Re: [Python-Dev] Python-Dev Digest, Vol 159, Issue 27
Hi Victor,
Thanks for the great contribution to the unified benchmark development! In addition to the OutReachy program that we are currently supporting, let us know how else we could help out in this effort.
Other than micros and benchmarking ideas, we'd also like to hear suggestions from the community on workload development around real world use cases, especially in the enterprise world, cloud computing, data analytics, machine learning, high performance computing, etc.
Thanks,
Peter
-----Original Message----- From: Python-Dev [mailto:python-dev-bounces+peter.xihong.wang=intel.com@python.org] On Behalf Of python-dev-request@python.org Sent: Thursday, October 20, 2016 9:00 AM To: python-dev@python.org Subject: Python-Dev Digest, Vol 159, Issue 27
Send Python-Dev mailing list submissions to python-dev@python.org
To subscribe or unsubscribe via the World Wide Web, visit https://mail.python.org/mailman/listinfo/python-dev or, via email, send a message with subject or body 'help' to python-dev-request@python.org
You can reach the person managing the list at python-dev-owner@python.org
When replying, please edit your Subject line so it is more specific than "Re: Contents of Python-Dev digest..."
Today's Topics:
1. Benchmarking Python and micro-optimizations (Victor Stinner) 2. Have I got my hg dependencies correct? (Skip Montanaro) 3. Re: Have I got my hg dependencies correct? (Skip Montanaro) 4. Re: Have I got my hg dependencies correct? (Victor Stinner) 5. Re: Benchmarking Python and micro-optimizations (Maciej Fijalkowski) 6. Re: Benchmarking Python and micro-optimizations (Eric Snow)
----------------------------------------------------------------------
Message: 1 Date: Thu, 20 Oct 2016 12:56:06 +0200 From: Victor Stinner victor.stinner@gmail.com To: Python Dev Python-Dev@python.org Subject: [Python-Dev] Benchmarking Python and micro-optimizations Message-ID: CAMpsgwYsAXJgHfJH_hDH6pj13pHacMWYbi_LExLoXdcoUo6F-Q@mail.gmail.com Content-Type: text/plain; charset=UTF-8
Hi,
Last months, I worked a lot on benchmarks. I ran benchmarks, analyzed results in depth (up to the hardware and kernel drivers!), I wrote new tools and enhanced existing tools.
* I wrote a new perf module which runs benchmarks in a reliable way and contains a LOT of features: collect metadata, JSON file format, commands to compare, render an histogram, etc.
* I rewrote the Python benchmark suite: the old benchmarks Mercurial repository moved to a new performance GitHub project which uses my perf module and contains more benchmarks.
* I also made minor enhancements to timeit in Python 3.7 -- some dev don't want major changes to not "break the backward compatibility".
For timeit, I suggest to use my perf tool which includes a reliable timeit command and has much more features like --duplicate (repeat the statements to reduce the cost of the outer loop) and --compare-to (compare two versions of Python), but also all builtin perf features (JSON output, statistics, histogram, etc.).
I added benchmarks from PyPy and Pyston benchmark suites to performance: performance 0.3.1 contains 51 benchmark scripts which run a total of 121 benchmarks. Example of tested Python modules:
* SQLAlchemy * Dulwich (full Git implementation in Python) * Mercurial (currently only the startup time) * html5lib * pyaes (AES crypto cipher in pure Python) * sympy * Tornado (HTTP client and server) * Django (sadly, only the template engine right now, Pyston contains HTTP benchmarks) * pathlib * spambayes
More benchmarks will be added later. It would be nice to add benchmarks on numpy for example, numpy is important for a large part of our community.
All these (new or updated) tools can now be used to take smarter decisions on optimizations. Please don't push any optimization anymore without providing reliable benchmark results!
