Message Passing Interface programming

W. Gropp, E. Lusk and A. Skjellum, Using MPI Portable Parallel Programming with Message-Passing Interface, Scientific and Engineering Computation series, 1994. 

We show in Figs. 2 and 3 the speed up for a Davidson iteration obtained with the PVM [17] ( Parallel Virtual Machine ) and PVMe [18] ( PVM enhanced ) message passing interfaces respectively. Speed-up factors are here relative to the sequential version of the program (n=l), and the theoretical mciximum has been defined according to the expression s n) — n — I appropriate to the used master/slave model. 

The scaling of the model performance with the number of processors is usually better with message passing programming (MPI message passing interface) than with a compiler parallelized code (Open MP) for relatively small problems. However, the preference of the different programming techniques is under debate. 

Includes appendices that discuss the Message-Passing Interface (MPI), multithreaded programming with Pthreads, and compiler extensions (OpenMP)... 

Gropp, W., Lusk, E., and Skjellum, A. (1999). Using MPI Portable Parallel Programming with the Message-Passing Interface, The MIT... 

The current protot3rpe system includes three Expert modules, the IR Expert, the STIRS Expert, and the Human. All modules are written in Lisp. The IR Expert is a rule-based infrared interpreter which we have developed. The STIRS Expert is an interface to the STIRS program, a pattern-matching mass spectrum interpreter developed by McLafferty and coworkers at Cornell University, which is written in Fortran. () ) The interface translates the output of STIRS into a form palatable to our program, and handles the message-passing protocol required by the Controller. The Human module controls communication with the user. It allows user-supplied elemental or substructure information to influence the course of the analysis. The power of... 

We now reexamine message passing as it pertains to software development and the interface of application software and library software. Compiler-managed parallelism is not yet ready for prime time. This means that efficient parallel programs are usually coded by hand, typically using point-to-point message passing and simple collective communications. There are many problems associated with this approach. 

---