gmbeard / shadow-cast

GPU accelerated screen and audio recording for Linux
GNU General Public License v2.0
39 stars 0 forks source link

feat(audio) Uses memory pool for audio data to reduce allocations #3

Closed gmbeard closed 1 year ago

gmbeard commented 1 year ago

This increases the throughput of the audio capture stage. Because of this, I've also had to change the audio encoding to a timer based approach instead of using notifications from the audio capture thread. This is because Pipewire can sometimes deliver small chunks of audio very quickly and this was saturating the processing loop, causing high CPU usage.