problem with cudaprof

When I use the cudaprof, I find that the number of gld and gst are not correct. In fact, it is very baffling.

Does anyone has this experience?

The version of cudaprof is 11.0 and the OS is OpenSUSE 10.3

Do you have a GTX260/280/285/295 ?

I am using 9800 GTX+.