WIP: Cmake refactor
What is wrong with the way things currently work:
- It is a hassle for the external projects.
- Rather than automatically pulling in the configuration parameters (such as include/library locations) already determined with the main VTK-m build, you must specify them all over again. And if you mess it up, things can go haywire.
The new way the build works:
-
Rendering components now building and usable. The build is now re-organized, and now the developer explicitly chooses what rendering backend they want ( None/GL/Mesa/EGL). This has simplified the build options and makes it easier for consumers to determine how VTK-m was built
-
Backported the CMake 3.10 rewrite of FindOpenGL and added it to VTK-m. This allows us to sanely choose GL versus Mesa versus EGL
-
Relaxed the minimum CMake version to 3.3!!!! Now you only need CMake 3.8/3.9 for CUDA