By mistake, mail was sent to Victor instead of speed. I was asking if the
benchmark code is specific to intel x86_64. When I am running this tool
(performance) on IBM Power machine, results are kind of promising enough as
multiple runs of this benchmark tool comes up with the same result and
benchmark seems to be stable.
On Thu, Jul 27, 2017 at 4:52 AM, Victor Stinner <victor.stinner(a)gmail.com>
wrote:
> Hum, you replied to me in private. I don't know if it was deliberate.
>
> 2017-07-26 19:39 GMT+02:00 Pintu Kumar <pintukumar79(a)gmail.com>:
> > Thanks Victor. I already tried with --fast and --rigorous option.
> Rigorous
> > spawns twice the number of worker processes. My result was almost same
> for
> > both options --fast as well as --rigorous.
>
> Oh ok, good.
>
> > I will look into third link you
> > have shared. I have one more query. Almost all the time you have
> discussed
> > about running this performance tool on Intel x86_64 architecture, even i
> saw
> > your video about how to tune systems to run benchmarks. I am running this
> > benchmark tool on IBM Power machine running ubuntu O.S. So, should I
> believe
> > the results what I get here on IBM Power machine.
>
> I focused on hardware that I have (access to), so Intel x86-64 CPUs.
> Sorry, I don't know IBM Power, I don't know how power management is
> handled on such CPU. Ask maybe on the speed mailing list?
>
> Victor
>
There is a --fast option to spawn less processes:
http://pyperformance.readthedocs.io/usage.html#run
But I don't suggest you to use it since it's less reliable ;-)
For me, it's really important to get stable benchmarks:
http://pyperformance.readthedocs.io/usage.html#how-to-get-stable-benchmarks
Most benchmarks can be run a standalone scripts: see
performance/benchmarks/ in the source code (performance.benchmarks
module). These scripts provide the full perf CLI, all options:
http://perf.readthedocs.io/en/latest/runner.html
Victor
2017-07-25 20:24 GMT+02:00 Pintu Kumar <pintukumar79(a)gmail.com>:
> Hi,
>
> I think performance is a kind of wrapper over perf if I am correct. As I
> come across thing that perf internally creates 20 worker process to run over
> benchmark program. So, is there a way to run any particular number of
> workers from command line of performance like perf.
>
> Thanks!
>
> _______________________________________________
> Speed mailing list
> Speed(a)python.org
> https://mail.python.org/mailman/listinfo/speed
>
Hi,
I think performance is a kind of wrapper over perf if I am correct. As I
come across thing that perf internally creates 20 worker process to run
over benchmark program. So, is there a way to run any particular number of
workers from command line of performance like perf.
Thanks!
Hi,
I released performance 0.6.0:
http://pyperformance.readthedocs.io/
This releaes uses the just released perf 1.4 which fix
parse_cpu_list() for the Linux setup of Xiang Zhang :-)
https://github.com/python/performance/issues/29
As any performance release, results produced by performance 0.6.0 are
incompatible with results produced by an older version. But in this
release, it's even more important because of a "fix" in bm_logging.py
benchmark: Logger.warning() is now used instead of Logger.warn(), but
warning() is faster!
Changes:
Version 0.6.0 (2017-07-06)
--------------------------
* Change ``warn`` to ``warning`` in `bm_logging.py`. In Python 3, Logger.warn()
calls warnings.warn() to log a deprecation warning, so is slower than
Logger.warning().
* Add again the ``logging_silent`` microbenchmark suite.
* compile command: update the Git repository before getting the revision
* Update requirements
- perf: 1.3 => 1.4 (fix parse_cpu_list(): strip also NUL characters)
- Django: 1.11.1 => 1.11.3
- mercurial: 4.2 => 4.2.2
- pathlib2: 2.2.1 => 2.3.0
- SQLAlchemy: 1.1.10 => 1.1.11
Victor