Also see under the help menu when you run SKA.
Also there is a section in openCL Programming Guide section 4.1
Is there any thing in particular you want to learn about them?
I will look at the OpenCL programming guide first. This is my first time use these tools. I just confused how to measure the performance of my OpenCL program. In the matrix multiplication example, there are setup time, kernel time, and total time (measured in secs.) in the main program. But also there is a time measurement inside the runKernels() which use CL_PROFILING.... (measured in ms). Usually which one to use for comparison?
Both can be used.Although clGetProifilingInfo is a standard OpenCL API which can be used to determine the submit time launch time of any API you want in a very easy and efficient manner.You can refer to openCL spec for more details about clGetProfilingInfo.