[Python-Dev] Computed Goto dispatch for Python 2
Gregory P. Smith
greg at krypto.org
Thu May 28 07:46:59 CEST 2015
Why now? We intentionally decided not to do this for 2.7 in the past
because it was too late for the release cutoff.
Has anyone benchmarked compiling python in profile-opt mode vs having the
computed goto patch? I'd *expect* the benefits to be the roughly the same.
Has this been compared to that? (Anyone not compiling their Python
interpreter in profile-opt mode is doing themselves a major disservice.)
Does it shows noteworthy improvements even when used with a profile-opt
build using a PROFILE_TASK of regrtest.py with the same test exclusion list
as the debian python2.7 package?
Could you please rerun benchmarks including the profile-opt build with and
without the patch for comparsion.
PS I recommend attaching the up to date patch against 2.7.10 to issue4753.
That is where anyone will go looking for it, not buried in a mailing list
On Wed, May 27, 2015 at 10:01 PM Terry Reedy <tjreedy at udel.edu> wrote:
> On 5/27/2015 9:31 PM, Nick Coghlan wrote:
> > +1 from me, for basically the same reasons Guido gives: Python 2.7 is
> > going to be with us for a long time, and this particular change
> > shouldn't have any externally visible impacts at either an ABI or API
> Immediately after a release, giving the patch plenty of time to be be
> tested, seems like a good time.
> Terry Jan Reedy
> Python-Dev mailing list
> Python-Dev at python.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Python-Dev