Re: Simple retargetable compiler (maybe only local optimizations)

"toby" <toby@telegraphics.com.au>
17 Feb 2006 00:05:59 -0500

          From comp.compilers

Related articles
Simple retargetable compiler (maybe only local optimizations) nkavv@skiathos.physics.auth.gr (Uncle Noah) (2006-02-11)
Re: Simple retargetable compiler (maybe only local optimizations) vivekm@sankhya.com (2006-02-14)
Re: Simple retargetable compiler (maybe only local optimizations) toby@telegraphics.com.au (toby) (2006-02-17)
| List of all articles for this month |

From: "toby" <toby@telegraphics.com.au>
Newsgroups: comp.compilers
Date: 17 Feb 2006 00:05:59 -0500
Organization: http://groups.google.com
References: 06-02-080
Keywords: available, lcc

Uncle Noah wrote:
> I have been looking for some time for an easily (not GCC!) retargetable
> compiler, generating assembly code for RISC processor ISAs from ANSI C.
> This seems to be utopic since in most attempts the specification for
> the target processor is tightly integrated in the code generator (e.g.
> AST->IR->assembly). Since this is a very difficult problem in practice,
> we can exclude some of the hard stuff. Anyway, these are the
> requirements:
>
> 1. Compact (not obfuscated though!) machine description language.
> 2. Single file machine descriptions.
> 3. Support for simple register allocation (not necessarily Chaitin
> variation).
> 4. No (global) optimizations.
>
> The latter is very important; I'm interested in generating correct code
> for different instruction sets, BUT don't care about optimizations. If
> it look ugly, I will deal with it later.
>
> Any suggestions? LCC looks nice and jackcc has some nice options
> (http://jackcc.sourceforge.net) but neither scores all 4 requirements.


I would say lcc meets all four? Unless you don't like its
machine-independent optimisations. All machine-dependent optimisations
are in a 1-file md.


--T


Post a followup to this message

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