Part II — The Parallel Execution Layer
This part focuses on the mechanisms used to execute parallel programs on supercomputing systems. Through classical programming models such as MPI and CUDA, it introduces the fundamental concepts of parallel execution, communication, synchronization, and acceleration.
Part II serves as a natural continuation of the infrastructure concepts introduced earlier, but it can also be approached selectively. Readers whose primary interest lies in deep learning frameworks may treat this part as contextual background, while those with a strong HPC orientation will find it essential for understanding performance behavior and execution models used later in the book.