|
|||||||
High Performance Computing: FORTRAN, OpenMP and MPITechnical University of DenmarkGeneral course objectives: FORTRAN is one of the principal languages used in scientific, numerical and engineering programming and knowledge in FORTRAN is an indispensible qualification for students, researchers, and engineers. With the two recent revisions of the language, the power of the language has been progressively enhanced, and most vendors (IBM, HP, SGI, Intel, Sun, Cray) provide highly optimizing FORTRAN compilers, based on more than 50 years of experience. MPI (Message Passing Interface) is the de-facto standard for programming portable message-passing parallel applications on networked computers (also known as clusters). MPI has bindings to C/C++ and FORTRAN, and it is available on all massively-parallel supercomputers. OpenMP (Open Multi-Processing) is a standard application programming interface (API) that supports multi-platform shared memory multiprocessing programming in C/C++ and FORTRAN. It consists of a set of compiler directives, library routines, and environment variables that enable shared memory parallelism and execution. The current and the future supercomputers consist of a cluster of computers (nodes) each with multiple processing cores (dual, quad, multicore) that share the memory of the node. The goal of the course it to outline strategies for porting and writing efficient, portable, maintainable code for such hybrid computer archictectures. Learning objectives: A student who has met the objectives of the course will be able to:
Contents: The Ph.D. course consists of 3 parts: Part I: Programming FORTRAN Part II: Introduction to OpenMP Part III: MPI |
|
||||||