1. Develop, test and distribute MPI applications for all major cluster configurations with a single MPI library.
2. Get a high performance implementation of the MPI-2 Standard.
3. Obtain portability with the most general multi-fabric support for all major network configurations.
4. Incorporate future network-fabrics without changing applications with a standard multi-fabric interface [DAPL].
5. Provides flexible support of multiple interconnects, e.g. InfiniBand*, TCP/IP, and 'Shared Memory' on advanced multi-core and shared-memory processor (SMP) configurations.
6. Easy to install, and comes with diligent documentation and reliable support.
7. Supports most major Linux* platforms and a range of compilers.
8. Invoke a parallel debugger of choice easily.
9. Benefit from excellent tuning tools through an enhanced performance analysis interface called Intel® Trace Analyzer and Collector.
10. Get a free runtime environment kit, available for pre-installation or redistribution.

• Provide an accelerated multi-fabric layer for fast interconnects via the Direct Access Programming Library (DAPL) methodology (Figure 1).
• Supports TCP, shared memory, and many DAPL-based interconnects, including InfiniBand, Myrinet*, and others.
• Utilize a fast shared-memory lane and a sockets fallback, when appropriate. Execution failure can be avoided even if interconnect selection fails. This is especially true for batch computing. For such situations, the sockets interface will automatically be selected (Figure 1) as a backup.

• Smart fabrics selection — Simplified usage by automatic choice of fastest transport protocol between MPI processes without additional environment settings. Fully configurable device and fallback selection by environment variables.
• Enhanced dynamic connection establishment — Reduced memory footprint by introducing lazy mode connection, which only establishes connections when needed.
• Two-phase communication buffer enlargement — Reduced memory consumption by allocating only the necessary communication buffer memory space.

• DAPL* intra-node communication mode — Bandwidth advantage for large messages by optional use of DAPL inside a multi-core or SMP node
• Further optimized collective operations — Significantly optimized versions of MPI_Reduce, MPI_Allreduce, MPI_Alltoall, MPI_Alltoallv, MPI_Broadcast, and others.
• Intelligent process pinning — Maximized performance on Intel multi-core and SMP nodes by automatic process pinning without additional environment settings.
• Scalable job startup protocol — Significantly improved application startup time with this release.
• Static version of libraries built without -fpic flag — Better performance for statically linked applications.

• Additional thread safe libraries at level MPI_THREAD_SERIALIZED — An MPI application process can be multi-threaded and multiple threads may make MPI calls, but only one at a time.
• Backward binary compatibility with Intel MPI Library v2.0 — Applications and objects compiled with Intel MPI library 2.0 will work with run-time library from Intel MPI Library 3.1.
• Enhanced handling of multi-homed environment — Simplified job and communication management and increased performance when working with multiple network interfaces per node.

• Intel® C++ Compiler for Linux, version 9.1
• Intel® Fortran Compiler for Linux, version 9.1
• GNU* Fortran 95 compiler, version 4.0 and higher

• Support for SLES* 10

• Support for Intel® Debugger version 8.1-23, 9.1-23 — Advanced productivity by running the Intel command line debugger on a parallel application using Intel MPI Library and allowing you to attach a running MPI program.
• Intel® Trace Analyzer and Collector 7.0 — Enhanced analysis capability for system network activity with Intel MPI Library 3.1, adding new features such as trace file comparison and performance counters.
• Intel® Math Kernel Library 9.0 Cluster Edition — Built for optimal performance with the Intel MPI Library.

Intel MPI Library is designed to work with interconnection fabrics through DAPL, sockets, and shared memory interfaces. Examples of interconnects and computing servers providing these interfaces are listed in the following table.

MPI-2 standard compliance and portability
Ensure substantial MPI-2 standard compliance, with support for multiple interconnect solutions, based on Argonne National Laboratory's MPICH-2 implementation of the MPI-2 specification. Intel MPI Library is targeted toward industry-wide standardization of the MPI-2 ABI. All MPI-1 features are supported, plus many MPI-2 features including the following:

• File I/O
• Active target one-sided communication
• Passive target one-sided communication
• Generalized requests
• Preliminary thread support

• Enable performance analysis and tuning with Intel Trace Analyzer and Collector.
• Increase interoperability with your Fortran tools with MPI_IN_PLACE parameter, and MPI_Type_create_darray/subarray library calls.

• Allinea Distributed Debugging Tool (DDT*) support, version 1.9.2 and higher
• Etnus TotalView* debugger support, version 7.1.0 and higher

Every purchase of an Intel® Software Development Product includes a year of support services, which provides access to Intel® Premier Support and all product releases during that time. Intel Premier Support gives you online access to our expert engineering support staff, technical notes, application notes, and documentation. Install the product, and then register to get support and product update information.