The Cluster Edition is intended to boost performance of clusters utilizing the Intel hardware. It includes all the components available in the Professional Edition plus the Intel® MPI Library and the Intel® Trace Analyzer and Collector—an MPI profiler.

Intel® MPI Library

Intel® MPI Library is a scalable performance library implementing the latest MPI standard. It is designed to perform on the Intel® microarchitecture processors and interconnects.

Using the Intel MPI Library is the first step in tuning your MPI application, as it offers competitive performance out-of-the-box without the need to manually configure the library. It can work with multiple interconnects and can automatically select the most appropriate communication mechanism for that interconnect. Since the Intel MPI Library is ABI-compatible with MPICH, switching to the Intel MPI Library is easy—no recompilation is required. For applications based on other MPI implementations, there is a set of compilation scripts that simplify recompilation.

Another advantage of the Intel MPI Library is its built-in analysis facilities and integration with other analysis tools. It can collect detailed application statistics at runtime and present it in a human-readable format. The mpitune utility is an automatic tuning tool that helps you adjust the library settings for the given cluster or application. On Linux*, the runtime options, -mps and -trace, help you easily analyze the application with low overhead, using the MPI Performance Snapshot and Intel® Trace Analyzer and Collector. On Windows*, the compile-time option, -trace, is available. For other tools, like Intel® VTune™ Amplifier, Intel® Advisor and debuggers, the Intel MPI Library provides a special option enabling you to target specific processes.

Intel® Trace Analyzer and Collector

The Cluster Edition also introduces the Intel® Trace Analyzer and Collector, the tool for profiling MPI applications. One of its major features is the MPI Performance Snapshot (MPS) tool, which serves as a starting point in MPI analysis. It provides essential performance information, exposing the application weakest spots, and also gives further guidance on the analysis.

While MPS provides a general performance overview, Intel Trace Analyzer and Collector collects all the details and visualizes the application flow. Various charts and diagrams help you identify weaknesses and bottlenecks in the communication pattern.

In addition to the information about MPI, these tools can analyze the application OpenMP* regions, giving you an idea about OpenMP threads effective use and load balancing.

Para obtener información más completa sobre las optimizaciones del compilador, consulte nuestro Aviso de optimización.