Programming in Vienna Fortran

Abstract

Exploiting the full performance potential of distributed memory machines requires a careful distribution of data across the processors. Vienna Fortran is a language extension of Fortran which provides the user with a wide range of facilities for such mapping of data structures. In contrast to current programming practice, programs in Vienna Fortran are written using global data references. Thus, the user has the advantages of a shared memory programming paradigm while explicitly controlling the data distribution. In this paper, we present the language features of Vienna Fortran for FoRTRAN 77, together with examples illustrating the use of these features

Efficient Index Generation for Compiling Two-Level Mappings in Data-Parallel Programs

“…Currently, parallelizing compilers try to fill the gap between users and machines in order to relieve the burden of such tedious and error-prone work from users. Many researchers have investigated the extension of existing languages to provide users with more convenient ways to manage data distribution, such as using Fortran D [6,26], High Performance Fortran (HPF) [8,16], and Vienna Fortran [2,3]. The major characteristic of these languages is that they provide users with a global addressing space and directives to specify data distribution at the language level.…”

Integrated range comparison for data-parallel compilation systems

“…Traditionally, distributed memory architectures have been programmed using message passing, where the user is responsible for explicitly inserting communication statements into a sequential program. The development of parallel languages, such as Vienna Fortran [2], Fortran D [3], and High Performance Fortran (HPF) [4], improved the situation by providing high-level features for the specification of data distributions. Among others, the Vienna Fortran Compilation System (VFCS) [5] and Fortran D compilation system [3] have been developed to support such languages and to automatically generate a message passing program.…”

The Nexus Approach to Integrating Multithreading and Communication

“…Because communication patterns are predictable, the programmer or compiler can place a matching receive for each send in a program. Libraries supporting this model, such as p4 [9], PVM [19], and MPI [24], are designed for programmer use but are also used as compiler targets by data-parallel languages such as High Performance Fortran (HPF) [37], Fortran-D [38], Vienna Fortran [11], and pC ++ [30]. In less regular, loosely synchronous problems, a programmer or compiler may possess global knowledge of the successive phases executed by a program, but not of the the precise sequence of operations and communications performed in each phase on each processor.…”