I am trying to do simple mathimetical operation - Log2. For this I call log2 function in my kernel. As a result I get - CL_BUILD_PROGRAM_FAILURE, when building an Opencl kernels on my GPU. When I start this programm on my CPU it works fine without any problems. Maybe I mast include some extentions for GPU? I use: nVidia Tesla M2090 and latest cuda drivers.
What is the compiler output? you can get an actual compilation error message and not just a flag. It would help to see what the compiler thinks is wrong.
The problem was in using invalid overloaded function.