Thursday, August 26, 2010

Random WebGL related gunk

There's a few things I've come across in the last few days that I feel are worth pointing out to the WebGL community:

  • Newer builds of Firefox have started validating the textures passed to generateMipmaps according to the OpenGL ES 2.0 spec. This means that attempting to generate mipmaps with non-power-of-two textures will start failing on you very soon! Please update you code/resources accordingly. (This already bit me with my Quake demos, which have now been tweaked accordingly.)
  • WebGL___Array types are going, going, gone! Newer browser builds don't even have them defined, so make sure you are using the appropriate analogs from the Typed Arrays. Again, my demos have been updated to account for this.
  • In an odd twist, whereas Firefox has been speeding up lately apparently Chrome is cracking down on crazy fast loops now. With the latest dev releases (7.xx) I've noticed that my Quake 3 demo is basically locked at 30fps, regardless of whether or not you check "Cap FPS". Curiously, this does not seem to be the case for my Quake 2 demo, which means that Chrome is actually limiting the internal message pump. Granted, it's a very stable 30fps, and I'd rather have a lower but stable framerate than a high but erratic one, so I'm not complaining. It did strike me as an interesting change, though. A bit of background on this one: For the Quake 3 demo in order to get the best framerates possible I used a hack picked up from Vladimir Vukicevic that continually posts messages to the page as a signal to draw (demo), which attempts to avoid some of the resolution limitations of setTimeout/Interval. This apparently causes severe performance issues if allowed to run rampant, however (as commenters on this site have pointed out) and so it's possible that this apparent capping may be Google trying to prevent abuse of the system. (Which is probably a good thing!) I'd love to know a bit more about the actual changes that took place in this regard, but in the end it appears that the timer-based method is the better way to go.
  • Finally, in a bit of news that has nothing to do with WebGL whatsoever, I find this to be supremely amusing. Chrome envy, perhaps? :) It's especially funny in light of some Google-bashing that Microsoft did a few months back about how a single address/search bar was a bad thing for users privacy.

No comments:

Post a Comment