This note was prepared and agreed by the International Federation for Information Processing Working Group on Numerical Software (WG 2.5). It proposes the use of common conventions for the parameterisation of the environment in FORTRAN programs.
History
(i) the design of the algorithm, (ii) its realization as a documented source-language An early draft of this note was used as a discussion subroutine or program and document during the first meeting of the IFIP Work-(iii) the testing of the compiled code on a given coning Group on Numerical Software (WG 2.5) in Oxford figuration (i.e. a machine together with its operin January 1975. The meeting requested a precise ating system, compiler and available libraries) specification of the purpose of the note and suggested and its detailed documentation for that confia number of other improvements which led to a guration. second draft. Written comment led to further changes.
While the final code must depend on the configu-A third draft was discussed during a workshop on ration, it is very desirable for the source-language transportable numerical software in the Applied version to be transportable (i.e. needing only a small Mathematics Division of the Argonne National Labor-number of mechanical changes prior to compilation atory in August 1975. A fourth draft was written in on a specific configuration) and for the algorithm January 1976 and distributed widely for comment to be adaptable (i.e. expressed in terms that permit and criticism. Discussion at the NSF/ERDA work-efficient and accurate computation on any configurashop on portability of numerical software and at the tion that is used later). Our first aim is to suggest second meeting of the IFIP WG 2.5 (both in June parameters that allow the algorithm designer to make 1976), together with correspondence from other his work adaptable in this sense. For example, he may parties led to the preparation of the present docu-need the "relative precision" for a root-finding algoment. Although some of the comment was contradic-rithm or "page size" for an algorithm designed to tory, it is our belief that this final document repre-solve linear equations efficiently on a machine with sents a consensus view, virtual memory. Our second aim is to provide a larger set of parameters that can be used in the transportable source-Objectives language [step (ii)] in the expectation that actual values will be readily available in the eventual run The development of numerical software entails [step (iii)]. The set has to be larger since it should three distinct steps: include such items as the standard input and standard output units, neither of which are likely to be of