- Prepare flow for multithreading support
- Optimize memory access of vertices
- More efficient back-face culling computation
- Added option to disable gprof in Cmake
AVX datas aren't correctly aligned in stack in following case:
- using inline functions (GCC not always inline function),
- not using reference to AVX data (copy of data in stack dynamically),
- not using O2 or O3 optimization flags (related behaviors of the previous points).
