[xslt] [Bart Terryn <bart@grafikon.com>] benchmark LibXSLT against MSXML

[Matt Sergeant <matt sergeant org>]

FYI. Amazing work. Keep it up!

Forwarded message:
> Hi,
> 
> In order to fulfil Matt's wish, I wrote an MSXML driver for the XSLT
> benchmark.pl test.
> And here are the results my machine came up with:
> (a slightly reworked versions of the benchmark.pl report)
> 
> I tested:
> MSXML3 SP1
> MSXML4 technology preview
> LibXSLT V 1.07
> 
> 
> Test Component   Iter  KB In  KB Out |    MSXML3    |    MSXML4    |
> LibXSLT
> ============================================================================
> ========
>                                      |   ms KB/s R  |  ms KB/s R   |      ms
> KB/s R
> =====================================|==============|=======================
> ========
> alphabetize        10    202     157 |  781  229 CO |   922  194 CO|    651
> 275 CO
> attsets           100    204     231 |  531  409 CO |   721  301 CO|    421
> 554 CO
> avts              100   2011    1128 | 1763  890 CO |  2073  757 CO|   3134
> 507 CO
> axis              100    163       0 |  270    0 ER |   351    0 ER|    110
> 1085 CO
> backwards         100    302     150 |  871  259 CO |   951  237 CO|    872
> 330 CO
> bottles           100    205       0 | 1772    0 ER |  2033    0 ER|   3365
> 205 CO
> breadth           100    453     377 |  882  470 CO |  1071  387 CO|    901
> 362 CO
> brutal            100    424     489 |  971  470 CO |  1192  382 CO|    731
> 578 CO
> chart             100    333     490 |  991  414 CO |  1172  350 CO|    821
> 427 CO
> creation           50   1010    1321 | 2273  512 CO |  2594  449 CO|   3455
> 372 CO
> current           100     79       0 |  240    0 ER |   260    0 ER|    140
> 407 CO
> dbonerow           10  19639       0 | 8542    0 ER | 11366    0 ER|   3996
> 2458 CH
> dbtail            100   2015     504 | 1922  655 CO |  2263  556 CO|   2774
> 453 CO
> decoy             100   2189       0 | 3425    0 ER |  3985    0 ER|  10164
> 211 CO
> depth             100    626     375 | 1020  490 CO |          0 SK|
> SK
> encrypt            10    204     157 |  511  352 CO |   601  300 CO|   1432
> 142 CO
> functions         100   2085       0 | 5418    0 ER |  5908    0 ER|   3239
> 104 CO
> game              100    369      62 |  391  550 CO |   341  631 CO|    221
> 971 CO
> html              100    130       0 |  190    0 ER |   250    0 ER|    120
> 799 CH
> identity           10   1965    1562 | 3265  540 CO |  4136  426 CO|   2494
> 805 CO
> inventory         100    264       0 |  671    0 ER |   741    0 ER|    901
> 259 CO
> metric            100    210       0 |  480    0 ER |   541    0 ER|    310
> 461 CO
> number            100    157       0 |  310    0 ER |   340    0 ER|    410
> 297 NO
> oddtemplate       100     99      10 |  260  209 CO |   301  181 CO|     90
> 643 CO
> patterns          100   2070       0 | 3045    0 ER |  3605    0 ER|   6340
> 329 CO
> prettyprint        10    216     996 | 4306  140 CH |  4697  128 CH|   4937
> 167 CO
> priority          100    233      58 |  410  355 CO |   501  290 CO|    240
> 605 CO
> products          100    271      51 |  401  401 CO |   551  292 CO|    871
> 186 CO
> queens             10     71       0 | 2464    0 ER |  2644    0 ER|   3215
> 13 CO
> reverser          100    223       0 | 1933    0 ER |  2163    0 ER|   2413
> 73 CO
> stringsort         10   1966    1562 | 3635  485 CO |  4406  400 CO|   3636
> 547 CO
> summarize         100    803      30 |  340 1225 CO |   521  799 CO|    471
> 884 CO
> total             100    200      43 |  281  431 CO |   280  433 CO|    270
> 456 CO
> tower              20     71    1381 | 5267  137 CO |  5959  121 CO|  10055
> 72 CO
> trend             100    239       0 |15872    0 ER | 17055    0 ER|   5277
> 21 CO
> union             100     79      13 |  291  159 CO |   270  171 CO|    221
> 208 CO
> xpath             100     89      14 |  270  190 CO |   291  176 CO|    150
> 370 CO
> xslbench1         100    929       0 |  531    0 ER |   611    0 ER|    461
> 1752 NO
> xslbench2          20   2942       0 |  902    0 ER |  1131    0 ER|   2844
> 899 NO
> xslbench3          20   2935       0 |  330    0 ER |   370    0 ER|   1362
> 1088 NO
> ============================================================================
> ========
> total            1710  13928   10164 |27328  440 22 | 31117  371 21| 110340
> 235 36
> 
> CO = correct
> CH = check output
> ER = error
> NO = no reference
> SK = skipped
> 
> Conclusions:
> 
> 1. MSXML4.0 is slower then its MSXML3.0 parent, although MS claims improved
> performance as one of the features of MSXML3.0
> 
> 2. MSXML4.0 does not score any better on conformance either. Had to skip the
> depth test because it crashed on it (the depth test mysteriously never shows
> up under the libXSLT results although it is not configured to be skipped)
> 
> 3. Yes, Matt, LibXSLT seems to be the faster (fastest?) engine.
> 
> PS: tests were done on 500MHz PIII/128MB running Win2000
> 
> Kind Regards,
> 
> Bart
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Bart Terryn - Systems Manager - Grafikon
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Steenbruggedijk, 1   8020 Oostkamp - Belgium
> phone: +32 50 367200 fax: +32 50 367230
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Life is easy if you know all the answers(Snoopy)
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> _______________________________________________
> Perl-XML mailing list
> Perl-XML@listserv.ActiveState.com
> http://listserv.ActiveState.com/mailman/listinfo/perl-xml
>