Re: Compilier Construction Tools?

Late last year I had a similar set of questions. I hoped to post a
summary, but had trouble developing an appropriate format, and then
got distracted. I have a few comments:


PCCTS: The real strengths of this tool are lexing, parsing, and error
reporting, i.e., the grammar side of languages. Semantics, code
transformation, and code production have received less emphasis,
although it does provide tools for code transformation.


Eli: This was the tool I finally chose primarilly because of Waite's
work on Fortran lexical, syntactic, and semantic analysis. However,
the additonal backend tools it provides relative to PCCTS were also
attractive. The support has been very good, but the system is not as
portable as I would have liked (Eli runs only on Unix platforms, and I
had some installation problems on a Red Hat Linux. ) Note that while
the system is not as portable as I would like the code it generates is
ANSI C and reputedly very portable. The system is large, and difficult
to comprehend, it has more documentation than any other tool that I
have found, which helps, but nothing teaches you a system better than
using it for a problem of direct interest to you.


GRDP/RDP: I like Johnson's postings, but my primary interest was
Fortran where I believe an LL(1) tool is inadequate and GRDP has yet
to be released.


Cocktail: the non-commercial version is used by a number of people,
but the real strength appears to be the commercial version.


Gentle: I also had trouble finding information on this tool.


SUIF: I may be wrong, but I believe SUIF is primarilly for
intermediate portions of the compiler, i.e., optimizations and code
transformations. You might be able to use it with one of the above
tools to fulfill part of your needs, but I don't think it provides
parsing or lexing capabilities. They seem to be planning on using the
New Jersey Machine Code Tookit for the final code generation portion
of their system.


William B. Clodius
--

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