[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] DFA compile speedup (trevor_3_3.2)

From: Trevor Morris
Subject: Re: [gnugo-devel] DFA compile speedup (trevor_3_3.2)
Date: Wed, 29 May 2002 12:08:23 -0400

Is it possible that building the DFA in a different order (as this
patch does) changes the order of the matching?  I think the answer
is obviously yes, and I think that's what's causing the regression
differences.  I removed the patch, and the regression differences
also disappeared.

I don't see how to remove dfa_add_string, as that is the code that
(in the patched version) manages the intermediate dfa bins.  It's
the function that mkpat calls for each pattern to add it to the dfa,
so I'm not sure how you recommend restructuring it.


At 05:11 PM 5/29/2002 +0200, Tanguy URVOY wrote:
>Trevor Morris wrote:
>> Note that all of the DFA functions used to take a dfa_t * parameter - this
>> is no longer used in dfa_add_string(..).  The upshot is that it's only
>> possible to build one DFA at a time (without calls to init_dfa() ).  I'm
>> not sure this wasn't the case before, and it's currently never attempted
>> anyway.  Not sure it this is considered a problem.
>My opinion about dfa_addstring() is that it should be
>removed from the code, dfa_finalize do the job better.
>My first idea was to write a kind of "generic" code,
>it is natural for functions like sync_prod(...) but
>not for local functions like dfa_add_string() or dfa_finalize().
>Your patch seems ok to me.
>Tanguy Urvoy
>gnugo-devel mailing list

reply via email to

[Prev in Thread] Current Thread [Next in Thread]