Hello,
I am a Windows C/C++ and also at times, assembly language Software Engineer.
The correct way to specifically target Intel Phi co-processors, would be to use Intel Compilers and OpenMP like libraries, to specifically target your program/code at the Xeon Phi coprocessor.
The code would probably involve custom C/C++ like languages that Intel Phi Compilers come with.
This type of high-performance super computing is from the realm of multi-cpu multi-processing (SMP, MPP, PPP) and requires understanding of multi-processing and super-computing concepts. :)