Profiling OpenGL with nVidia drivers

[gmueckl]

I'm writing a game engine with quite unusual OpenGL usage characteristics: the engine is often updating a couple of big textures. This seems to be the main bottleneck. However, the engine turns out to run too slow and I'm at a loss as to what the reason really is as opengl calls don't show up at all in gprof output. So I cannot compare the times my own code takes to execute to that of the OpenGL calls I use. I currently suspect that the driver developers have put some nasty tricks in their lib to prevent that (gdb can't deal with that libGL.so properly, either).

Does anyone know of a trick to make the OpenGL calls appear in the profiler output?