
Videos are forthcoming, but Nicolai's technical talk did produce a TODO list for the R300 GLSL compiler. While most GLSL shaders should work today, as proclaimed Nicolai, there still is missing flow control in vertex programs, loop support in fragment and vertex programs, support for additional instructions, and there is an immense amount of optimization work that can be done to Mesa and the OpenGL drivers.
Daniel's talk was on how users expect "every frame must be perfect" and some of the current problems include issues with RandR reconfiguration, video display programs, server implementations are awful, video tearing is common, and window reconfiguration is brutal. Daniel also briefly commented on the Wayland Display Server, but as he said, "X is the best since everything else doesn't work or doesn't exist." Daniel thinks someday Wayland might function according to him.
6 Comments