On Sat, Nov 17, 2001 at 01:49:43PM +0100, H.Merijn Brand wrote:
> On Sat 17 Nov 2001 01:05, Jarkko Hietaniemi <[EMAIL PROTECTED]> wrote:
> > On Fri, Nov 16, 2001 at 11:53:41PM +0000, Nicholas Clark wrote:
> > > On Sat, Nov 17, 2001 at 01:05:14AM +0200, Jarkko Hietaniemi wrote:
> > > > Never mind the above... the overall difference doesn't seem to be that
> > > > significant. Yes, dropping the preservation speeds up arith, but not
> > > > by that much... then again, speeding up arith may help other parts.
> > > > Isn't benchmarking and tuning fun? Wait for the numbers and make
> > > > your own guestimate.
> > >
> > > As you say, swings and roundabouts
> >
> > ...and questions like: why does dropping the preservation code speed
> > up wantarray() by 11%, but slows down regexes by about 5%... (in alpha)
>
> Still need hppa benches for that?
Yes, seeing the following comparisons
32bit 5.6.1 vs 32bit blead
32bit blead vs 64bitall blead
32bit blead vs 32bit blead with -DNO_PERL_PRESERVE_IVUV
64bitall blead vs 64bitall blead with -DNO_PERL_PRESERVE_IVUV
would be nice, at least the first two.
> Personally I'm all in favour of /documenting/ stability, and (also
> personally)
It's a behavioral guarantee for which I think we would need Larry's say-so.
--
$jhi++; # http://www.iki.fi/jhi/
# There is this special biologist word we use for 'stable'.
# It is 'dead'. -- Jack Cohen