My first major action was to close the latest attempt to micro-optimize int+int in Python/ceval.c, http://bugs.python.org/issue21955 : I closed the issue as rejected, because there is no significant speedup on benchmarks other than two (tiny) microbenchmarks. To make sure that no one looses its time on trying to micro-optimize int+int, I even added a comment to Python/ceval.c :-)
https://hg.python.org/cpython/rev/61fcb12a9873 "Please don't try to micro-optimize int+int"
The perf and performance are now well tested: Travis CI runs tests on the new commits and pull requests, and the "tox" command can be used locally to test different Python versions, pep8, doc, ... in a single command.
Next steps:
* Run performance 0.3.1 on speed.python.org: the benchmark runner is currently stopped (and still uses the old benchmarks project). The website part may be updated to allow to download full JSON files which includes *all* information (all timings, metadata and more).
* I plan to run performance on CPython 2.7, CPython 3.7, PyPy and PyPy 3. Maybe also CPython 3.5 and CPython 3.6 if they don't take too much resources.
* Later, we can consider adding more implementations of Python: Jython, IronPython, MicroPython, Pyston, Pyjion, etc. All benchmarks should be run on the same hardware to be comparable.
* Later, we might also allow other projects to upload their own benchmark results, but we should find a solution to groups benchmark results per benchmark runner (ex: at least by the hostname, perf JSON contains the hostname) to not compare two results from two different hardware
* We should continue to add more benchmarks to the performance benchmark suite, especially benchmarks more representative of real applications (we have enough microbenchmarks!)
Links:
* perf: http://perf.readthedocs.io/ * performance: https://github.com/python/performance * Python Speed mailing list: https://mail.python.org/mailman/listinfo/speed * https://speed.python.org/ (currently outdated, and don't use performance yet)
See https://pypi.python.org/pypi/performance which contains even more links to Python benchmarks (PyPy, Pyston, Numba, Pythran, etc.)
Victor
------------------------------
Message: 2 Date: Thu, 20 Oct 2016 06:47:42 -0500 From: Skip Montanaro skip.montanaro@gmail.com To: python-dev Dev python-dev@python.org Subject: [Python-Dev] Have I got my hg dependencies correct? Message-ID: CANc-5UwuWxYkN5LgOdDPEO5-7xYDCLK5Xv6dFP=sLBaxbH1K9g@mail.gmail.com Content-Type: text/plain; charset=UTF-8
I've recently run into a problem building the math and cmath modules for 2.7. (I don't rebuild very often, so this problem might have been around for awhile.) My hg repos look like this:
* My cpython repo pulls from https://hg.python.org/cpython
* My 2.7 repo (and other non-tip repos) pulls from my cpython repo
I think this setup was recommended way back in the day when hg was new to the Python toolchain to avoid unnecessary network bandwidth.
So, if I execute
hg pull hg update
in first cpython, then 2.7 repos I should be up-to-date, correct? However, rebuilding in my 2.7 repo fails to build math and cmath. The compiler complains that Modules/_math.o doesn't exist. If I manually execute
make Modules/_math.o make
after the failure, then the math and cmath modules build.
Looking on bugs.python.org I saw this closed issue:
http://bugs.python.org/issue24421
which seems related. Is it possible that the fix wasn't propagated to the 2.7 branch? Or perhaps I've fouled up my hg repo relationships? My other repos which depend on cpython (3.5, 3.4, 3.3, and 3.2) all build the math module just fine.
I'm running on an ancient MacBook Pro with OS X 10.11.6 (El Capitan) and XCode 8.0 installed.
Any suggestions?
Skip
------------------------------
Message: 3 Date: Thu, 20 Oct 2016 07:13:47 -0500 From: Skip Montanaro skip.montanaro@gmail.com To: python-dev Dev python-dev@python.org Subject: Re: [Python-Dev] Have I got my hg dependencies correct? Message-ID: CANc-5Uy+xE3OWaVDNy+EOchCPBhgiHmbetsBQxP9TLVGQo_8ZA@mail.gmail.com Content-Type: text/plain; charset=UTF-8
On Thu, Oct 20, 2016 at 6:47 AM, Skip Montanaro skip.montanaro@gmail.com wrote:
Is it possible that the fix wasn't propagated to the 2.7 branch? Or perhaps I've fouled up my hg repo relationships?
