On Mon, Feb 08, 2010 at 02:54:57PM +0000, Nicholas Clark wrote:
> On Sat, Feb 06, 2010 at 09:24:03PM +0000, Tim Bunce wrote:
> >
> > Well, the buffer could be made larger :)
>
> Yes, but make something foolproof and they build a better fool.
:)
> I made it dynamically resize (r1045)
Cool.
> But joining the filenames isn't *that* useful when you're merging 538 runs.
> (See attached screengrab, intentionally shrunk to be less legible)
By 'eck that's long!
How long does it take to merge them? (And are you using stmts=0?)
> > More usefully, it would be good if the final report could show
> > attributes from each of the files that were merged.
> > So perhaps attributes could include an index:
> > $profile->attribute->[0]->{application}
> >
> > The [0]'th element would be original attributes in the case of an
> > unmerged file. For a merged file they could be the "smart merged"
> > 'summary' that you currently generate. The 'application' attribute
> > could then simply say something like "(merge of N files)".
> >
> > Just a thought.
>
> Yes, but I can't see how to do that inside the existing file format:
I don't have a major objection to changing the file format.
I doubt many people are inconvenienced by it.
> > Another alternative would be to take some things that are currently
> > attributes and move them to become fields of NYTP_TAG_PID_START.
>
> I suspect that this is the way, but that would be (another) file format
> change, and presumably another major version bump.
Major version bump for the file format perhaps but not for NYTProf
itself. I'd like to save v4.0 for abstract clocks & memory profiling :)
> There's a bit more I can do within the current format.
>
> (Although it feels like I'm close to where I'd suggest that a release would
> be good. For example r1040 might give a small speedup in all profiling)
Yes, that and assorted other optimizations. Lots of people will also be
grateful for the 100ns timer code from Markus Peter.
I'd like to finish off the usecputime fix that not quite complete
(still gives figures that are x10 on my Mac.)
I'd also like to see some regression tests for nytprofmerge.
Tim.
--
You've received this message because you are subscribed to
the Devel::NYTProf Development User group.
Group hosted at: http://groups.google.com/group/develnytprof-dev
Project hosted at: http://perl-devel-nytprof.googlecode.com
CPAN distribution: http://search.cpan.org/dist/Devel-NYTProf
To post, email: [email protected]
To unsubscribe, email: [email protected]
