Media decode and composition: bridging the gap

This proposal has been accepted as a session.

*

One Line Summary

A look at the issues preventing better usage of media decode and 2D composition hardware

Abstract

A lot of graphics hardware, particularly in non-desktop media-led cases such as set-top boxes, has dedicated hardware for both media decoding (e.g. MPEG4 to raw YUV), as well as 2D composition. This hardware is much more capable than is exposed through general purpose APIs: for example, overlays give much higher-quality scaling and filtering than using EGL to do the media composition, also at higher performance.

Unfortunately the state of the art for a zerocopy media pipeline involves OpenMAX, passing physical addresses around through userspace, and a huge amount of platform-specific hacks. This talk will cover the pipeline in general, the gaps and issues facing us, and some potential ways forward with both new and existing frameworks.

Tags

gstreamer, V4L2, drm, kms, graphics, media, openmax, wayland, x11

Presentation Materials

slides

Speaker

Leave a private comment to organizers about this proposal