Parallel Programming with Intel® Parallel Studio XE

Chapter 3

Parallel Studio XE for the Impatient

What's in This Chapter?

An overview of the four-step methodology for adding parallelism

Using Cilk Plus to add parallelism

Using OpenMP to add parallelism

The previous chapter introduced three ways of using Intel Parallel Studio XE: Advisor-driven design, compiler-centric development, and a four-step methodology.

This chapter describes the four-step methodology for transforming a serial program into a parallel program. The chapter's hands-on content guides you through the steps to create a completely parallelized program.

In the examples in this chapter, you use two parallel models, Intel Cilk Plus and OpenMP, to add parallelism to the serial code. Cilk Plus is regarded as one of the easiest ways to add parallelism to a program. OpenMP is a well-established standard that many parallel programmers have traditionally used.

You use various key components of Intel Parallel Studio XE to achieve the parallelization. This chapter describes how to use Intel VTune Amplifier XE 2011, an easy-to-use yet powerful profiling tool, to identify hotspots in the serial application, as well as analyze the parallel program for synchronicity, efficiency, and load balancing.

You use Composer XE to build the newly parallelized application, and then use Intel Inspector XE 2011 to reveal threading and memory errors. Finally, you return to Amplifier XE to check for thread concurrency and fine-tuning.

The Four-Step Methodology

Initially, parallelizing a serial ...