Quote:
Originally Posted by Shadow
Some day I'll hopefully have time to do some bigger optimization work like glsl, moving vertex work to GPU, and much more multi-threading.
|
I don't think these things are alike. GLSL isn't a huge transition, and once you get there, adding transforms into them isn't that hard either. In general, if anything fails, it'll be immediate and fast.
Multithreading, on the other hand, is a huge undertaking. There *will* be many subtle bugs. The safest way to do multithreading in C++ is probably
OpenMP: when you get to a very large computation in a loop, farm it out to other threads using a simple #pragma. Getting a real performance improvement though will take a lot of work.