Re: [Python-Dev] OS-X builds for 3.7.0
On Jan 30, 2018 6:47 PM, "Joni Orponen" <j.orponen@4teamwork.ch> wrote: On Tue, Jan 30, 2018 at 6:50 PM, Ray Donnelly <mingw.android@gmail.com> wrote:
While we're making such macOS-build requests, any chance of building a static interpreter too? We've been doing that on the Anaconda Distribution since the 5.0 release in September and it seems to be working well.
PyPy is also currently eyeing doing their macOS builds better: https://bitbucket.org/pypy/pypy/issues/2734/establish-a-build-and-release- pipeline-for What do the Anaconda static builds get built on? We have our own clang pseudo cross-compilers and use a macOS 10.9 SDK for all of our package compilation to achieve compatibility (this means we can compile on newer macOS just fine). We see a 1.1 to 1.2 times performance benefit over official releases as measured using 'python performance'. Apart from a static interpreter we also enable LTO and PGO and only build for 64-bit so I'm not sure how much each bit continues. Our recipe for python 3.6 can be found at: https://github.com/AnacondaRecipes/python-feedstock/tree/master/recipe -- Joni Orponen _______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/ mingw.android%40gmail.com
On Jan 31, 2018 8:31 AM, "Ray Donnelly" <mingw.android@gmail.com> wrote: On Jan 30, 2018 6:47 PM, "Joni Orponen" <j.orponen@4teamwork.ch> wrote: On Tue, Jan 30, 2018 at 6:50 PM, Ray Donnelly <mingw.android@gmail.com> wrote:
While we're making such macOS-build requests, any chance of building a static interpreter too? We've been doing that on the Anaconda Distribution since the 5.0 release in September and it seems to be working well.
PyPy is also currently eyeing doing their macOS builds better: https://bitbucket.org/pypy/pypy/issues/2734/establis h-a-build-and-release-pipeline-for What do the Anaconda static builds get built on? We have our own clang pseudo cross-compilers and use a macOS 10.9 SDK for all of our package compilation to achieve compatibility (this means we can compile on newer macOS just fine). We see a 1.1 to 1.2 times performance benefit over official releases as measured using 'python performance'. Apart from a static interpreter we also enable LTO and PGO and only build for 64-bit so I'm not sure how much each bit continues. Our recipe for python 3.6 can be found at: s/continues/contributes/ https://github.com/AnacondaRecipes/python-feedstock/tree/master/recipe -- Joni Orponen _______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/mingw. android%40gmail.com
On Wed, Jan 31, 2018 at 9:31 AM, Ray Donnelly <mingw.android@gmail.com> wrote:
We see a 1.1 to 1.2 times performance benefit over official releases as measured using 'python performance'.
Apart from a static interpreter we also enable LTO and PGO and only build for 64-bit so I'm not sure how much each bit continues. Our recipe for python 3.6 can be found at:
Do you metrify LTO and PGO independent of each other as well or only the "enable everything" combo? I've had mixed results with LTO so far, but this is probably hardware / compiler combination specific. -- Joni Orponen
On Wed, Jan 31, 2018 at 11:16 AM, Joni Orponen <j.orponen@4teamwork.ch> wrote:
On Wed, Jan 31, 2018 at 9:31 AM, Ray Donnelly <mingw.android@gmail.com> wrote:
We see a 1.1 to 1.2 times performance benefit over official releases as measured using 'python performance'.
Apart from a static interpreter we also enable LTO and PGO and only build for 64-bit so I'm not sure how much each bit continues. Our recipe for python 3.6 can be found at:
Do you metrify LTO and PGO independent of each other as well or only the "enable everything" combo? I've had mixed results with LTO so far, but this is probably hardware / compiler combination specific.
I've never found enough time to take detailed metrics, sorry. Maybe one day? Looking at my performance graphs again: Against the official CPython 3.6 (probably .3 or .4) release I see: 1 that is 2.01x faster (python-startup, 24.6ms down to 12.2ms) 5 that are >=1.5x,<1.6x faster. 13 that are >=1.4x,<1.5x faster. 21 that are >=1.3x,<1.4x faster. 14 that are >=1.2x,<1.3x faster. 5 that are >=1.1x,<1.2x faster. 0 that are < 1.1x faster/slower. Pretty good numbers overall I think.
-- Joni Orponen
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/mingw.android%40gmail.com
Against the official CPython 3.6 (probably .3 or .4) release I see: 1 that is 2.01x faster (python-startup, 24.6ms down to 12.2ms) 5 that are >=1.5x,<1.6x faster. 13 that are >=1.4x,<1.5x faster. 21 that are >=1.3x,<1.4x faster. 14 that are >=1.2x,<1.3x faster. 5 that are >=1.1x,<1.2x faster. 0 that are < 1.1x faster/slower.
