tools to understand Altera OpenCL scheduling

The kernel does not have multiple threads. If I remember correctly the profiler can tell you something about where the kernel spends its time. But if you mean host threads, you can use a profiler for your os.

Not the host threads. But threads in the sense of the each new iteration launched into the pipeline. Is there any tool for how long it took for the entire pipeline to be executed once?

I believe this is the profiler, It has been a couple of months since I used it, and remember it sometimes crashed, but it should do what you want. See Profiling Your OpenCL Kernel in the Programming Guide.