Which OpenCL APIs load kernels to FPGA?

The clCreateProgramWithBinary call configures the FPGA; you can put a printf before and after it and look for the "Reprogramming device [0] with handle 1" line in stdout and its position compared to the printf outputs to make sure. If you are measuring time, you should make sure to use clFinish() before reading the end time because many of OpenCL's call functions are non-blocking. Obviously, unless you call the clCreateProgramWithBinary multiple times, the FPGA will not be reconfigured multiple times, and you can call clEnqueue with the same kernel as many times as you want after the initial configuration.