How does arrayfun specifically implement GPU computing?

I would like to better understand and analyze the performance of how MATLAB through arrayfun does necessary computations using the GPU. Does arrayfun do the computing exactly as CUDA? Does MATLAB follow a default configuration in allocating operations to threads and other necessary settings since the user has less flexibility in using arrayfun compared to MEX CUDA functions?

Also, instead of tic/toc or gputimeit functions, is there a more detailed way to profile the GPU's execution in the hardware (number of computations, operation allocation to threads, etc) when using arrayfun?