High Performance Fortran

A basic problem in programming for parallel architectures is that each machine has its own design, and also its own specific software for accessing its hardware features. In 1992, as a response to this situation, the High Performance Fortran Forum (HPFF) was founded, under the guidance of Professor Ken Kennedy, to produce a portable Fortran-based interface to parallel machines for the solution of data-parallel problems, mainly involving regular grids. Over 40 organizations participated, and the work used existing dialects, such as Fortran D (Rice University), Vienna Fortran and CM Fortran (Thinking Machines), as inspiration. It was realised early on that much of the desired functionality already existed in the, then, very new Fortran 90 (see [1]), and this was selected as the base language on which to build HPF itself. The array processing features of Fortran 90 are particularly relevant. This enabled the goal of producing an ad hoc standard for HPF within a year to be met, and full details can be found in [2]. The standard document itself is obtainable by anonymous ftp at titan.cs.rice.edu in the directory public/HPFF/draft as the file hpf-v10-final.ps.Z. At the time of writing, HPFF has reconvened to produce a second, more advanced, version – HPF II. One new feature is expected to be an extension for irregular grids. The basic approach adopted was that of designing a set of directives that may be added to Fortran 90 programs, together with a few syntactical additions and some extra libraries, thus creating a data parallel programming language that is independent of the details of the architecture of the parallel computer it is run on. The principle is to arrange for locality of reference of the data by aligning related data sets to one another, and distributing the aligned sets over memory regions such that, usually, calculations on a given processor are performed between operands already on that processor. Any message passing that might nevertheless be necessary to communicate data between processors is handled by the compiler and run-time system.