The Next Generation OpenGL Initiative isn't about an API cleanup, rewrite, tooling, benchmarking, or profiling. Those are all areas that could use improvement, but a standards body is
not the place to do those things. Trying to do those things in a standards body is a recipe for logjams and delayed and bloated standards. "Longs Peak" is a prime example actually.
This initiative is about one major thing: reducing overhead. It's a response to Mantle/DX12 and their low CPU overhead rendering capabilities. OpenGL risks being left behind performance-wise because drivers are too complex. Because the API hasn't evolved as fast as the capabilities of modern hardware changed, drivers today do too much automatic work "behind the scenes". Things like memory allocation, texture swizzling, parameter validation, and shader recompilation often happen implicitly, outside of the control of the application. This adds unexpected CPU overhead which varies per driver.
Mantle/DX12 change the semantics of the API to match what modern hardware actually does. Applications have more control, drivers do much less implicit automatic work and CPU overhead is greatly reduced. That's what the "Next Generation OpenGL Initiative" is hopefully going to do for OpenGL.
This initiative is about one major thing: reducing overhead. It's a response to Mantle/DX12 and their low CPU overhead rendering capabilities. OpenGL risks being left behind performance-wise because drivers are too complex. Because the API hasn't evolved as fast as the capabilities of modern hardware changed, drivers today do too much automatic work "behind the scenes". Things like memory allocation, texture swizzling, parameter validation, and shader recompilation often happen implicitly, outside of the control of the application. This adds unexpected CPU overhead which varies per driver.
Mantle/DX12 change the semantics of the API to match what modern hardware actually does. Applications have more control, drivers do much less implicit automatic work and CPU overhead is greatly reduced. That's what the "Next Generation OpenGL Initiative" is hopefully going to do for OpenGL.