Portability Problems Research Articles

Observations on the portability of Ada I/O

I have been involved recently in developing reusable Ada components which are portable (in theory) across multiple systems using multiple compilers. One of the biggest problem areas in Ada portability aside from Chapter 13 issues is Chapter 14, Ada I/O. This paper represents an informal collection of my working notes on the portability problems of Chapter 14. These notes are based on my experience with a wide variety of compilers on an even larger set of hardware. I have purposefully not mentioned any compilers by name as I only wish to show the differences between implementations and do not wish to imply that any compiler is better than any other. While I/O is inherently non-portable, packages of common I/O functions such as those described in Chapter 14 can be specified in a manner which ensures maximum portability. There are too many loopholes and areas for implementor interpretation in current Ada 1.0; I hope to see more stringent I/0 specifications in 1988.

Electromagnetic field modeling on supercomputers

The relatively new area of problem solving on supercomputers offers exciting new possibilities for numerical solution of 3-D electromagnetic field problems. The solution of systems of equations with tens of thousands of unknowns and bandwidths in the thousands is all but impossible on conventional computers. The availability of new vector supercomputers such as the CRAY 1 and the CYBER 205 brings both hope and new challenges. The faster machine cycle, vector instructions and parallel computing combined with advanced I/O devices promises to speed up the solution of large problems considerably. On the other hand, efficient programming on such machines is far from being trivial and in most cases, one has to compromise portability of the code. In addition, these machines require a front end computer and are not interactive, complicating and lenghtening the program development process. This work presents an overview of the work done to date on electromagnetic field modeling on supercomputers. Some data on field solution on an attached array processor is also included since array processors are a special class of supercomputers and for comparison purposes. Using some very large eddy current problems the various aspects of computing are presented together with results showing dramatic improvements in solution times. The problem of code portability, vector complex computation and program conversion are also addressed and, finally, the need for new algorithms, specifically designed for vector computing is presented.

A hydrostatic profiler for nearshore surveying

A portable beach-profiling system is described that utilizes the hydrostatic pressure principle to measure vertical differences between a pressure sensor that follows the profile and a reference sensor on the shoreline. The two instruments are connected by a fluid-filled tube. Horizontal distance is determined by measuring the extension of the tube. The complete profiling system, including a data-logging capability on magnetic tape, is carried in a small four-wheel-drive vehicle. Vertical accuracy in ocean surveys is estimated to be 2.5 cm and horizontal accuracy approximately 1 m. This accuracy is sufficient to resolve sediment transport volumes in most applications. Surveying the underwater portion of nearshore beach profiles has presented significant accuracy problems in the past. The highest accuracy is required at the greatest offshore distance to resolve sediment motions. Boat-supported surveying systems, using the sea surface as a horizontal datum, suffer from the presence in the record of surface waves with a broad range of periods which are difficult to remove from the profile. A number of systems have been developed to eliminate the sea-surface intermediate reference and to measure the submerged portion of the profile directly. These systems involve large vehicles that are propelled or pulled along the bottom and are restricted by portability limitations. The system described eliminates many of the portability problems.

The P-4 Deck (Portable Patient Problem Pack): Educational Tool of the Future?

This paper describes the structure and use of the Portable Patient Problem Pack (P-4 Deck) as a clinical simulation and educational tool. Originally designed by Dr. Howard Barrows of McMaster University to present medical diagnostic problems, the deck has been adapted by the author for occupational therapy. This tool is described and its current use in this context and future trends are outlined.

System design methodology in C.A.D. programs based on the finite element method

The general structure of a Finite Element package is analysed and we show how it is possible to build an evolutive system, both highly structured and without a great problem of portability. The help of a program generator may reduce the time devoted to the construction or the adaptation of a package and may increase its reliability. Problem solving techniques are available for the optimization of the general structure of the system and the reduction of computation time. A global structure well adapted to the present needs of computer aided design using finite elements is given in an example.

Fortran 77 portability

AbstractThis paper is the result of a study of the portability problems which might arise in the use of the Fortran 77 language. The study involved both an examination of the text of the Standard and discussions with a number of compiler‐writing teams. The paper identifies areas of Fortran 77 which are likely to cause problems, either because of an incomplete language specification, difficulties in compilation, or user inattention. It will be of interest not only to users writing large packages in Fortran, and to Fortran compiler writers, but also to those involved in language standardization and portability in general.

The construction of a portable editor

AbstractThe paper describes the implementation of an editor designed and constructed with portability in mind. We give first a brief introduction to the editor from the user's point of view, then we discuss the methods used to ease the problems of portability, and finally we explain some data structures used in its implementation, structures which, we believe, are interesting in their own right.

