Hi Miroslaw,
I think I can reproduce the issue - I see a very large obviously incorrect number of cycles reported on G2xx parts, where when I run the same code on an F5xx part I see the correct number of cycles reported. It sounds like this issue: http://e2e.ti.com/support/microcontrollers/msp430/f/166/p/293077/1021885.aspx#1021885
I'm going to follow up to see if Walter ever heard back from the CCS team on this issue.
This is different than the original CCSv5.1 issue at the top of this thread though (in that case the clock did not count at all). Can you confirm that this is what you are seeing - the clock starts counting but returns bad (extremely large) values? If not, please provide more detailed information like what code you are using and where you put the breakpoints.
Regards,
Katie