Front page | perl.perl5.porters |
Postings from February 2003
On Sun, Feb 23, 2003 at 03:31:38AM +0000, Ton Hospel wrote:
> - if the value isn't going to change (much), why did the user use a (??{$p}) ?
> What is the scenario in which you expect the cache to be hit ?
Please do something like
dd if=/dev/zero of=/tmp/k count=4096
and run the script below on it with/without the optimization.
( Dropping the optimization is matter of commenting out the branch
if (!(SvFLAGS(ret) ...
at regexec.c:2892 - line number of the unpatched file ).
I get:
ratsnest# time perl /tmp/x /tmp/k # without the optimization
real 0m54.794s
user 0m54.752s
sys 0m0.041s
ratsnest# time perl /tmp/x /tmp/k # with the optimization in
real 0m4.332s
user 0m4.316s
sys 0m0.016s
Notice that instead of the silly quotes there the user may want to put
some convoluted regular expressions, without putting it in qr// for
some reason or another - these may be build up by some sub, etc.
> - why is the r-magic not invalidated when the variable gets assigned ?
That's just what my patch try to do: invalidate the r-magic when
the variable gets assigned.
( I admit the test is silly - I suspect however that a more serious
investigation will give the same results).
Regards
Adi
/tmp/x
--------------------------------------------
$entremets = "
Tancowala ryba z rakiem, a pietruszka z pasternakiem.
Cebula sie dziwowala, jak pietruszka tancowala.
|#";
$appero = "
Pisse une goutte dans ce vermouth,
c'est pas beau mais c'est salaud.
|%";
$entremets = "a";
$appero = "b";
$/=\4096;
while(<>) {
$a = not $a;
print if /(??{
$a ? $entremets : $appero;
})/
}
Thread Previous
|
Thread Next