Either way, I went ahead and opened a ticket:
http://bugs.python.org/issue28487
S
------------------------------
Message: 4 Date: Thu, 20 Oct 2016 14:35:10 +0200 From: Victor Stinner victor.stinner@gmail.com To: Skip Montanaro skip.montanaro@gmail.com Cc: python-dev Dev python-dev@python.org Subject: Re: [Python-Dev] Have I got my hg dependencies correct? Message-ID: CAMpsgwa0eK0nMa2PZk7PjRJGTVu1c5yBuZP0ROH_dSaWH4PyMw@mail.gmail.com Content-Type: text/plain; charset=UTF-8
Are you on the 2.7 branch or the default branch?
You might try to cleanup your checkout:
hg up -C -r 2.7 make distclean hg purge # WARNING! it removes *all* files not tracked by Mercurial ./configure && make
You should also paste the full error message.
Victor
2016-10-20 13:47 GMT+02:00 Skip Montanaro skip.montanaro@gmail.com:
I've recently run into a problem building the math and cmath modules for 2.7. (I don't rebuild very often, so this problem might have been around for awhile.) My hg repos look like this:
My cpython repo pulls from https://hg.python.org/cpython
My 2.7 repo (and other non-tip repos) pulls from my cpython repo
I think this setup was recommended way back in the day when hg was new to the Python toolchain to avoid unnecessary network bandwidth.
So, if I execute
hg pull hg update
in first cpython, then 2.7 repos I should be up-to-date, correct? However, rebuilding in my 2.7 repo fails to build math and cmath. The compiler complains that Modules/_math.o doesn't exist. If I manually execute
make Modules/_math.o make
after the failure, then the math and cmath modules build.
Looking on bugs.python.org I saw this closed issue:
http://bugs.python.org/issue24421
which seems related. Is it possible that the fix wasn't propagated to the 2.7 branch? Or perhaps I've fouled up my hg repo relationships? My other repos which depend on cpython (3.5, 3.4, 3.3, and 3.2) all build the math module just fine.
I'm running on an ancient MacBook Pro with OS X 10.11.6 (El Capitan) and XCode 8.0 installed.
Any suggestions?
Skip _______________________________________________ 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%40gm ail.com
------------------------------
Message: 5 Date: Thu, 20 Oct 2016 15:15:43 +0200 From: Maciej Fijalkowski fijall@gmail.com To: Victor Stinner victor.stinner@gmail.com Cc: Python Dev Python-Dev@python.org Subject: Re: [Python-Dev] Benchmarking Python and micro-optimizations Message-ID: CAK5idxRK1Muf5X8OWot2pkUCvvAG2SgU6D+LDKnnaeUH3P7erA@mail.gmail.com Content-Type: text/plain; charset=UTF-8
Hi Victor
Despite the fact that I was not able to find time to run your stuff yet, thanks for all the awesome work!
On Thu, Oct 20, 2016 at 12:56 PM, Victor Stinner victor.stinner@gmail.com wrote:
Hi,
Last months, I worked a lot on benchmarks. I ran benchmarks, analyzed results in depth (up to the hardware and kernel drivers!), I wrote new tools and enhanced existing tools.
- I wrote a new perf module which runs benchmarks in a reliable way
and contains a LOT of features: collect metadata, JSON file format, commands to compare, render an histogram, etc.
- I rewrote the Python benchmark suite: the old benchmarks Mercurial
repository moved to a new performance GitHub project which uses my perf module and contains more benchmarks.
- I also made minor enhancements to timeit in Python 3.7 -- some dev
don't want major changes to not "break the backward compatibility".
For timeit, I suggest to use my perf tool which includes a reliable timeit command and has much more features like --duplicate (repeat the statements to reduce the cost of the outer loop) and --compare-to (compare two versions of Python), but also all builtin perf features (JSON output, statistics, histogram, etc.).
