Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!usc!apple!snorkelwacker.mit.edu!bloom-beacon!eru!hagbard!sunic!mcsun!hp4nl!charon!dik From: dik@cwi.nl (Dik T. Winter) Newsgroups: comp.lang.fortran Subject: Re: Fortran vs. C for numerical work (SUMMARY) Message-ID: <2619@charon.cwi.nl> Date: 29 Nov 90 02:13:10 GMT References: <7097@lanl.gov> <17680:Nov2806:04:1090@kramden.acf.nyu.edu> <7928@uwm.edu> Sender: news@cwi.nl Organization: CWI, Amsterdam Lines: 33 All that stuff. A basic problem still is that the major numerical libraries (IMSL, NAG) are in fortran. Some suggested f2c. While pretty good it does not help if you do not have the source (and IMSL and NAG are not distributed with source). And really, a very large amount of numerical programming depends on those libraries. Some suggested calling the Fortran routines from C; may I respectful ask how to call the following Fortran routine from C on a Sun without writing intermediate Fortran code? REAL FUNCTION F() F = 1.0 RETURN END Dan Bernstein suggested multiple compilations in order to allow for all possible aliasing problems. But how to do that if you do not have the source? How to do a program wide flow analysis if you do not have the source, but only the libraries? A large body of numerical software still depends on Fortran. There are many reasons for this: 1. They use common libraries, written in fortran. (I know that NAG is considering a library in C; they are already considering a long time!) 2. There are still systems that do not have C (surprise!). 3. There are systems that do have C, but where C is much less efficient than Fortran (Alliant, CDC Cyber, Cray ...). 4. And of course the always recurring question about arrays. Fortran is half adequate, C is not adequate. (There are a few languages with full support for arrays; I know Algol 60, Algol 68 and Ada; there are probably a few more, but Pascal as defined by Jensen and Wirth is not amongst them.) Oh, well.. -- dik t. winter, cwi, amsterdam, nederland dik@cwi.nl