[Jython-dev] some microbenchmarks results

[Samuele P. <ped...@op...>]

Hi,

now that Jython supports properly new style classes etc it can run
the suite of microbenchmarks that have been and are used by the PyPy 
project:

http://codespeak.net/svn/pypy/dist/pypy/translator/microbench/

I suppose people here may be interested even if performance is not the 
only or main goal right now. These are the results of a run comparing 
python2.5 and jython2.2.1 on a 1st gen macbook pro (intel core duo, not 
core duo 2)
still using Java 5 for Jython:

You run microbench.py with a list of python interpreters, the first is 
used as reference to produce relative results, for jython you need a 
larger heap than the default one (I know that -Xmx512m works):

exe: /Users/pedronis/sw/Python/2.5/Python.framework/Versions/2.5/bin/python
exe: /Users/pedronis/Software/jython2.2.1/jython
 0.86x slower on test_exception.test_try_except_finally()
 1.01x slower on test_count1.test_loop_other_count()
 1.02x slower on test_exception.test_try_except_else()
 1.03x slower on test_exception.test_try_except()
 1.03x slower on test_unicode.test_bad_case_python2_5()
 1.09x slower on test_exception.test_except_multiple_builtin_exception()
 1.10x slower on test_exception.test_except_multiple_user_exception()
 1.14x slower on test_count1.test_call_nested_function_other_count()
 1.14x slower on test_exception.test_reraise()
 1.17x slower on test_list.test_list_setitem()
 1.21x slower on test_count1.test_count_in_dict()
 1.29x slower on test_count1.test_call_nested_function()
 1.30x slower on test_count1.test_loop_unrolled()
 1.38x slower on test_count1.test_call_nested_function_many_args()
 1.40x slower on test_count1.test_loop()
 1.43x slower on test_count1.test_call_function_with_arguments_in_cellvars()
 1.53x slower on test_exception.test_except_specific_builtin_exception()
 1.53x slower on test_exception.test_except_specific_user_exception()
 1.56x slower on 
test_count1.test_call_function_without_arguments_in_cellvars()
 1.89x slower on test_dict.test_dict_instance_setattr_instance_dict()
 1.89x slower on test_dict.test_dict_instance_setnewattr_instance_dict()
 1.94x slower on test_count1.test_call_function()
 1.96x slower on test_bltn.test_isinstance1()
 2.00x slower on test_list.test_list_extend()
 2.02x slower on test_count1.test_call_method_of_old_style_class()
 2.02x slower on test_formatting.test_repr_formatting()
 2.03x slower on test_list.test_list_getitem()
 2.09x slower on test_count1.test_count_in_global2()
 2.09x slower on test_list.test_list_append()
 2.09x slower on test_count1.test_count_in_global()
 2.10x slower on test_dict.test_dict_creation_mode2()
 2.11x slower on test_create1.test_simple_loop_with_new_style_class_new()
 2.12x slower on test_bltn.test_isinstance2()
 2.13x slower on test_bltn.test_call_sin()
 2.15x slower on test_exception.test_raise_user_exception()
 2.16x slower on test_exception.test_raise_builtin_exception()
 2.17x slower on test_dict.test_dict_getitem()
 2.25x slower on test_formatting.test_dict_formatting()
 2.26x slower on 
test_create1.test_simple_loop_with_old_style_class_creation()
 2.28x slower on test_create1.test_simple_loop()
 2.34x slower on test_dict.test_dict_creation_mode1()
 2.34x slower on test_bltn.test_call_fabs()
 2.52x slower on test_dict.test_dict_instance_getattr_instance_dict()
 2.59x slower on test_bltn.test_isinstance3()
 2.85x slower on test_unicode.test_find_pattern1()
 3.06x slower on test_formatting.test_format_long()
 3.15x slower on test_unicode.test_find_pattern2()
 3.19x slower on test_formatting.test_format_unicode()
 3.34x slower on test_dict.test_dict_class_dict_getmethod()
 3.38x slower on test_formatting.test_simple_formatting()
 3.46x slower on test_dict.test_dict_creation_mode3()
 3.53x slower on test_dict.test_dict_raw_range()
 3.84x slower on test_dict.test_dict_setitem1()
 4.10x slower on test_dict.test_dict_setitem2()
 4.71x slower on 
test_create1.test_simple_loop_with_new_style_class_creation()
 4.78x slower on test_unicode.test_find_pattern4()
 4.80x slower on test_count1.test_call_method_of_new_style_class()
 5.14x slower on test_formatting.test_number_formatting()
 5.26x slower on test_list.test_list_slice()
 6.12x slower on test_exception.test_instantiate_user_exception()
 6.26x slower on test_count1.test_count_in_slot()
 7.06x slower on test_count1.test_count_in_attr()
 7.30x slower on test_unicode.test_find_pattern8()
 9.07x slower on test_unicode.test_count_worstcase()
 9.11x slower on test_unicode.test_find_worstcase()
 9.40x slower on test_exception.test_instantiate_builtin_exception()
10.29x slower on test_unicode.test_find_pattern16()


we are not too shabby, someone may want to look at the worst offenders 
though.

regards.