Beebe@SCIENCE (Nelson H.F. Beebe) (01/21/88)
Both Fortran and C are currently in the ANSI draft standardization process, and it is an appropriate time to reflect on the vagaries of floating-point arithmetic and on the question of providing decent floating-point support in programming languages. The IEEE P754 floating-point standard, now implemented (though not always completely) in coprocessors from Inmos, Intel, Motorola, National Semiconductor, Weitek, Western Digital, and perhaps others, provides a much more satisfactory model of floating-point arithmetic than has previously been the case on commercial computers. David Hough (pronounced "huff") (Arpanet: dgh@sun.com, Usenet: sun!dhough), one of the designers of IEEE P754, has been making a careful analysis of draft ANSI C with respect to floating-point support, and his extended comments have been evolving over the last few months on the basis of feedback from net colleagues who review each new draft. For a recent contents summary of his work, see the posting of 14-Dec-87 on the NA (numerical analysis) newsgroup distributed from SCORE.STANFORD.EDU. Since the topic of Fortran to C convertors pops up frequently in both the Fortran and C newsgroups, there is clear evidence that many Fortran users are contemplating, or are already in progress with, a conversion from Fortran to C. It is also significant, I believe, that the trend on supercomputers is toward Unix operating systems, and to investigation of the use of facilities, like remote procedure calls permitting dynamic communication between workstations and supercomputer, wherein C will play an increasingly important role in scientific computations. I therefore feel that it is relevant to bring Hough's work to the attention of this newsgroup, because many of his comments APPLY EQUALLY WELL TO FORTRAN. He will send copies of his comments to anyone on the net who requests it. If you feel that the quality of floating-point support in programming languages is important, I urge you to send for his comments. -------
john@nmtsun.nmt.edu (John Shipman) (01/26/88)
In article <12359761762.8.BEEBEX@SCIENCE.UTAH.EDU>, Beebe@SCIENCE (Nelson H.F. Beebe) writes: > ... [most of posting deleted] > Since the topic of Fortran to C convertors pops up > frequently in both the Fortran and C newsgroups, there is > clear evidence that many Fortran users are contemplating, or > are already in progress with, a conversion from Fortran to > C.... > [C may] play an > increasingly important role in scientific computations. > What about type COMPLEX? Presumably the FORTRAN-to-C translators are generating calls to library routines to handle COMPLEX calculations. Questions for you folks in net-land: 1. Will complex-number users mind writing a lot of function calls, instead of just writing expressions as they do now in FORTRAN? 2. Has the standards committee considered a complex type for C? -- John Shipman/Zoological Data Processing/Socorro, New Mexico USENET: ihnp4!lanl!unm-la!unmvax!nmtsun!john CSNET: john@nmt.csnet ``If you can't take it, get stronger.'' --Falline Danforth