6502/CLK
1
0
Fork 0
mirror of https://github.com/TomHarte/CLK.git synced 2025-11-01 11:16:16 +00:00
Code Issues Projects Releases Wiki Activity

Adds an additional buffer. To reduce latency. No, really.

Browse Source 
Specifically: there's no way to guarantee no overbuffering due to the startup race, other than having QAudioOutput obtain data by pull rather than push. But if it's pulling then that implies an extra buffer. And since the sizes it may pull are not explicit, there's guesswork involved there.

So: no extra buffer => uncontrollable risk of over-buffering. Extra buffer => a controllable risk of over-buffering.
This commit is contained in:
Thomas Harte
2020-06-09 00:01:22 -04:00
parent bcb23d9a15
commit d9f02aecdf
4 changed files with 120 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
OSBindings/Qt/mainwindow.cpp
View File

@@ -12,7 +12,7 @@
* it seems like Qt doesn't offer a way to constrain the aspect ratio of a view by constraining
the size of the window (i.e. you can use a custom layout to constrain a view, but that won't
affect the window, so isn't useful for this project). Therefore the emulation window
affect the window, so isn't useful for this project). Therefore the emulation window resizes freely.
*/
MainWindow::MainWindow(QWidget *parent)
@@ -149,6 +149,7 @@ void MainWindow::launchMachine() {
// Install audio output if required.
const auto audio_producer = machine->audio_producer();
if(audio_producer) {
static constexpr size_t samplesPerBuffer = 256;
const auto speaker = audio_producer->get_speaker();
if(speaker) {
const QAudioDeviceInfo &defaultDeviceInfo = QAudioDeviceInfo::defaultOutputDevice();
@@ -158,7 +159,6 @@ void MainWindow::launchMachine() {
// are available, and — at least for now — assume a good buffer size.
audioIsStereo = (idealFormat.channelCount() > 1) && speaker->get_is_stereo();
audioIs8bit = idealFormat.sampleSize() < 16;
const int samplesPerBuffer = 256;
speaker->set_output_rate(idealFormat.sampleRate(), samplesPerBuffer, audioIsStereo);
// Adjust format appropriately, and create an audio output.
@@ -166,17 +166,12 @@ void MainWindow::launchMachine() {
idealFormat.setSampleSize(audioIs8bit ? 8 : 16);
audioOutput = std::make_unique<QAudioOutput>(idealFormat, this);
// Start the output. setBufferSize is supposed to set the audio buffer size, but
// appears not to work properly on macOS; experimenting with this on that platform
// revealed that setting the buffer size to 1024, then reading it back to find out
// what I actually got results in 2048. If I use a QIODevice-based audio output,
// i.e. one that pulls data as required, it then pulls in units of... 16384 bytes.
// That's almost 1/6th of a second — epic latency.
//
// So I have no idea. TODO: find an alternative audio library, I guess.
// Start the output. The additional `audioBuffer` is meant to minimise latency,
// believe it or not, given Qt's semantics.
speaker->set_delegate(this);
audioOutput->setBufferSize(samplesPerBuffer * sizeof(int16_t));
audioIODevice = audioOutput->start();
audioOutput->start(&audioBuffer);
audioBuffer.setDepth(audioOutput->bufferSize());
}
}
@@ -217,13 +212,14 @@ void MainWindow::launchMachine() {
}
void MainWindow::speaker_did_complete_samples(Outputs::Speaker::Speaker *, const std::vector<int16_t> &buffer) {
const char *data = reinterpret_cast<const char *>(buffer.data());
size_t sizeLeft = buffer.size() * sizeof(int16_t);
while(sizeLeft) {
const auto bytesWritten = audioIODevice->write(data, qint64(sizeLeft));
sizeLeft -= bytesWritten;
data += bytesWritten;
}
audioBuffer.write(buffer);
// const char *data = reinterpret_cast<const char *>(buffer.data());
// size_t sizeLeft = buffer.size() * sizeof(int16_t);
// while(sizeLeft) {
// const auto bytesWritten = audioIODevice->write(data, qint64(sizeLeft));
// sizeLeft -= bytesWritten;
// data += bytesWritten;
// }
}
void MainWindow::dragEnterEvent(QDragEnterEvent* event) {