Howto find out which primitives/render-operations are causing a slowdown?

[Clemens Eisserer]

Hi there,

I am currently trying to analyze why some programs perform so weak
with the proprietary nvidia driver and it turned out that it seems for
some operations the nvidia driver falls back to  software rendering.

Does anybody know a way how I could find out which RENDER/xlib calls
are that problematic without having deep insights to the analyzed
programs?

Maybe recompiling xlib/librender with debugging symbols, forcing xlib
into synchronous mode and using oprofile to analyze? Would oprofile
report the waiting as used cpu-time?
Any others/better ideas which do not mean recompiling xlib/libxrender?
Are there some protocoll-analyzers which could be used for this task?

Sorry for bothering and thank you in advance, lg Clemens