[antlr-interest] pleasant new table-based DFA speed

Terence Parr parrt at cs.usfca.edu
Sun Jun 4 13:40:19 PDT 2006


On Jun 4, 2006, at 1:35 PM, Martin Probst wrote:

> Hi,
>
> that sounds very promising. What are the overall savings, or are  
> you still optimizing parts of that?

Hi Martin,

Well, it's pretty good.  The wallclock shows an overall drop from an  
average of 4.29s to 3.57; about a 17% savings.

Just working on the DFA to drop the 1.9M of class files really ;)   
Just happens to be faster too :)

Ter




More information about the antlr-interest mailing list