At the bottom is the video I recorded of Anholt's presentation about the Intel Linux graphics driver developed out of their Open-Source Technology Center. Here are some of the keynotes from his FOSDEM 2012 presentation:
- As far as what's next for the Intel Linux graphics team is they will be focusing upon more performance improvements, increased stability, and MSAA.
- The proper multi-sampling anti-aliasing (MSAA) support for the Intel Linux driver is finally on the horizon! MSAA will become more important with the greater usage of cairo-gl.
- Anholt reiterated that APITrace is awesome and offers lots of features, plus more capabilities are coming up. APITrace is very useful for OpenGL debugging, finding driver issues, performance monitoring, and more.
- The Intel Glamor acceleration architecture (using OpenGL acceleration for 2D) is still being worked on. Glamor is still slower than using the CPU-based software rasterizer, but it's faster than the current (UXA) 2D acceleration in the xf86-video-ati DDX. (Phoronix benchmarks of Glamor acceleration are still on my TODO list.)
- The SNA acceleration architecture continues to be very promising and Chris Wilson is still working extensively on it. SNA is more than 70k lines of code (this is just for 2D acceleration, their entire OpenGL driver in comparison is around 40k lines of code). SNA can be faster than using software acceleration and overall is very promising, but continues to be disabled by default. (Phoronix benchmarks of SNA continue to show much progress.)
- As far as what will succeed UXA (either Glamor or SNA), it's still being decided. The problem with SNA is that new back-ends need to be created for each new generation of Intel graphics hardware while Glamor is just going to be using the OpenGL code-paths. However, for whether a switch is made to Glamor or SNA, replacing UXA will likely be gradual. The Intel developers will slowly replace the various UXA operations with the new SNA/Glamor calls gradually. When switching from EXA to UXA with the Intel driver it was immediate and initially caused some nasty fallout for users; Intel developers are being careful to not cause great pain again with more invasive changes all at once.
- Cairo-gl is coming along nicely as well.
- The libdrm component has received some useful improvements recently for the Intel driver developers like the AUB file dumping and batch-buffer decoding.
Below is Eric's FOSDEM presentation in full about the Linux graphics user-space. There's also some other exciting Intel Linux news that wasn't publicly brought up during the presentation but will be mentioned soon.