I'm getting strange errors wheny trying to profile a program which uses clAmdFft. Please see attachment for console output of CodeXL. The error seems to come from the definition of a float value in the clAmdFft kernel where comma is used instead of point. I checked the chached kernel and indeed, commas are used instead of points. This problem didn't occur on my previous ubuntu installation and the code works fine without CodeXL.
Can it be caused by my langauge settings or so? In Germany we use comma instead of point. Maybe the part of clAmdFft that generates the kernels adopts this from the system settings. But why it then works without CodeXL and doesn't with it?