Re: push parser implemenation

[Bob Rossi]

On Thu, Sep 14, 2006 at 03:46:17PM -0400, Bob Rossi wrote:
> On Thu, Sep 14, 2006 at 12:44:00PM -0700, Paul Eggert wrote:
> > Bob Rossi <address@hidden> writes:
> > 
> > > Honestly, I don't understand why a pull parser needs this feature.
> > 
> > If nobody knows why it'd be useful then let's just omit it for now.
> > All we need to do is document the fact that the combination of
> > features is not supported.  Someone who actually needs the feature
> > will be a better judge of how it should be implemented.
> 
> OK, thank you. I'm going to incorporate all the changes we've discussed
> so far and post a patch to this list.

Here it is, this diff would have been much smaller if the file 
src/parse-gram.y wasn't re-bisoned.

Also note that push.c is attached separatly. A diff between it and
yacc.c will reveal that they are not that much different.

I've also attached the current version of the benchmark program that
I've been using.

The results of the testsuite and benchmark are below. 

Benchmarks:

            Rate push-push yacc-pure push-pure yacc-yacc push-yacc
push-push 2.93/s        --      -27%      -27%      -29%      -29%
yacc-pure 4.02/s       37%        --       -0%       -2%       -2%
push-pure 4.04/s       38%        0%        --       -2%       -2%
yacc-yacc 4.12/s       40%        2%        2%        --       -0%
push-yacc 4.12/s       40%        2%        2%        0%        --

Note that there is no time delay between yacc and push skeletons.

Testsuite run with push.c. (ie. cp push.c yacc.c before running):
    90. Calculator %push-parser %locations (calc.at:564): ok
    96. Calculator %push-parser %error-verbose %debug %locations 
        %defines %name-pref

    179 tests were successful.
    2 tests were skipped.

Please let me know what the next steps will be.

Thanks,
Bob Rossi

push.diff
Description: Text document

push.c
Description: Text Data

bench-yacc.pl
Description: Text Data