This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| LAM/MPI | |
|---|---|
| Name | LAM/MPI |
| Title | LAM/MPI |
| Developer | Los Alamos National Laboratory; University of Tennessee contributors |
| Initial release | 1995 |
| Stable release | 7.1.4 |
| Programming language | C (programming language), Fortran |
| Operating system | UNIX, Linux, Solaris |
| Platform | x86, PowerPC, SPARC, Alpha (microprocessor) |
| License | BSD license |
LAM/MPI
LAM/MPI is a portable implementation of the Message Passing Interface standard designed for distributed-memory parallel computing on clusters and supercomputers. The project provided tools and libraries to support high-performance scientific computing on heterogeneous networks, emphasizing process management, communication protocols, and profiling utilities. It was widely used in research institutions and national laboratories for codes developed in Fortran and C (programming language), and for integration with toolchains from vendors such as Intel and Cray Inc..
LAM/MPI implemented the MPI-1 and parts of MPI-2 specifications to enable message-passing parallelism across nodes in a cluster or networked workstation environment. The distribution included a runtime system for job startup and teardown, a communication substrate with support for TCP/IP and vendor-specific interconnects, and debugging/profiling tools to work with development environments such as GNU Compiler Collection and TotalView (software). It was adopted by research groups at Los Alamos National Laboratory, Argonne National Laboratory, Lawrence Livermore National Laboratory, and universities including Stanford University and Massachusetts Institute of Technology.
Development began in the mid-1990s at Los Alamos National Laboratory to address the need for a flexible MPI implementation on commodity clusters. Early releases coincided with concurrent efforts such as MPICH at Argonne National Laboratory and commercial offerings from IBM and HP (Hewlett-Packard). Over its lifecycle, the project incorporated contributions from academic and national laboratory partners including University of Tennessee and entities working on Beowulf clustering. LAM/MPI influenced subsequent middleware and was a stepping stone toward more modern implementations like Open MPI. The project was eventually superseded as standards and interconnect technologies evolved, with source and binaries archived by several institutions.
The architecture combined several components: a process manager for launching and monitoring tasks, a message-passing library implementing MPI calls, and auxiliary tools for tracing and resource discovery. The runtime used a daemon-based model derived from concepts used in p4 (parallel programming), with a head node coordinating workers on compute nodes such as Cray XT or commodity x86 blades. Communication plugins abstracted transport layers including TCP/IP sockets, shared memory segments on UNIX System V, and vendor fabrics like Myrinet and InfiniBand. The toolset included a profiling interface compatible with MPI Profiling Interface hooks and integrated with visualization tools used at Sandia National Laboratories.
LAM/MPI provided collective operations, point-to-point messaging, communicators, and support for derived datatypes as specified by MPI-1; partial MPI-2 features covered dynamic process management and one-sided communications in limited form. It included a startup system that managed heterogeneous environments across subnets and implemented fault-detection and logging facilities used by researchers at Oak Ridge National Laboratory. Tools such as lamtrace and lamtest assisted developers using environments like Sun Microsystems Solaris and GNU/Linux distributions, and it interoperated with debuggers such as gdb and TotalView (software).
LAM/MPI was optimized for low-latency and high-bandwidth paths on networks used by national laboratories, with tunings for Myrinet and InfiniBand fabric characteristics and CPU architectures like Alpha (microprocessor) and PowerPC. Benchmarks conducted at institutions such as Lawrence Berkeley National Laboratory compared LAM/MPI to implementations like MPICH and vendor MPI stacks from IBM and SGI (company), demonstrating competitive performance for mid-size clusters. Scalability was influenced by the daemon-based process manager and the efficiency of collective algorithms, factors studied in academic venues including ACM conferences and workshops hosted by IEEE.
Distributions were built for many UNIX-like operating systems and integrated with compilers from GNU Compiler Collection, Intel, and vendor toolchains such as Sun Studio. LAM/MPI supported interoperable execution models and could run user applications developed for standards-compliant MPI environments, enabling portability between systems at National Energy Research Scientific Computing Center and university clusters. Interfacing layers allowed it to exploit vendor-specific networking libraries and kernel modules employed on platforms from Dell and HP.
LAM/MPI was used to run scientific codes in computational physics, climate modeling, and materials science at institutions including Los Alamos National Laboratory, Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, and universities such as University of California, Berkeley and Massachusetts Institute of Technology. Applications included parallel solvers, domain-decomposition codes, and ensemble simulations leveraging MPI semantics in projects funded by agencies like National Science Foundation and Department of Energy (United States). It also served as an educational tool in high-performance computing courses at Stanford University and University of Illinois at Urbana–Champaign.
Category:Message Passing Interface Category:Parallel computing software