I have a Radeon 6970 and running some OpenCL applications (under Linux). Things went fine but when I tried to profile a program using sprofile, it is slowed down 15 times (from 60seconds to 900 seconds!). It is the same even I profile only one event (performance counter).
I used profiler version 2.5. I think I used Catalyst 13.11 beta but I'm not so sure because I don't know which way to check this information.
Any help would be appreciated
Where exactly are you exactly seeing the degradation in performance? I think the profiler does add some instrumentation code which will increase the overall application time. But you should not see any difference in the kernel execution time.
Did you try the latest CodeXL 1.3 for profiling? http://developer.amd.com/tools-and-sdks/heterogeneous-computing/codexl/#one