KVC@ENGVAX.UUCP.UUCP (02/18/87)
> I have heard that if I you want it to go fast and yet use a high-level > language, use Fortran. Fortran is the fastet language under VMS, they say. > > Anyway, I have just gone into a real-time project for VMS and currently > they're using Fortran and Macro. What I'd like to know if there is any one > who can give me *facts* on this issue. Are the code produced by the Pascal > compiler really so much slower than the one of the Fortran compiler? > Especially if you are using all these fancy Fortran-8X features, which > doesn't to be aimed to make quick programmes. > If you have facts that are supporting my view, please give me references > so I can convince the other project members. > Erland Sommarskog I'm afraid I don't have "facts", of the sort that are going to convince die-hards. Die-hards generally think of benchmarks as "facts". If someone out there has benchmarks to offer, fine, though my personal opinion is that benchmarks tend to benchmark unrealistic situations. I offer my own generalizations from 8 years of VAX/VMS experience. Currently (that is, with current releases) all of the DEC VMS language products produce good code. Some may be tuned to producing slightly better code given a certain type or style of program. I argue that, taken as a set, the VMS compilers are good enough that your choice of language should not be based on ancient mythology but on what language is best suited to expressing the program and hence allowing the compiler to most easily convert it to code. I offer proof as follows: Given your example of a system-programming application, I doubt very seriously that any FORTRAN programmer could create a FORTRAN program that produces better code *for that application* than a competent BLISS programmer could with a BLISS program. The two languages are just NOT equally applicable to the problem at hand. (I would argue that the C compiler is also better applied to this problem than FORTRAN). Note! I am not saying that VAX FORTRAN is not an excellent FORTRAN compiler and I'm not saying that the FORTRAN solution is a bad one. If it does the job and the programmer was comfortable with it, then it's a fine solution. I'm merely stating that it's ridiculous to claim that the FORTRAN compiler produces the "fastest code". The project in question may not want to use BLISS, but I think I've irrefutably argued against their blanket statement. I would say that VAX Pascal would produce code at least as good as FORTRAN given a system programming application. If the programmer (you!) is more comfortable in Pascal than FORTRAN for this application (I know I would be, I love VAX Pascal) than you're gonna get a better application in less time. Everyone wins. So, while there certainly may be perfectly good reasons for spec'ing the project in FORTRAN and MACRO (programmer experience, etc...) claiming you'll get "the fastest code" that way is blatant misinformation. Note that the entire premise of my argument is based on the belief that all the major VAX language products have now matured to the point where they provide excellent support for the languages they compile, allowing you to choose your language based on the language, not the implementation of the compiler. /Kevin Carosso kvc%engvax.uucp@usc-oberon.usc.edu Hughes Aircraft Co. ps. I don't condone slacking off from you DEC compiler-writers either! If we can't find you a bug here and there to keep you busy, you'd best just keep teaking for more speed!