Pretty good numbers overall I think.
Yay!! Congrats for all of us! -- INADA Naoki <songofacandy@gmail.com>
On Wed, Jan 31, 2018 at 4:20 AM, INADA Naoki <songofacandy@gmail.com> wrote:
Against the official CPython 3.6 (probably .3 or .4) release I see: 1 that is 2.01x faster (python-startup, 24.6ms down to 12.2ms) 5 that are >=1.5x,<1.6x faster. 13 that are >=1.4x,<1.5x faster. 21 that are >=1.3x,<1.4x faster. 14 that are >=1.2x,<1.3x faster. 5 that are >=1.1x,<1.2x faster. 0 that are < 1.1x faster/slower.
Pretty good numbers overall I think.
Yay!! Congrats for all of us!
I'm confused -- I _think_ these are performance improvements of the Anaconda build over the python.org build for OS-X -- so congrats to the Anaconda team :-) But a hint that maybe we should do the python.org builds differently! -CHB -- Christopher Barker, Ph.D. Oceanographer Emergency Response Division NOAA/NOS/OR&R (206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception Chris.Barker@noaa.gov
On 1/31/2018 6:23 PM, Chris Barker wrote:
On Wed, Jan 31, 2018 at 4:20 AM, INADA Naoki <songofacandy@gmail.com <mailto:songofacandy@gmail.com>> wrote:
> Against the official CPython 3.6 (probably .3 or .4) release I see: > 1 that is 2.01x faster (python-startup, 24.6ms down to 12.2ms) > 5 that are >=1.5x,<1.6x faster. > 13 that are >=1.4x,<1.5x faster. > 21 that are >=1.3x,<1.4x faster. > 14 that are >=1.2x,<1.3x faster. > 5 that are >=1.1x,<1.2x faster. > 0 that are < 1.1x faster/slower. > > Pretty good numbers overall I think.
Yay!! Congrats for all of us!
I'm confused -- I _think_ these are performance improvements of the Anaconda build over the python.org <http://python.org> build for OS-X -- so congrats to the Anaconda team :-)
But a hint that maybe we should do the python.org <http://python.org> builds differently!
Ned Deily is in charge of the Mac build (as well as current release manager). Within the last week, he revised the official builds (now two, I believe) for 3.7.0b1, due in a day or so. One will be a future oriented 64-bit build. The PR and What's New have more. He may not be reading this thread, but will read MacOS tracker issues with a specific proposal, data and a patch. Comparisons should be against the current master or an installed 3.7.0b1. -- Terry Jan Reedy
Ned Deily is in charge of the Mac build (as well as current release manager). Within the last week, he revised the official builds (now two, I believe) for 3.7.0b1, due in a day or so. One will be a future oriented 64-bit build. The PR and What's New have more.
What's New doesn't mention it, but Ned's annoucemtn does: """ Attention macOS users: with 3.7.0b1, we are providing a choice of two binary installers. The new variant provides a 64-bit-only version for macOS 10.9 and later systems; this variant also now includes its own built-in version of Tcl/Tk 8.6. We welcome your feedback. """ So that's a start -- thanks Ned! He may not be reading this thread, but will read MacOS tracker issues with
a specific proposal, data and a patch. Comparisons should be against the current master or an installed 3.7.0b1.
I hope the folks on this thread that know what they are doing can test and make suggestions :-) -CHB -- Christopher Barker, Ph.D. Oceanographer Emergency Response Division NOAA/NOS/OR&R (206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception Chris.Barker@noaa.gov
There is https://speed.python.org/comparison/ to compare Python 2.7, 3.5, 3.6 and master (future 3.7). Victor Le 31 janv. 2018 13:14, "Ray Donnelly" <mingw.android@gmail.com> a écrit :
On Wed, Jan 31, 2018 at 9:31 AM, Ray Donnelly <mingw.android@gmail.com> wrote:
We see a 1.1 to 1.2 times performance benefit over official releases as measured using 'python performance'.
Apart from a static interpreter we also enable LTO and PGO and only
build
for 64-bit so I'm not sure how much each bit continues. Our recipe for python 3.6 can be found at:
Do you metrify LTO and PGO independent of each other as well or only the "enable everything" combo? I've had mixed results with LTO so far, but
On Wed, Jan 31, 2018 at 11:16 AM, Joni Orponen <j.orponen@4teamwork.ch> wrote: this
is probably hardware / compiler combination specific.
