[antlr-interest] Profiler 2
Terence Parr
parrt at cs.usfca.edu
Fri Dec 2 09:59:35 PST 2005
On Dec 1, 2005, at 11:50 PM, Martin Probst wrote:
> I consider myself being part of that group, too. However we don't have
> any performance problems using Java ANTLR with a handwritten Lexer and
> no ()=> at all. Or at least no real problems - faster is always
> better,
> but our parser munges through all queries in a couple of ms.
Yes, I'm pretty sure the lexer mechanism is *the* main bottleneck in
v2 and then overuse of (...)=> predicates.
The new lexers are much more efficient and there is PLENTY of room to
remove the obvious stupid generated code; I'll get to that later when
I finish all functionality.
Ter
More information about the antlr-interest
mailing list