From ljm@slac.stanford.edu Thu Mar 23 05:33:36 1995 Received: from SCSW6.SLAC.Stanford.EDU by dkuug.dk with SMTP id AA21382 (5.65c8/IDA-1.4.4j for ); Thu, 23 Mar 1995 22:31:08 +0100 Received: from [134.79.128.74] (MOZART.SLAC.Stanford.EDU) by SCSW6.SLAC.STANFORD.EDU (PMDF V4.3-10 #6987) id <01HOH6Q296DC0033HG@SCSW6.SLAC.STANFORD.EDU>; Thu, 23 Mar 1995 13:32:00 -0800 (PST) Date: Thu, 23 Mar 1995 13:33:36 -0800 From: ljm@slac.stanford.edu (Leonard J. Moss) Subject: REPOST: Conditional Compilation in F90 X-Sender: ljm@popserv To: SC22WG5@dkuug.dk (SC22/WG5 Mailing List) Message-Id: X-Envelope-To: SC22WG5@dkuug.dk Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7BIT X-Charset: ASCII X-Char-Esc: 29 Here is the complete text, as posted by Mallory to the X3J3 list, of the request from James Billen for a conditional compilation facility. -- Len Moss >Date: Wed, 22 Mar 1995 13:52:02 -0820 >From: NORTHCM%MHS@mhs.rose-hulman.edu >Subject: (x3j3.1995-131) Conditional Compilation in F90 >To: x3j3@ncsa.uiuc.edu > > X3J3/95-108 > >To: C. Mallory North, Jr., X3J3 Librarian >From: James H. Billen, Los Alamos National Laboratory >Date: March 17, 1995 > >This message is my formal request to the Standards Committee >for inclusion of conditional compilation in the F90 standard. Please >assign this request a paper number for the next meeting of the >committee. Thank you for your assistance. > >Several of my colleagues and I develop FORTRAN codes used for >accelerator design, including rf cavity modeling, static electric and >magnetic field solvers, beam-dynamics simulations of space-charge >dominated beams, generation of the machine-tool instructions >for fabrication of accelerator hardware, and measurement of the >accelerator fields and other rf-cavity properties. Our programs, >which involve a few million lines of code, are used extensively by >the worldwide accelerator community. For example, Lloyd Young >and support the POISSON/SUPERFISH codes, which have become >popular with over 300 users. > >Most of our codes are targeted for the PC platform and for this we >have been using products from Lahey Computer Systems, Inc. Our >codes already include FORTRAN 90 features, such as allocatable >arrays, that are available as extensions of the FORTRAN 77 standard >in Lahey's F77L3-EM/32. > >We were very disappointed to learn that conditional compilation, an >extension of the standard available in EM/32, was not added to Lahey's >new LF90 compiler. Tom Lahey explained that they were waiting to see >if conditional compilation became part of the standard so they would not >have to redo its implementation at a later date. > >Conditional compilation is a common extension in implementations >of high-level languages including FORTRAN. According to Kernigan >and Richie (The C Programming Language, Second Edition, Prentice >Hall, 1988), conditional compilation was (in 1988) part of the proposed >C-language standard, ANSI X3.159.1989. A colleague who programs >in C told me he believes this standard was adopted. > >Many people use conditional compilation to produce executable >versions targeted for different machines. Our codes make much more >extensive use of the conditional compilation directives. For example, >various SUPERFISH codes include an electromagnetic-field solver that >can be compiled with either real fields and material properties or with >complex fields and material properties. Only a few lines of code need >to be changed to switch from real to complex and it is very convenient >to maintain only one source file, using compiler directives to include >the >appropriate lines. There are numerous other examples. We believe that >Lahey's implementation of conditional compilation in their EM/32 product >was instrumental in helping us to develop the POISSON/SUPERFISH >codes into a robust, self-consistent package. > >One can, of course, use a stand-alone or third-party preprocessor to >handle the compiler directives, but this eliminates some significant >advantages. The "built-in" feature makes using the debugger more >straightforward. When a preprocessor creates a new file, then one >is debugging and editing a file other than the original source. Changes >must be made twice (increasing the possibility of error), or, if one >edits >the original file between debugging sessions, one does not have access >to correct line numbers from the compiled source and the turn-around >time between tests is longer. > >I hope that Standards Committee will be able to consider this request >for conditional compiler directives as part of this year's agenda. It >would seem to be a fairly benign addition to the language standard for >those programmers who do not use it. But, it will seriously influence >how long it takes some programmers to migrate to a new compiler. >It may decide whether or not some of us make the switch at all. > >Thank you for considering this request. > > >__________________________________________________________ >James H. Billen (jbillen@lanl.gov) >Los Alamos National Laboratory >Accelerator Operations and Technology Division >Group AOT-1, Mailstop H817 >Los Alamos, NM 87545 >Phone: 505-667-6627 >Fax: 505-665-6590 > -- Leonard J. Moss | My views don't necessarily Stanford Linear Accelerator Center | reflect those of SLAC, MS 97; P.O. Box 4349; Stanford, CA 94309 | Stanford or the DOE