Pixar and DreamWorks have announced the availability of v3.0 of the set of open source libraries that implement high-performance subdivision surface evaluation on CPU and GPU architectures. New features and improvements include:
- Faster subdivision using less memory (performance improvements of up to 4x on GPUs, 10x on CPUs)
- Support for non-manifold topology
- Face-Varying data specified topologically
- Elimination of fixed valence tables
- Single-crease patch for semi-sharp edges
- Additional irregular patch approximations
- Introduction of Stencil Tables
- Faster, simpler GPU kernels
- Unified adaptive shaders
- Updated coding style with namespaces
- More documentation and tutorials
More on the OpenSubdiv project’s page.
Would be really nice if this would get implemented in the next update to Max(or any other package for that matter) without any fuss, I hope we don’t have to make another request for what would be simply a logical course of action.
Blender 2.76 will contain the new OpenSubdiv 3.0.