Intel® MPI Library Development Kit

Published By:

Intel

The Intel MPI Library implements the MPI-2 specification across multiple fabrics to enhance application performance on clusters using Intel processors. In addition it provides an abstraction layer than enables you to modify cluster interconnects without having to make major changes to the software. The high performance multifabric message-passing library works on either Windows or Linux and delivers ultimate high performance computing capabilities.

Product Variants

Intel® MPI Library Development Kit for Linux

Intel® MPI Library Development Kit for Windows

New Features

Enhanced Performance

Optimised shared memory path for multicore platforms allows more communication throughput and lower latencies
Native InfiniBand interface (OFED verbs) support for lower latencies
Multi-rail capability for higher bandwidth and increased interprocess communication
Tag Matching Interface (TMI) support for higher performance on Qlogic PSM and Myricom MX interconnects

Improved usability

Updated MPI tuner to optimise Intel MPI library options for faster applications
MPI 2.1 standard conformance for compatibility
Backward compatibility with Intel MPI 3.x based applications to re-use existing implementations
Experimental fault tolerance support for more robust cluster applications
Experimental scalable process startup to shorten the startup phase

And more

Greater scalability with connectionless DAPL UD support
Application support in a wider range of Intel Cluster Ready environments

Features

Performance

Get high-performance interconnects, including InfiniBand, Myrinet, as well as TCP, shared memory, and others
Efficiently work through the Direct Access Programming Library (DAPL), making it easy for you to test and run applications on a variety of network fabrics
Get users up and running faster with the ability to install under root or through an ordinary user ID
Use the provided mpivars.sh and mpivars.csh shell scripts for easy environment setup
Reduce hand-coding work by using the mpirun script, which automates multiprocessing daemon (MPD) startup and cleanup
Take advantage of flexible system-, user-, and session-specific configuration files
Give the end user a reliable runtime with transparent support for fallback Internet Protocol (IP) interfaces
Increase performance with the ability to use device-specific and collective-protocol thresholds
Boost performance with memory registration cache
Get more accurate measurements with platform-specific finegrain timers

Compatibility

Based on Argonne National Laboratory’s MPICH-2 implementation
Simplified Integration with leading Linux job schedulers
MPI-2 standard compliance and portability
Support for ROMIO (a high-performance, portable MPI-IO implementation)
Support for leading Linux Parallel Debuggers
Support for GNU compilers