Released testing 240923-334ce8:
- [Pi5] Optimized H264 software decoder yet again by only using a single pooled DMA buf per video frame.
- [Pi4/5] Provide executable name to content-process. This can be used to improve content process start up time by avoiding to load the content process into an overlay if configured in node.json.
- [Pi4/5] Prevent content process content from being drawn prior to being marked ready.
- [Pi4/5] Allow raw video players to transition to gl player. This can allow fancy transitions by switching from faster raw playback to gl during transitions like rotations.
- [Pi4/5] The same mechanism can now seamlessly mode rendering from DRM to GL and vice versa on demand or when required by how video layers are stacked. This can help render more videos at once without using too many DRM planes. This can also be used while taking screenshots to avoid having to double the active DRM planes as would be required without this feature.
- [Pi4/5] Reduce CPU usage by avoiding syncing of video frames if they are not being drawn.
- [Pi4/5] Use optimal DMA heap depending on Pi model: system heaps on Pi5, CMA on Pi4. This also means that there isn’t any real memory restriction on how many videos can be loaded on a Pi5.
- [Pi3] Fixed the screenshot problem from the previous release. The later Pi3 models should now also use the Pi4/Pi5 OS method of video rendering, including content process features like embedding a live browser. Get in contact if you’re interested in testing this.