Intel's Gallium3D Driver After Google's Work

[89c51]

I'm primarily interested in seeing Intel chips move to Gallium due to the abundance of useful state trackers that are in development. Once Clover, OpenVG and the GL3 state tracker start to mature, being able to run them on my 965GM (with features not supported by the hardware running through optimized LLVM bytecode) would be very nice. No Can Do with i965c though, and i965g isn't really being worked on. i915 users should consider themselves lucky to get this nice boost despite the hardware being really, really ancient (even my 965GM is old, and it isn't getting that degree of work).

in order to use the state trackers the underlying hardware must support the function right???? ie if i915 hw doesn't support opengl 5 or OpenCL 3 the state tracker supporting those won't be able to do much right??


ps
versions numbers are fictional & used for illustration purposes in order not to create confusion with the current hw capabilities

[devius]

I use the latest ubuntu with the default movie player (totem), I also tried VLC no better result.

I try to play 720p movies with x264 codec.

I'm using Ubuntu on my Eee PC 1000H and I'm in the same situation. Even low bitrate 720p movies are practically unplayable, however on Arch they used to play almost perfectly (maybe 1-2 dropped frames every 10 seconds or so) and I have no idea why. The kernel I used back then was the same version of the most recent Ubuntu. Also tried VP8 encoded file and it's the same situation.

In regards to marcheu's work it's amazing to see this kind of performance on something that hasn't been touched for years until recently! Nice work and, not that it matters, but you have my vote on the performance optimizations business I just want to be able to play Neverwinter Nights on this piece of crap 945gm.

[chithanh]

For those wondering about five-day-old-benchmarks... it's because I had to get them done in advance and build up a publishing queue in advance as I was away for 3~4 days due to needing to meet with some of Valve contacts in Washington, as mentioned on my Twitter feed.

That's cool and it was still the right decision to publish this article. But better journalism is to explicitly mention what has changed in the i915g driver between the time the benchmarks were done and the article was published, e.g. the 500 lines of changes that marcheu mentioned. Even better if you make an estimate about which of your benchmarks are affected by the changes to what extent.

[bridgman]

It sounds as if the article was written right after running the benchmarks, before the subsequent changes happened.