So I may start this one.
Trying to make an in engine time profiler that monitors each steps of each frame on the cpu and on the gpu.
It's not yet perfect, because some gpu time is not monitored but I'm looking into it.
For example here is the TestSSAO2 test scene with a lot of geoms and the ssao filter with approximated normals.
Don't mind the 19 fps, the screen capture software is killing it.
EDIT: oh and left value is CPU time right value is GPU time