Front page | perl.perl5.porters |
Postings from May 2006
On Mon, May 01, 2006 at 07:55:21PM +0100, Dave Mitchell wrote:
> I should imagine it's copying the scratchpad that takes most of the time.
> Hmm, perhaps we could just share the scratchpads for non-closure anon
> subs?
I've knocked up a proof-of-concept patch (it fails loads of tests, but
works for basic code).
The timings I get are as follows:
orig perl (share CV and scratchpad):
0m2.292s sub {} for 1..10_000_000
hacked perl (copy CV, but share scratchpad):
0m6.564s sub {} for 1..10_000_000
orig perl, but full closure (CV and scratchpad copied):
0m18.229s my $x; sub {$x} for 1..10_000_000
However, there's a problem with copying the CV while sharing the pad:
If one of the sub calls another copy of the sub, then they share the
same set of lexicals and temps. The following contrived example shows
how the inner call corrupts the outer sub's $x:
push @s, sub { my $x=1; $s[0] && &{pop @s}; print "x=$x\n" } for 1,2;
&{pop @s};
which gives:
x=1
x=
whereas in normal perl it gives
x=1
x=1
this is becuase with the CV being shared, they share CvDEPTH and thus
when one sub calls the other, CvDEPTH gets incremented and so they use
different levels in the pad.
--
But Pity stayed his hand. "It's a pity I've run out of bullets",
he thought. - "Bored of the Rings"
Thread Previous
|
Thread Next