out of the scope of this PR. Therefore, we
need to keep Buffer around, but to make it clear that it's the old
buffer type which will be removed soon, we rename it to LegacyBuffer.
Most of the users will be gone after the next commit anyways.
It seems like this happens in quite some valid situations, so my
initially sensible failsafe doesn't make sense. As the buffer system is
hopefully gone soon, it won't be an issue in the future either way.
This was regressed at some point though I never saw it working.
Basically, while jump to slider works correctly it doesn't even get
actioned. While the user is clicking the slider it's very likely that a
buffer finishes playing and the callback for that changes the slider
value. This means that the user click just gets lost. There's some
additional weird behavior where values are lost in even more cases, so
an additional fix that is needed is to store the slider value in the
AutoSlider while we're dragging and apply it on mouse up.
Now that we have y-axis (gain) logarithmic display, we should also have
x-axis (frequency) logarithmic display; that's how our ears work. This
can be turned off with an option, but it generally looks much nicer.
For DSP reasons I can't explain myself (yet, sorry), short-time Fourier
transform (STFT) is much more accurate and aesthetically pleasing when
the windows that select the samples for STFT overlap. This implements
that behavior by storing the previous samples and performing windowed
FFT over both it as well as the current samples. This gives us 50%
overlap between windows, a common standard that is nice to look at.
The input to the FFT was distorted by the usage of fabs on the samples.
It led to a big DC offset and a distorted spectrum. Simply removing fabs
improves the quality of the spectrum a lot.
The FFT input should be windowed to reduce spectral leakage. This also
improves the visual quality of the spectrum.
Also, no need to do a FFT of the whole buffer if we only mean to render
64 bars. A 8192 point FFT may smooth out fast local changes but at 44100
hz samplerate that's 200 ms worth of sound which significantly reduces
FPS.
A better approach for a fluent visualization is to do small FFTs at the
current playing position inside the current buffer.
There may be a better way to get the current playing position, but for
now it's implemented as an estimation depending on how many frames where
already rendered with the current buffer.
Also I picked y-axis log scale as a default because there's usually a
big difference in energy between low and high frequency bands. log scale
looks nicer.
Visualization widgets should only have to tell how many samples they
need per frame and have a render method which receives all data relevant
to draw the next frame.
Although it's nice to have this as an option, it should be the default
to adjust higher frequencies as they intrinsically have less energy than
lower energies.
Several related improvements to our Fast Fourier Transform
implementation:
- FFT now operates on spans, allowing it to use many more container
types other than Vector. It's intended anyways that FFT transmutes the
input data.
- FFT is now constexpr, moving the implementation to the header and
removing the cpp file. This means that if we have static collections
of samples, we can transform them at compile time.
- sample_data.data() weirdness is now gone.
Display the album cover for the current playing song in the
visualization area for the "None" Visualization.
For now only "cover.png" and "cover.jpg" are looked for in the same
directory for the album cover image.
When no cover image is found the serenity background is shown instead as
a fallback.
This adds a new start_new_file() function to VisualizationWidget which
is called when the player starts a new file, passing the filename of the
file. This allows VisualizationWidget subclasses to do any setup needed
when a new file is started.
When the bars visualization receives a new buffer, it checks if it needs
a new buffer, which is only the case after it has repainted. However,
after then setting m_is_using_last, which is the flag for this, it
checks the buffer size of the passed buffer and returns if that is too
small. This means that if the visualizer receives a buffer that is too
small, and because of external circumstances the update doesn't run
after the buffer modification routine, the m_is_using_last variable is
stuck at true, which means that the visualization incorrectly believes
that the passed buffer is old and we need not update. This simply fixes
that by resetting m_is_using_last if the buffer we're passed is too
small, because in that case, we're clearly not using the last buffer
anymore.
Note: This bug is not exposed by the current SoundPlayer behavior. It
will become an issue with future changes, so we should fix it
regardless.
Enable the warning project-wide. It catches when a non-virtual method
creates an overload set with a virtual method. This might cause
surprising overload resolution depending on how the method is invoked.
Previously, SoundPlayer would read and enqueue samples in the GUI loop
(through a Timer). Apart from general problems with doing audio on the
GUI thread, this is particularly bad as the audio would lag or drop out
when the GUI lags (e.g. window resizes and moves, changing the
visualizer). As Piano does, now SoundPlayer enqueues more audio once the
audio server signals that a buffer has finished playing. The GUI-
dependent decoding is still kept as a "backup" and to start the entire
cycle, but it's not solely depended on. A queue of buffer IDs is used to
keep track of playing buffers and how many there are. The buffer
overhead, i.e. how many buffers "too many" currently exist, is currently
set to its absolute minimum of 2.
This shortcut let us mute/unmute the player, but it still doesn't update
the volume slider because the actual volume widget can't display a muted
state.
This fix syncs up the AudioPlayer's internal state for showing
playlist information with the AudioPlayer's GUI. Before, if the
AudioPlayer was opened with a playlist file (.m3u or .m3u8) it would
automatically show the playlist information in the GUI and set the
loop mode to playlist, but the menu options would be unchecked. In
order to hide the playlist information, the menu option would then
have to be toggled twice -- once on and again off.
Previously, a libc-like out-of-line error information was used in the
loader and its plugins. Now, all functions that may fail to do their job
return some sort of Result. The universally-used error type ist the new
LoaderError, which can contain information about the general error
category (such as file format, I/O, unimplemented features), an error
description, and location information, such as file index or sample
index.
Additionally, the loader plugins try to do as little work as possible in
their constructors. Right after being constructed, a user should call
initialize() and check the errors returned from there. (This is done
transparently by Loader itself.) If a constructor caused an error, the
call to initialize should check and return it immediately.
This opportunity was used to rework a lot of the internal error
propagation in both loader classes, especially FlacLoader. Therefore, a
couple of other refactorings may have sneaked in as well.
The adoption of LibAudio users is minimal. Piano's adoption is not
important, as the code will receive major refactoring in the near future
anyways. SoundPlayer's adoption is also less important, as changes to
refactor it are in the works as well. aplay's adoption is the best and
may serve as an example for other users. It also includes new buffering
behavior.
Buffer also gets some attention, making it OOM-safe and thereby also
propagating its errors to the user.
LibDSP can greatly benefit from this nice FFT implementation, so let's
move it into the fitting library :^)
Note that this now requires linking SoundPlayer against LibDSP. That's
not an issue (LibDSP is rather small currently anyways), as we can
probably make great use of it in the future anyways.
The path returned by GUI:FilePicker is stored on the stack when the
callback is executed. The player only stored a StringView to the path
however it should take ownership of the path instead since the path is
accessed even after the file menu open action has returned.
