[antlr-interest] REAL or int or Range

Jim Idle jimi at temporal-wave.com
Wed Jan 23 07:53:14 PST 2008


> -----Original Message-----
> From: Gavin Lambert [mailto:antlr at mirality.co.nz]
> 
> At 08:09 23/01/2008, Jim Idle wrote:
>  >
>  >But when you define using fragments, then you will quite
>  >often make a function call in the lexer, rather than just
>  >matching.
> 
> True.  Although the ANTLR compiler ought to peephole optimise that
> away.  (It doesn't at the moment though, sadly.  Still, the target
> compiler might.)
> 

There are quite a few codegen optimizations that could be (and I am sure 
will be in the end) achieved. Of course, the first thing you want is a 
accurate code generation and bugs removed from the analysis engine. Then 
effort can be put into the generated code.

In general, I think that the better C compilers would see the fragment 
functions as excellent targets for inlining, but some compilers are 
better than others of course ;-)

Jim



More information about the antlr-interest mailing list