I've never found enough time to take detailed metrics, sorry. Maybe one day? Looking at my performance graphs again:
Against the official CPython 3.6 (probably .3 or .4) release I see: 1 that is 2.01x faster (python-startup, 24.6ms down to 12.2ms) 5 that are >=1.5x,<1.6x faster. 13 that are >=1.4x,<1.5x faster. 21 that are >=1.3x,<1.4x faster. 14 that are >=1.2x,<1.3x faster. 5 that are >=1.1x,<1.2x faster. 0 that are < 1.1x faster/slower.
Pretty good numbers overall I think.
-- Joni Orponen
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/
mingw.android%40gmail.com
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/ victor.stinner%40gmail.com
The horizontal axis labelling in that graph is useless with so many tests included! Would a graphic with hover labels over the bars be more useful? Steve Holden On Wed, Jan 31, 2018 at 1:06 PM, Victor Stinner <victor.stinner@gmail.com> wrote:
There is https://speed.python.org/comparison/ to compare Python 2.7, 3.5, 3.6 and master (future 3.7).
Victor
Le 31 janv. 2018 13:14, "Ray Donnelly" <mingw.android@gmail.com> a écrit :
On Wed, Jan 31, 2018 at 9:31 AM, Ray Donnelly <mingw.android@gmail.com> wrote:
We see a 1.1 to 1.2 times performance benefit over official releases as measured using 'python performance'.
Apart from a static interpreter we also enable LTO and PGO and only
build
for 64-bit so I'm not sure how much each bit continues. Our recipe for python 3.6 can be found at:
Do you metrify LTO and PGO independent of each other as well or only the "enable everything" combo? I've had mixed results with LTO so far, but
On Wed, Jan 31, 2018 at 11:16 AM, Joni Orponen <j.orponen@4teamwork.ch> wrote: this
is probably hardware / compiler combination specific.
I've never found enough time to take detailed metrics, sorry. Maybe one day? Looking at my performance graphs again:
Against the official CPython 3.6 (probably .3 or .4) release I see: 1 that is 2.01x faster (python-startup, 24.6ms down to 12.2ms) 5 that are >=1.5x,<1.6x faster. 13 that are >=1.4x,<1.5x faster. 21 that are >=1.3x,<1.4x faster. 14 that are >=1.2x,<1.3x faster. 5 that are >=1.1x,<1.2x faster. 0 that are < 1.1x faster/slower.
Pretty good numbers overall I think.
-- Joni Orponen
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/mingw.
android%40gmail.com
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/victor. stinner%40gmail.com
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/ steve%40holdenweb.com
Click on "[x] horizontal" to exchange the two axis ;-) Victor 2018-01-31 16:08 GMT+01:00 Steve Holden <steve@holdenweb.com>:
The horizontal axis labelling in that graph is useless with so many tests included!
Would a graphic with hover labels over the bars be more useful?
Steve Holden
On Wed, Jan 31, 2018 at 1:06 PM, Victor Stinner <victor.stinner@gmail.com> wrote:
There is https://speed.python.org/comparison/ to compare Python 2.7, 3.5, 3.6 and master (future 3.7).
Victor
Le 31 janv. 2018 13:14, "Ray Donnelly" <mingw.android@gmail.com> a écrit :
On Wed, Jan 31, 2018 at 11:16 AM, Joni Orponen <j.orponen@4teamwork.ch> wrote:
On Wed, Jan 31, 2018 at 9:31 AM, Ray Donnelly <mingw.android@gmail.com> wrote:
We see a 1.1 to 1.2 times performance benefit over official releases as measured using 'python performance'.
Apart from a static interpreter we also enable LTO and PGO and only build for 64-bit so I'm not sure how much each bit continues. Our recipe for python 3.6 can be found at:
Do you metrify LTO and PGO independent of each other as well or only the "enable everything" combo? I've had mixed results with LTO so far, but this is probably hardware / compiler combination specific.
I've never found enough time to take detailed metrics, sorry. Maybe one day? Looking at my performance graphs again:
Against the official CPython 3.6 (probably .3 or .4) release I see: 1 that is 2.01x faster (python-startup, 24.6ms down to 12.2ms) 5 that are >=1.5x,<1.6x faster. 13 that are >=1.4x,<1.5x faster. 21 that are >=1.3x,<1.4x faster. 14 that are >=1.2x,<1.3x faster. 5 that are >=1.1x,<1.2x faster. 0 that are < 1.1x faster/slower.
Pretty good numbers overall I think.
-- Joni Orponen
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe:
https://mail.python.org/mailman/options/python-dev/mingw.android%40gmail.com
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/victor.stinner%40gmail.co...
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/steve%40holdenweb.com
participants (7)
-
Chris Barker -
INADA Naoki -
Joni Orponen -
Ray Donnelly -
Steve Holden -
Terry Reedy -
Victor Stinner