I added benchmarks from PyPy and Pyston benchmark suites to performance: performance 0.3.1 contains 51 benchmark scripts which run a total of 121 benchmarks. Example of tested Python modules:
- SQLAlchemy
- Dulwich (full Git implementation in Python)
- Mercurial (currently only the startup time)
- html5lib
- pyaes (AES crypto cipher in pure Python)
- sympy
- Tornado (HTTP client and server)
- Django (sadly, only the template engine right now, Pyston contains
HTTP benchmarks)
- pathlib
- spambayes
More benchmarks will be added later. It would be nice to add benchmarks on numpy for example, numpy is important for a large part of our community.
All these (new or updated) tools can now be used to take smarter decisions on optimizations. Please don't push any optimization anymore without providing reliable benchmark results!
My first major action was to close the latest attempt to micro-optimize int+int in Python/ceval.c, http://bugs.python.org/issue21955 : I closed the issue as rejected, because there is no significant speedup on benchmarks other than two (tiny) microbenchmarks. To make sure that no one looses its time on trying to micro-optimize int+int, I even added a comment to Python/ceval.c :-)
https://hg.python.org/cpython/rev/61fcb12a9873 "Please don't try to micro-optimize int+int"
The perf and performance are now well tested: Travis CI runs tests on the new commits and pull requests, and the "tox" command can be used locally to test different Python versions, pep8, doc, ... in a single command.
Next steps:
- Run performance 0.3.1 on speed.python.org: the benchmark runner is
currently stopped (and still uses the old benchmarks project). The website part may be updated to allow to download full JSON files which includes *all* information (all timings, metadata and more).
- I plan to run performance on CPython 2.7, CPython 3.7, PyPy and PyPy
- Maybe also CPython 3.5 and CPython 3.6 if they don't take too much
resources.
- Later, we can consider adding more implementations of Python:
Jython, IronPython, MicroPython, Pyston, Pyjion, etc. All benchmarks should be run on the same hardware to be comparable.
- Later, we might also allow other projects to upload their own
benchmark results, but we should find a solution to groups benchmark results per benchmark runner (ex: at least by the hostname, perf JSON contains the hostname) to not compare two results from two different hardware
- We should continue to add more benchmarks to the performance
benchmark suite, especially benchmarks more representative of real applications (we have enough microbenchmarks!)
Links:
- perf: http://perf.readthedocs.io/
- performance: https://github.com/python/performance
- Python Speed mailing list:
https://mail.python.org/mailman/listinfo/speed
- https://speed.python.org/ (currently outdated, and don't use
performance yet)
See https://pypi.python.org/pypi/performance which contains even more links to Python benchmarks (PyPy, Pyston, Numba, Pythran, etc.)
Victor _______________________________________________ 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/fijall%40gmail.com
------------------------------
Message: 6 Date: Thu, 20 Oct 2016 09:31:42 -0600 From: Eric Snow ericsnowcurrently@gmail.com To: Victor Stinner victor.stinner@gmail.com Cc: Python Dev Python-Dev@python.org Subject: Re: [Python-Dev] Benchmarking Python and micro-optimizations Message-ID: CALFfu7AZsqrazOvfSDMMruqqTumUvxELWrijV08uAsgnbRnwfw@mail.gmail.com Content-Type: text/plain; charset=UTF-8
On Thu, Oct 20, 2016 at 4:56 AM, Victor Stinner victor.stinner@gmail.com wrote:
Hi,
Last months, I worked a lot on benchmarks. I ran benchmarks, analyzed results in depth (up to the hardware and kernel drivers!), I wrote new tools and enhanced existing tools.
This is a massive contribution. Thanks!
All these (new or updated) tools can now be used to take smarter decisions on optimizations. Please don't push any optimization anymore without providing reliable benchmark results!
+1
-eric
------------------------------
Subject: Digest Footer
_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev
------------------------------
End of Python-Dev Digest, Vol 159, Issue 27 *******************************************
participants (1)
-
Wang, Peter Xihong