Pyperf

--------------------------

Major change:

* Display ``Mean +- std dev`` instead of ``Median +- std dev``

Enhancements:

* Add a new ``Runner.bench_command()`` method to measure the execution time of
a command.
* Add ``mean()``, ``median_abs_dev()`` and ``stdev()`` methods to ``Benchmark``
* ``check`` command: test also minimum and maximum compared to the mean

Major API change, rename "sample" to "value":

* Rename attributes and methods:

- ``Benchmark.bench_sample_func()`` => ``Benchmark.bench_time_func()``.
- ``Run.samples`` => ``Run.values``
- ``Benchmark.get_samples()`` => ``Benchmark.get_values()``
- ``get_nsample()`` => ``get_nvalue()``
- ``Benchmark.format_sample()`` => ``Benchmark.format_value()``
- ``Benchmark.format_samples()`` => ``Benchmark.format_values()``

* Rename Runner command line options:

- ``--samples`` => ``--values``
- ``--debug-single-sample`` => ``--debug-single-value``

Changes:

* ``convert``: Remove ``--remove-outliers`` option
* ``check`` command now tests stdev/mean, instead of testing stdev/median
* setup.py: statistics dependency is now installed using ``extras_require`` to
support setuptools 18 and newer
* Add setup.cfg to enable universal builds: same wheel package for Python 2
and Python 3
* Add ``perf.VERSION`` constant: tuple of int
* JSON version 6: write metadata common to all benchmarks (common to all runs
of all benchmarks) at the root; rename 'samples' to 'values' in runs.

--------------------------

* Add ``--python-names`` option to the :ref:`Runner CLI <runner_cli>`
* ``system show`` command now checks if the system is ready for benchmarking
* Fix ``--compare-to`` option: the benchmark was run twice with the reference
Python, instead of being run first with reference Python and then changed
Python.
* Runner now raises an exception if a benchmark name is not unique.
* ``compare_to`` command now keeps the original order of benchmarks, only
sort if ``--by-speed`` option is used.
* Fix ``system`` command on macOS on non-existent ``/proc`` and ``/sys``
pseudo-files.
* Fix ``system`` bugs on systems with more than 32 processors.

--------------------------

New features:

* Add ``--compare-to`` option to the :ref:`Runner CLI <runner_cli>`
* :ref:`compare_to <compare_to_cmd>` command: Add ``--table`` option to render a table

Bugfixes:

* Fix the ``abs_executable()`` function used to find the absolute path to the
Python program. Don't follow symbolic links to support correctly virtual
environments.

--------------------------

* Fix the Windows support.
* system: Don't try to read or write CPU frequency when the
/sys/devices/system/cpu/cpu0/cpufreq/ directory doesn't exist. For example,
virtual machines don't have this directory.
* Fix a ``ResourceWarning`` in ``BenchmarkSuite.dump()`` for gzip files.

--------------------------

* Issue 15: Added ``--no-locale`` command line option and locale environment
variables are now inherited by default.
* Add :meth:`Runner.timeit` method.
* Fix ``stats`` command: display again statistics on the whole benchmark suite.
* Fix a ResourceWarning if interrupted: Runner now kills the worker process
when interrupted.
* ``compare`` and ``compare_to``: add percent difference to faster/slower
* Rewrite timeit internally: copy code from CPython 3.7 and adapt it to
PyPy.

--------------------------

* ``system tune`` now also sets the maximum sample rate of perf event.
* ``system show`` command now also displays advices, not only ``system tune``
* ``system`` now detects when running on a laptop with the power cable
unplugged.
* ``system tune`` now handles errors when /dev/cpu/N/msr device is missing:
log an error suggesting to load the ``msr`` kernel module
* Fix a ResourceWarning in Runner._spawn_worker_suite(): wait until the worker
completes.