[antlr-interest] greedy vs nongreedy lexer rules
Terence Parr
parrt at cs.usfca.edu
Sun Apr 18 16:56:18 PDT 2010
On Apr 18, 2010, at 4:52 PM, Cliff Hudson wrote:
> On the subject of multi-threaded implementation, let me vote +1 for providing a mechanism for target implementors to take advantage of this. :) Multi-core systems are the norm now. In my job, we spend a LOT of time determining how best to extract maximum work in minimum time, and parallel programming is a big part of that.
Yeah, best we can do is linear for NFA evaluation given infinite processors, though, unless I allow lexer to race ahead of parser, buffering up all the tokens. It's a good idea to explore.
Synchronization might kill idea of threads to handle a single token though. I need n-1 threads to die when one of the threads succeeds. actually that could be unguarded boolean i think. anyway, i'm still thinking about action execution. grrr. a hard nut to crack.
Ter
More information about the antlr-interest
mailing list