rapidobj VS SpeciaLUT

I wrote a rapidobj library for parsing Wavefront .obj files. It's an old text format for 3D data. This library was optimised to quickly process large files (see benchmarks); it can parse millions and even tens of millions of triangles per second.

I just released rapidobj v1.0. It's a fast .obj file parser. May be useful to those who need to quickly load large .obj models (of course, you will still need to transform the data into something that's GPU friendly). Third party benchmarks: https://aras-p.info/blog/2022/05/14/comparing-obj-parse-libraries/

I made SpeciaLUT to convert bool/enum runtime tests to compile-time conditionals — by compiling all branching combinations in hot functions and saving them in a lookup-table, so the optimal one can be called at runtime. Reason: as an HPC consultant I encountered many codes that grew without good architecture, in which features would just be added and branching would propagate through all levels. This yields 10% to 50% performance increase in such codes.

I have implemented this as a C++ library, where you extract const states as template parameters, and the library compiles all specializations and allows you to choose the optimal one at runtime.

So here's my C++20 implementation of the above: SpeciaLUT