mirror of
https://github.com/moonlight-stream/moonlight-qt
synced 2024-09-20 14:11:54 +00:00
Remove latency drop logic and replace by single frame writes
This commit is contained in:
parent
cfabaf334e
commit
14cb2e1999
2 changed files with 12 additions and 28 deletions
|
@ -17,6 +17,7 @@ SoundIoAudioRenderer::SoundIoAudioRenderer()
|
|||
m_Device(nullptr),
|
||||
m_OutputStream(nullptr),
|
||||
m_RingBuffer(nullptr),
|
||||
m_Latency(0),
|
||||
m_Errored(false)
|
||||
{
|
||||
|
||||
|
@ -24,6 +25,10 @@ SoundIoAudioRenderer::SoundIoAudioRenderer()
|
|||
|
||||
SoundIoAudioRenderer::~SoundIoAudioRenderer()
|
||||
{
|
||||
SDL_LogInfo(SDL_LOG_CATEGORY_APPLICATION,
|
||||
"Audio latency: %f",
|
||||
m_Latency);
|
||||
|
||||
if (m_OutputStream != nullptr) {
|
||||
soundio_outstream_destroy(m_OutputStream);
|
||||
}
|
||||
|
@ -330,41 +335,19 @@ void SoundIoAudioRenderer::sioWriteCallback(SoundIoOutStream* stream, int frameC
|
|||
(me->m_OpusChannelCount * stream->bytes_per_sample);
|
||||
int bytesRead = 0;
|
||||
|
||||
// Clamp framesLeft to frameCountMax
|
||||
framesLeft = qMin(framesLeft, frameCountMax);
|
||||
|
||||
// Ensure we always write at least a buffer, even if it's silence, to avoid
|
||||
// busy looping when no audio data is available while libsoundio tries to keep
|
||||
// us from starving the output device.
|
||||
frameCountMin = qMax(frameCountMin, (int)(stream->sample_rate * k_RawSampleLengthSec));
|
||||
frameCountMin = qMin(frameCountMin, frameCountMax);
|
||||
|
||||
// Place an upper-bound on audio stream latency to
|
||||
// avoid accumulating packets in queue-based backends
|
||||
// like WASAPI, ALSA, and PulseAudio.
|
||||
//
|
||||
// This bound was set by testing on several Windows machines.
|
||||
// The highest latency was found on a XPS 9343 running Windows 7
|
||||
// in Steam Big Picture and the 5.1 audio test clip from Fraunhofer.
|
||||
if (me->m_SoundIo->current_backend != SoundIoBackendCoreAudio && me->m_SoundIo->current_backend != SoundIoBackendJack) {
|
||||
double latency;
|
||||
if (soundio_outstream_get_latency(stream, &latency) == SoundIoErrorNone) {
|
||||
if (latency > 0.050) {
|
||||
// If our latency is higher than desired, drop these samples to gracefully lower
|
||||
// the latency without glitching too much. Dropping the whole buffer causes
|
||||
// a much more noticeable glitch. This approach also ensures that we don't
|
||||
// accidentally underflow if the driver/kernel side is delayed and isn't
|
||||
// consuming data fast enough. Dropping a frame at a time and re-evaluating
|
||||
// each time ensures that we'll stop dropping if latency returns to normal.
|
||||
readPtr += framesLeft * stream->bytes_per_sample * me->m_OpusChannelCount;
|
||||
bytesRead += framesLeft * stream->bytes_per_sample * me->m_OpusChannelCount;
|
||||
framesLeft = 0;
|
||||
// Clamp framesLeft to frameCountMin to ensure that we never write more than one sample.
|
||||
// This makes sure that we never increase our latency beyond what the sink is consuming.
|
||||
framesLeft = qMin(framesLeft, frameCountMin);
|
||||
|
||||
SDL_LogInfo(SDL_LOG_CATEGORY_APPLICATION,
|
||||
"Latency exceeded drop cap: %f",
|
||||
latency);
|
||||
}
|
||||
}
|
||||
// Track latency on queueing-based backends
|
||||
if (me->m_SoundIo->current_backend != SoundIoBackendCoreAudio && me->m_SoundIo->current_backend != SoundIoBackendJack) {
|
||||
soundio_outstream_get_latency(stream, &me->m_Latency);
|
||||
}
|
||||
|
||||
for (;;) {
|
||||
|
|
|
@ -32,6 +32,7 @@ private:
|
|||
struct SoundIoOutStream* m_OutputStream;
|
||||
struct SoundIoRingBuffer* m_RingBuffer;
|
||||
struct SoundIoChannelLayout m_EffectiveLayout;
|
||||
double m_Latency;
|
||||
bool m_Errored;
|
||||
|
||||
static const double k_RawSampleLengthSec;
|
||||
|
|
Loading…
Reference in a new issue