Re: [patch] Fix of the CPU bottleneck in pack_vector

[Akim Demaille]

Le 21 déc. 2012 à 02:32, Yuri <address@hidden> a écrit :

> Hello,

Hi Yuri,

> I reported this issue with the testcase and patch in Jan 2012.
> Now I see that you didn't apply it as of 2.7 (Dec 2012).
> 
> Why didn't you apply it?

Because it segfaults in some situations, and because it does
not use our bitset routines, although it looks like it would
be useful.

> It fixes the algorithmic bottleneck in bison that I am sure someone else will 
> hit it sooner or later.

We did see an improvement on your test case, but not as large
as originally reported.  Yet, no, your patch is not forgotten,
and, time permitting, Théophile (Théophile, your cue here) is
working on it.

I'd be happy to see that patch mature and be integrated soon.

I wonder, though, if there are really world cases with such
massive sort of grammar.  AFAIR, you made up that grammar to
hurt the engine, right?

> Also why there is no bug tracking page for bison?

There is one, https://savannah.gnu.org/support/?group=bison

> How can people meaningfully track issues without the bug tracker?

Fortunately there are not too many issues to track, but you
are right that using the/a bug tracker more systematically
would help.

Yet the main problem, currently, is the work force.  We need
more hands :)