As you know, radeon-rewrite, with all its KMS & co. goodiness is being developped. And as much as we're trying to make things work out well, the switch will be painful for some people. So I thought to myself, why not try to sweeten the deal a bit by seriously setting a performance goal?
(Caveat: I don't have much time to work on this - my involvement is a one-afternoon-per-week kind of involvement.)
Now I'm a person who likes to have verifiable results and tests. Sure I can try to make things faster and get fuzzy "feels better" results, and in the end, but I'd also like to have some objective metrics to back me up.
I've used the PTS (thanks, by the way, it's a great tool!), but I haven't found a test or test suite that would hold up as a useful metric for Compiz desktop performance.
There are RENDER tests, and OpenGL tests, but I'm not sure what a good indicator for Compiz-related performance would be. And if there is none, take it as a hint that such a test would be a great feature for future versions!
Any input on this topic would be highly appreciated.
I would love to see a wonderful Compiz benchmark and would be happy to push it into PTS. However, at this time I am not aware of any real world ones. There is the FPS counter plug-in for Compiz, but that doesn't work out too well for automated tasks and then also needing to automate the window movement and such while managing what Compiz plug-ins are used, etc. I haven't tried writing my own pure Compiz benchmark, yet.
Anyone else aware of any Compiz tests that may be out there?
Michael: Thank you for the feedback. I guess for the time being I will just run a couple of glxgears, add the compiz benchmark plugin, and let the glxgears wobble. Right now this use case has enough interesting problems as it is. In the long run, a real automatic benchmark of (to put it desktop neutrally) composited desktop performance would be awesome.