Tels wrote:
> attached is the patch, and a testversion of t/uni/class.t (so that you can
> toy around with it).
Thanks, applied.
I just changed this part :
+ is ('first N-1 chars did not match', $blk);
(and the similar one below) to use the fail() function of test.pl instead.
> The testversion contains two subroutines, one that constructs the qr//
> objects in the subroutine, and one that takes them as arguments. The
> latter is slower, so the final patch contains the other variant.
>
> I have no idea what's the difference. In any event, the patch changes two
> more places to use the subroutine, but on my system the timings do not
> change in a measurable way for these, only the very last loop is affacted
> by the changes.
>
> On my system (2.0 Ghz AMD XP 2400+) the time goes down from about 31
> seconds to about 19 seconds after the patch.
I've similar figures.
