[Scons-dev] Python slots value (Was: catching up)

[Gary Oberbrunner]

We do have continuous measurements of SCons performance and memory use on a
selected set of test cases, at http://buildbot.scons.org/timings/.  I'm not
completely sure this is up to date; Bill, do you know?


On Tue, Aug 19, 2014 at 10:54 AM, Kenny, Jason L <jason.l.kenny at intel.com>
wrote:

> The problem for me is that we have been looking at this in our builds, but
> I cannot share much. We have made prototype tools to collect this
> information to show within our products. We have seen a number of issues in
> different area ( caching for example is a big one. Parts and SCons both
> "over" cache for speed, as such waste lots of memory and we may not be
> getting a real speed gain for it.) I know given a drop of SCons with the
> Slots I should be able to publish some results on what we saw. The issues
> is that I cannot give out our build to our products for open testing
> comparisons.
>
> I think we will want some case of an open source build that is large and
> "real" in that it has build dependency patterns that are real, vs something
> we might try to make up. I know the builds I have are large and real, but I
> cannot share them. I can share metrics at most. I sure there has to be
> something...
>
> Jason
>
> -----Original Message-----
> From: Scons-dev [mailto:scons-dev-bounces at scons.org] On Behalf Of anatoly
> techtonik
> Sent: Tuesday, August 19, 2014 8:57 AM
> To: SCons developer list
> Subject: [Scons-dev] Python slots value (Was: catching up)
>
> 1. Real build with probes that dumb memory and time statistics in an
> obvious way.
> 2. Patch that modifies SCons to use slots.
> 3. Part that allows to easily compare stats from both builds and add more
> info to comparison.
>
> On Tue, Aug 19, 2014 at 4:10 PM, Kenny, Jason L <jason.l.kenny at intel.com>
> wrote:
> > By this are we looking for a fake build that make usage of this memory
> usage, or the test to show the value of the __slots__ in python that can be
> run in different python impls to so the value of slots.
> >
> > The test case to show the value of slots is easy to remake ( or to find
> in the archives), and to post results for different python impls.
> >
> > Jason
> >
> > -----Original Message-----
> > From: Scons-dev [mailto:scons-dev-bounces at scons.org] On Behalf Of
> > anatoly techtonik
> > Sent: Tuesday, August 19, 2014 3:41 AM
> > To: SCons developer list
> > Subject: Re: [Scons-dev] catching up
> >
> > On Mon, Aug 18, 2014 at 8:17 PM, Kenny, Jason L <jason.l.kenny at intel.com>
> wrote:
> >> On the slots usage.
> >>
> >> I believe for me the value of slots was that it reduced memory a lot. I
> posted tests some time ago to show this. Ideally I would be a fan of python
> cleaning this up internally, but at the end of the day the __slots__ just
> seem to be a memory allocator tweak.
> >
> > Thanks for the reminder. Do we have a reproducible environment to repeat
> experiment and test with different conditions? It may become useful for
> PyPy optimizations if not for CPython itself.
> > _______________________________________________
> > Scons-dev mailing list
> > Scons-dev at scons.org
> > http://two.pairlist.net/mailman/listinfo/scons-dev
> > _______________________________________________
> > Scons-dev mailing list
> > Scons-dev at scons.org
> > http://two.pairlist.net/mailman/listinfo/scons-dev
>
>
>
> --
> anatoly t.
> _______________________________________________
> Scons-dev mailing list
> Scons-dev at scons.org
> http://two.pairlist.net/mailman/listinfo/scons-dev
> _______________________________________________
> Scons-dev mailing list
> Scons-dev at scons.org
> http://two.pairlist.net/mailman/listinfo/scons-dev
>



-- 
Gary
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://two.pairlist.net/pipermail/scons-dev/attachments/20140819/1200039c/attachment-0001.html>