Hi,

2010/11/25 SASADA Koichi <ko1 / atdot.net>:
> (2010/11/25 13:55), Shugo Maeda wrote:
>> I have run "make benchmark" 5 times, and it shows that the modified
>> version is slower than average 2.5% than the original version.
>>
>> The benchmark result is available at:
>>
>> http://shugo.net/tmp/refinement-benchmark-20101107.csv
>
> My benchmark result is here:
> http://www.atdot.net/fp_store/f.132gcl/file.graph.png
> (Y-axis is a speed-up ratio (up is better))


I also conducted retest.
Up to 24.9% performance degradation occured on my environment.

  (become slower)
  24.9%: vm2_super
  24.5%: loop_for
  21.9%: vm1_ivar
  20.9%: vm2_zsuper
  20.7%: vm1_swap
  15.7%: vm2_poly_method
  15.0%: app_tak
  12.7%: vm3_thread_mutex
  11.6%: vm2_mutex
  11.3%: so_nsieve
  ....
  -1.6%: so_partial_sums
  -2.6%: vm2_array
  -3.0%: so_random
  -4.0%: so_nbody
  -4.3%: app_tarai
  -5.0%: so_binary_trees
  -5.1%: vm1_const
  -7.1%: vm1_ensure
  -7.5%: vm1_simplereturn
  -7.8%: so_array
  (become faster)


raw data:
http://www.atdot.net/sp/view/gu5gcl


csv:
program,original,refinement,time ratio (refinement / original)
app_answer,0.0971895694732666,0.10370779037475586,1.0670670827828104
app_erb,0.7143167495727539,0.724087905883789,1.0136790244900171
app_factorial,0.3495138168334961,0.35308108329772947,1.0102063675094504
app_fib,1.3584553718566894,1.383923053741455,1.0187475292986308
app_mandelbrot,0.31960186958312986,0.34449019432067873,1.0778729009627253
app_pentomino,31.527720880508422,33.666850090026855,1.0678491546415878
app_raise,0.8428834438323974,0.8716522693634033,1.0341314398111803
app_strconcat,0.5739905834197998,0.5915801525115967,1.0306443513184473
app_tak,1.676273727416992,1.9272963523864746,1.149750378392131
app_tarai,1.4925862312316895,1.427922534942627,0.9566767434028238
app_uri,1.4658241271972656,1.4655370235443115,0.9998041349929865
io_file_create,0.42700395584106443,0.44702463150024413,1.0468863938736708
io_file_read,0.39592618942260743,0.40063791275024413,1.0119005093714764
io_file_write,0.2036591053009033,0.20306005477905273,0.9970585625377982
loop_for,2.656430149078369,3.306622934341431,1.2447618603819273
loop_generator,0.8340113639831543,0.8517292499542236,1.0212441781206079
loop_times,2.4262061595916746,2.4653957366943358,1.0161526162761274
loop_whileloop,1.10532808303833,1.1343581676483154,1.026263771865985
loop_whileloop2,0.22827577590942383,0.23306760787963868,1.020991416855883
so_ackermann,1.4401548862457276,1.5601153373718262,1.083296909430914
so_array,2.876900815963745,2.6525750160217285,0.9220251881131089
so_binary_trees,0.7443643569946289,0.7073556423187256,0.950281452452498
so_concatenate,0.6735883235931397,0.6702359199523926,0.9950230674681766
so_count_words,0.33946728706359863,0.3388693332672119,0.99823855252281
so_exception,1.7248088836669921,1.778290367126465,1.0310071938786456
so_fannkuch,28.669377708435057,30.272033739089967,1.0559013190643263
so_fasta,3.9797119140625,3.990163469314575,1.0026262090014968
so_k_nucleotide,2.4159430503845214,2.584695339202881,1.0698494481446907
so_lists,0.5479694843292237,0.5965001106262207,1.0885644687977543
so_mandelbrot,9.189719247817994,9.212599325180054,1.00248974715604
so_matrix,0.7333712100982666,0.7263244152069092,0.9903912305332886
so_meteor_contest,9.425535678863525,9.882782745361329,1.0485115204139714
so_nbody,6.576598167419434,6.311960506439209,0.9597607069424975
so_nested_loop,2.17801456451416,2.1627822875976563,0.9930063475402418
so_nsieve,4.845090913772583,5.39297251701355,1.1130797363747225
so_nsieve_bits,5.009188795089722,5.168913412094116,1.0318863240213594
so_object,1.4976745128631592,1.5413443088531493,1.029158402319677
so_partial_sums,8.703983974456786,8.56481056213379,0.9840103781519562
so_pidigits,2.3049992084503175,2.294805479049683,0.9955775562250678
so_random,0.5043463706970215,0.48915743827819824,0.9698839264019453
so_reverse_complement,2.5664955139160157,2.577986478805542,1.0044772978667682
so_sieve,0.1414027690887451,0.15396690368652344,1.0888535258449785
so_spectralnorm,5.634414100646973,5.971523380279541,1.0598304053643943
vm1_block,4.044720840454102,4.3206627368927,1.06822272965756
vm1_const,2.2964776039123533,2.180025339126587,0.9492909207617028
vm1_ensure,1.3428500652313233,1.247154951095581,0.9287373053675523
vm1_ivar,2.4599918842315676,2.9994019985198976,1.2192731275846573
vm1_ivar_set,2.4699801445007323,2.4521598339080812,0.9927852413581838
vm1_length,2.1807806491851807,2.199760341644287,1.0087031643765723
vm1_neq,1.9046989917755126,1.9770766735076903,1.0379995380082125
vm1_not,1.5559280872344972,1.6081242561340332,1.0335466461000196
vm1_rescue,1.2895635128021241,1.297832202911377,1.0064120068745475
vm1_simplereturn,3.474906158447266,3.21527214050293,0.9252831569816115
vm1_swap,1.6546783447265625,1.9964598178863526,1.2065546299370165
vm2_array,1.432089138031006,1.3952114582061768,0.9742490332162336
vm2_case,0.46059327125549315,0.4548778057098389,0.9875910789359233
vm2_eval,28.152580356597902,29.735985946655273,1.0562437108783984
vm2_method,3.3265981674194336,3.469304418563843,1.0428985540069338
vm2_mutex,2.5152658462524413,2.8074283599853516,1.1161557193519764
vm2_poly_method,4.341252231597901,5.024772596359253,1.157447742793274
vm2_poly_method_ov,0.6244135856628418,0.6301047325134277,1.009114386652149
vm2_proc,1.251944398880005,1.3809085845947267,1.1030111128178643
vm2_regexp,2.2819092750549315,2.2643832683563234,0.9923195865452688
vm2_send,0.7383904933929444,0.8028009414672852,1.0872308739761414
vm2_super,1.1586995601654053,1.4471320629119873,1.2489277744313703
vm2_unif1,0.6489530563354492,0.6893436431884765,1.0622396126479587
vm2_zsuper,1.2690173149108888,1.5337296962738036,1.2085963510919495
vm3_gc,1.2378501892089844,1.2501022815704346,1.0098978797824312
vm3_thread_create_join,2.7109290599822997,2.7120433807373048,1.0004110475524624
vm3_thread_mutex,237.79514360427856,268.030443572998,1.1271485174610414

-- 
Yusuke Endoh <mame / tsg.ne.jp>