Inside the Message Passing Interface

Chapter 2 Shared Memory

In this chapter, we will create a simple MPI subset capable of basic blocking synchronous point-to-point communication over shared memory. While not being a complete MPI implementation, this subset will take a sensible first step toward the complexities expected in the following chapters, and at the same time set a useful reference point in performance, since shared memory is certainly the most sensitive and capricious of all communication fabrics we will deal with.

2.1 Subset Definition

Let’s make our first MPI subset very simple. Shared memory is lightning fast, and the simpler the subset, the lower latency we will be able to achieve. To this end, we will:

–Provide the basic MPI functionality:

• The initialization and ...