The C++ AMP code that you want to run on the accelerator is specified as an argument in a call to the parallel_for_each method. You can provide either a lambda expression or a function object (functor ...