Re: Intermediate forms (again?): worthwhile?

Tony wrote:


> Something just a tad higher level than assembly and a bit more regular
> would be nice, but so far, I haven't seen anything I'd want to commit
> to. Surely if I could just concentrate on the compiler front end and
> attach it to a back end black box supporting multiple platforms, that
> would be great.


Mod said:


> [The usual suggestion these days is LLVM. -John]


Maybe I just need to "grow into it". My first proof-of-concept
implementation will be "walk the AST and generate x86-32 assembly",
the AST being the intermediate form. I believe MS is readying it's AST
to become prominent and available via APIs (if that strategy hasn't
changed by now, as I read that years ago).


I can categorize and comment on the 4 IRs I know of now (note that
cross-platform targeting is not necessarily a requirement):


      1. Assembly Language: write to one HW architecture and translate from
that to all others.
      2. GCC, LLVM: proprietary instruction stream, but inherently
cross-platform.
      3. C--, VooDoo: intermediate LANGUAGE.
      4. C: intermediate language only suitable for C-like derivative
languages.


(1) is nice because it does not introduce a 3rd party, esoteric step,
save for the assembler.
(2) GCC is nice because it allows "instant" availability on a number of
platforms, but the big drawback is GPL.
(3) is probably better than C for non-C-like languages, but suffers from
the ubiquitousness of something C and THAT is pretty much all of what
makes a LANGUAGE IR lucrative.
(4) Ubiquitousness across all platforms but limits to C-derivative
languages.


For my project, C and GCC are immediately dismissed as unappropriate, for
separate reasons. LLVM will get another look after I do (1). I find (3)
harder to grok than (1) so I have no plans to visit them in the future.

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