Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code

wilson@ann-arbor.applicon.slb.com (David Wilson)
Wed, 13 Jan 1993 16:11:47 GMT

          From comp.compilers

Related articles
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code nickh@CS.CMU.EDU (1993-01-08)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code jlg@cochiti.lanl.gov (1993-01-11)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code eifrig@beanworld.cs.jhu.edu (1993-01-12)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code tmb@arolla.idiap.ch (1993-01-12)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code jlg@cochiti.lanl.gov (1993-01-12)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code drw@euclid.mit.edu (1993-01-12)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code wilson@ann-arbor.applicon.slb.com (1993-01-13)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code nickh@CS.CMU.EDU (1993-01-13)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code chased@rbbb.Eng.Sun.COM (1993-01-13)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code bart@cs.uoregon.edu (1993-01-14)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code rv@erix.ericsson.se (1993-01-14)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code smk@dcs.edinburgh.ac.uk (1993-01-14)
Re: Compile Time vs. Run Time, Mixed Language Compiling, Fat Code tmb@arolla.idiap.ch (1993-01-15)
[1 later articles]
| List of all articles for this month |
Newsgroups: comp.compilers
From: wilson@ann-arbor.applicon.slb.com (David Wilson)
Organization: Schlumberger CAD/CAM; Ann Arbor, Michigan (USA)
Date: Wed, 13 Jan 1993 16:11:47 GMT
Keywords: optimize, debug

jlg@cochiti.lanl.gov (J. Giles) writes:
>It seems to me that the main problem with compiler writers ... is that
>compiler writers don't actually *use* the language the compiler is written
>to compile.


I spent 13 years implementing/maintaining/adding code generators to a
compiler for an extended version of Pascal. It was written in Pascal.
About two years into its life cycle, I bootstrapped it, and from there on
out we worked exclusively in the same language we were implementing. I
can attest to the value of compiler writers USING the language they are
compiling. Without our own full-time experience using the language
ourselves, the quality of our diagnostics, just for one issue, wouldn't
have been nearly so high.


It was also very educational to see how differently application
programmers (CAD/CAM software, in our case) use a language than compiler
writers do. The language features they would use (or not use), and the
ways in which they would use them, was totally different. The extent of
the differences caught me by surprise many times over the years. Even the
obvious fact that a primarily numeric application places different demands
on the language/comp- iler than the non-numeric application a compiler
itself is, doesn't explain it all. The entirely different mind-set of
compiler writers, who appreciate languages in their own right, as opposed
to people to whom the language is a means to an end, is important too.


My conclusion is that you're right that it is a problem that many compiler
writers don't actually *use* the language they're compiling. But, the
kind of programming they do, in contrast to their user community, is
important too. Having the compiler written in the same language it
implements is a good step towards solving this particular problem, but
isn't a complete solution in itself. I don't think that there is any
magic answer, other than to just try and make sure the compiler writers
are aware of the issue, and give it a lot of thought.


Dave Wilson
--


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.