An initial evaluation of learning units to facilitate problem solving and self-directed study (portable patient problem pack)

This paper describes the initial evaluation of the portable patient problem pack, a learning unit designed to facilitate the development of problem-solving skills and stimulate self study. A cross-over design using two neurological patient problems, each presented in two formats (simulated patient and P4), was used with a sample of twenty-nine volunteer nursing and medical students. The results indicate that the P4 appears to be feasible and effective in stimulating problem solving and self-directed study. The suggested concurrent validity, low cost of production, and implementation make it a potentially potent tool for evaluation.

Un Langage-Machine Pour L’ecriture De Programmes Completement Portables En Contexte Systeme*

With the very fast evolution of mini- and microcomputers, the problem of software portability has become very important. The λ-machine has been designed and implemented for a complete portability of system programs. It especially allows the realization of operating systems common to different machines. The λ-machine is examined in detail; in particular, its basic language is studied; five implementations have been already realized.

RATMAC: A preprocessor for writing portable scientific software

AbstractIn preparing a revision of XRAY, a large scientific crystallographic software package, it was decided to use a structured language as the primary coding language. RATFOR was chosen as the most appropriate preprocessor. Initial implementation experience showed that while RATFOR improved the quality of the coding it did little to ease the problems of portability. By augmenting RATFOR with a powerful macro processor, a versatile tool for generating portable scientific software was developed. The resulting preprocessor, RATMAC, is portable and can be optimized, via macros, to a local operating environment. Its characteristics, strengths and weaknesses are described.

Managing Data in FORTRAN Programs

The concept of managing data in FORTRAN programs is introduced and illustrated by following the evolution, through several stages of development, of a simplified program that has the essential features of larger nonlinear structural analysis programs. The relevant characteristics of FORTRAN are reviewed and the techniques of dynamic dimensioning are considered. A FORTRAN data manager is provided in both simple and production versions. Listings of all essential coding are included along with some tips on data management. The paper concludes that the problems of program portability and adaptability can be reduced considerably if FORTRAN data management functions are relegated to an appropriate set of data managing routines. The emphasis in the paper is on the management of main memory data but a brief description is given of a family of programs for managing secondary storage.

Portability and the national energy software center

The software portability problem is examined from the viewpoint of experience gained in the operation of a software exchange and information center. First, the factors contributing to the program interchange to date are identified, then major problem areas remaining are noted. The import of the development of programming language and documentation standards is noted, and the program packaging procedures and dissemination practices employed by the Center to facilitate successful software transport are described. Organization, or installation, dependencies of the computing environment, often hidden from the program author, and data interchange complexities are seen as today's primary issues with dedicated processors and network communications offering an alternative solution.

UNIX Time-Sharing System: RBCS/RCMAS-Converting to the MERT Operating System

The paper presents a case history in applying the MERT executive to a large software project based on the UNIX∗ system. The work illustrates some of the basic architectural differences between the MERT and UNIX systems as well as the problems of portability. Emphasis is on matters pertaining to software engineering and administration as they affect development and support of a manufactured product.

The design of a portable programming language

We consider the design of a strongly-typed language with userdefined types in which it is arranged that, given that a type is available, it is immaterial to the user whether it is a user-defined type or one of the "primitive types" with representations selected by the implementer. This scheme provides unprecedented freedom in choosing the primitive types; by making these machine-dependent we can ensure production of programs that are easily and efficiently portable between computers of different architectures. A general discussion of the implementer's responsibilities in choosing primitive types appropriate to his machine is illustrated by considering implementation choices for translation of the language into BCPL. Finally we discuss the contribution of the language to the solution of the portability problem.

Program portability in high-energy physics

This paper discusses the computing problems in high-energy physics and reviews the portability achieved in software in the field. The techniques employed at CERN for developing programs are described and suggestions put forward for the future solution of the problems of portability.

Name standardization and value specification for machine dependent constants

There are a number of factors which limit the ease of transfer of mathematical software from one computer sytem to another. For FORTRAN implementations of mathematical and statistical algorithms, a primary portability problem is caused by necessary variation in machine dependent constants. The problem occurs for both decimal and machine-specific representations. Redish and Ward (1971) discuss the problem, and propose a solution. This note extends their ideas to additional constants and computer types.

PORTABILITY: A NEW LOOK AT AN OLD PROBLEM.

Increased interest in portability and load-carrying problems among Army human factors specialists is hading to the eventual compilation of a handbook for these areas. Available research material and handbook data are summarized and evaluated. The nature of portability research is examined, and the work required for the preparation of a useful handbook is indicated. Emphasis is placed upon the critical areas of criteria selection, training and motivation.