More performance analysis

[Michael G Schwern]

Eric Wilhelm wrote:
> # from Michael G Schwern
> # on Friday 14 September 2007 03:52 pm:
> 
>> Anyhow, looking at how many times we call each routine it seems some
>> clever memoization might help a lot.
> 
> I don't think "memoization" applies unless the results are the same.  
> What's needed could maybe be best expressed as a source filter where we 
> create (even at `build dist` time) TAP::Parser::Fast as a stripped-down 
> version with all of the options turned off.  (That would mean 
> memoization does (sort of) apply at _get_parser_args() or 
> TAP::Parser->new or thereabouts.)

What about a pure Perl source filter to inline subroutine calls?

Hmm...


-- 
The interface should be as clean as newly fallen snow and its behavior
as explicit as Japanese eel porn.