X.Org's GLAMOR 2D Performance Continues To Be Tuned

While GLAMOR has already been around for a number of years as a means of providing generic X11 2D acceleration over OpenGL for the X.Org Server, it's a seemingly never-ending process to optimize its code-paths for best performance. More improvements are en route for making GLAMOR 2D faster, which should especially be helpful for Raspberry Pi users making use of the VC4 driver stack on this very slow-speed hardware.

Benefits to the GLAMOR code in the X.Org Server obviously have the potential to benefit all users of this acceleration mechanism for code going into the xorg-server code-base as opposed to an individual GL driver, but for Raspberry Pi users in particular there is some efforts ongoing by Broadcom's Eric Anholt as well as Keith Packard's never-ending tinkering with the X Server code. GLAMOR continues to be used by default for all AMD GCN GPUs, Nouveau for the latest generations of GPU too, VC4 2D is only supported with GLAMOR, and optionally by other DDX drivers too.

At last week's XDC2016 conference, Eric Anholt was talking with Keith Packard and others about further tuning on GLAMOR. What they decided was hurting their rendering performance currently is excessive flushing and lack of bounds on rendering operations.

Fixes for these are in the works by limiting how often glFlush() is called rather than flushing the OpenGL command stream every time. Bounding of the rendering operations is more difficult but work is going on in that direction along with some underlying improvements to GLAMOR of const-ifying all op arguments, etc.

Excited Raspberry Pi users over the prospects of faster 2D can read Eric's blog post for all of the details.