Classic HPC Developer Using Visual C++ Tutorial
< Go Back
This tutorial presents a classic HPC development scenario of data parallelism in the context of image processing. This scenario is explored using both shared-memory (OpenMP) and distributed-memory (MPI) techniques. Written for the C and C++ software developer, this tutorial walks you through the steps of designing, writing, debugging, and profiling a parallel application for Windows HPC Server 2008. When you complete the tutorial, you will have the skills and expertise necessary to deliver high-performance, cluster-wide applications for Windows HPC Server 2008.
The tutorial includes lab exercises, program solutions, and miscellaneous support files. It begins with the problem domain, and then discusses data parallelism and contrast stretching. It then presents a sequential version of the contrast stretching application, and steps you through a lab exercise. The tutorial continues with a look at working with Windows HPC Server 2008, accompanied by a second lab exercise.
The tutorial then takes a look at parallelizing the contrast stretching algorithm for execution on a shared-memory machine, which can be your local workstation or a single node in the Windows HPC Server 2008-based cluster. Following that, the tutorial steps through a distributed-memory parallel version using MPI, which involves installing and configuring Microsoft Message Passing Interface (MS-MPI). A lab exercise involving MPI debugging, profiling, and event tracing is next, and the tutorial concludes with a look at hybrid OpenMP /MPI designs and managed solutions with MPI.NET.
