Add notes about benchmarks.

Author: ionelmc

README.rst

@@ -71,3 +71,57 @@ Development
 To run the all tests run::
 
     tox
+
+Benchmarks
+==========
+
+The tested implementations:
+
+They include these less complete proxy implementations:
+
+* ``lazy_object_proxy.slots.Proxy``, available as ``lazy_object_proxy.Proxy`` if the C extension is not available.
+* ``lazy_object_proxy.cext.Proxy``, available as ``lazy_object_proxy.Proxy``.
+* `objproxies <https://pypi.python.org/pypi/objproxies>`_ - fork of PJE's `ProxyTypes <https://pypi.python.org/pypi/ProxyTypes>`_ with Python 3 support. [1]_
+* `SimpleLazyObject <https://github.com/django/django/blob/stable/1.7.x/django/utils/functional.py#L337>`_ from Django. [1]_
+* ``lazy_object_proxy.simple.Proxy`` - uses the non-data descriptor `trick <http://blog.ionelmc.ro/2014/11/04/an-interesting-python-descriptor-quirk/>`_. [1]_
+
+For Python 2.7::
+
+    --- benchmark: 5 tests, min 5 rounds (of min 25.00us), 5.00s max time, timer: time.clock ---
+    Name (time in ns)                    Min        Max       Mean    StdDev  Rounds  Iterations
+    --------------------------------------------------------------------------------------------
+    test_perf[slots]                705.3837  3091.0072   720.4805   27.2588  180453          38
+    test_perf[cext]                  93.6346   278.9731    95.1876    2.5474  164373         312
+    test_perf[simple]               398.1636  1863.2017   405.4325   13.1411  207521          59
+    test_perf[django]               471.6515   983.0809   480.4264   10.2059  193043          53
+    test_perf[objproxies]          1204.7003  4818.8010  1475.3419  114.8213   55898           1
+    --------------------------------------------------------------------------------------------
+
+For Python 3.4::
+
+    --- benchmark: 5 tests, min 5 rounds (of min 25.00us), 5.00s max time, timer: time.perf_counter ---
+    Name (time in ns)                            Min        Max       Mean   StdDev  Rounds  Iterations
+    ---------------------------------------------------------------------------------------------------
+    test_perf[slots]                        803.1335  2728.8286   826.0882  20.5562  182436          33
+    test_perf[cext]                         103.2911   254.4223   104.6869   2.6294  164373         277
+    test_perf[simple]                       448.6894  2040.6147   462.0342  17.1546  218443          49
+    test_perf[django]                       518.3013  1225.7125   532.0523  14.4672  218443          43
+    test_perf[objproxies]                  1116.8575  3827.4331  1157.8419  43.9602  174755          24
+    ---------------------------------------------------------------------------------------------------
+
+For PyPy 2.4:
+
+    Note that the mean difference between ``simple``, ``django`` and ``objproxies`` is smaller the standard deviance, thus it cannot be considered.
+
+::
+
+    --- benchmark: 4 tests, min 5 rounds (of min 25.00us), 15.00s max time, timer: time.clock ---
+    Name (time in ns)                        Min        Max     Mean   StdDev  Rounds  Iterations
+    ---------------------------------------------------------------------------------------------
+    test_perf[slots]                      1.9577   260.2979   2.2457   0.4882  498050       15384
+    test_perf[simple]                    31.3536  4374.4246  34.7433  11.7566  478894         999
+    test_perf[django]                    32.5269  4443.8381  35.8729  16.9071  461157        1000
+    test_perf[objproxies]                33.1293  4529.3718  36.7435  13.3377  452773        1000
+    ---------------------------------------------------------------------------------------------
+
+.. [1] Lacks support for certain more tricky operations like ``vars(obj)`` and other operators.