[antlr-interest] attempt to compare antlr vs bison/flex performance

Terence Parr parrt at cs.usfca.edu
Fri Dec 28 09:56:36 PST 2007


On Dec 20, 2007, at 2:27 PM, Minas Hambardzumyan wrote:

> I guess I could compare antlr java parser with the bison version in  
> the gcc compiler. I will send an update when I get some results...
>
> One conclusion seams to be obvious though. If the speed is of  
> essence, then one has to simplify the grammar definition as close to  
> LL(1) as possible.

Just LL(*) or perhaps LL(k).  It's the backtracking that is inefficient.

> Doing so though moves the grammar definition away from the simple to  
> read EBNF format -- one of the most important advantages of using  
> antlr. So it is up to the user to pick -- easy to use/read vs not os  
> easy but fast...

The point of LL(*) is to get power w/o forcing unnatural grammars or w/ 
o deadly performance.  90% of the time, you'll be fine with ANTLR.

Ter


More information about the antlr-interest mailing list