Fixed a few bugs with the Physx implementation on our engine. It's one of those things where you fix one thing, and it reveals about ten other things you're doing wrong.
I really like version 3.2, it seems to fix everything I didn't like about 2.7. Though I still hate using programming API's, and use them sparingly. If a system isn't hardware implemented I have a hard time justifying its utility.
First off, there's a learning curve. And it takes a long time to really know a system. This includes spending hours on a bug that ends up being an API system that you're using wrong. It's just much faster to fix your own code (unless the code base it huge and you didn't design/document it well.)
Second, upgrades can be a pain. Often there are whole sale changes, which repeats the above. Pretty much everyone was touched in the Physx upgrade. It was needed, it is appreciated, it's just time consuming to redo.
Third, API's often cost money. Physx is free for what we're doing (at the moment) so that's not a concern. But other API's are not, and given the first two that makes the utility of using them questionable. Take Granny for stance (a skinning API.) Are bones really that hard, now that information to do them is all over the place? It seems all the work was really getting animation info our of whatever 3d animation system you're using. If what 3ds MAX did made more sense (and maybe it will with later releases) this will be even less of an issue.
We use OpenGl, OpenAl and Physx. I really don't want that list to grow. Each one has a hard ware implementation and in all fairness I think they're all robust enough for any need.
Okay, less ranting. More pretty pictures later.
No comments:
Post a Comment