Commits on Oct 3, 2020

1. HtmlParser: Improve performance on JVM …

   The HTML parser incurs a significant slowdown as the nesting level
   increases:
   
   ```shell
   $ bloop run pine-bench-jvm -- slow
   [...]
   Benchmark: Parse HTML w/o attributes
   - depth=2:
     units: 7
     iterations: 591733
     run time: 3384 μs/it ± 7
   - depth=6:
     units: 127
     iterations: 11610
     run time: 171586 μs/it ± 504
   - depth=10:
     units: 2047
     iterations: 56
     run time: 36148809 μs/it ± 74820
   - depth=14:
     units: 32767
     iterations: 1
     run time: 9353666666 μs/it ± 174704194
   
   Summary:
     Unit growth: 18.1x, 16.1x, 16.0x
     Run time growth: 50.7x, 210.7x, 258.8x
   ```
   
   This slow down can be attributed to the `rest()` function in `Reader`.
   It calls `data.drop()` which on the JVM creates a copy of the string
   rather than pointing to the same memory.
   
   Scala.js' `drop()` implementation has the expected semantics such that
   the run time performance is roughly linear to the number of nodes in the
   tree:
   
   ```shell
   $ bloop run pine-bench-js -- slow
   [...]
   Benchmark: Parse HTML w/o attributes
   - depth=2:
     units: 7
     iterations: 92592
     run time: 21624 μs/it ± 128
   - depth=6:
     units: 127
     iterations: 5229
     run time: 382531 μs/it ± 182
   - depth=10:
     units: 2047
     iterations: 312
     run time: 6479611 μs/it ± 134542
   - depth=14:
     units: 32767
     iterations: 17
     run time: 119013071 μs/it ± 1455066
   
   Summary:
     Unit growth: 18.1x, 16.1x, 16.0x
     Run time growth: 17.7x, 16.9x, 18.4x
   ```
   
   After applying the optimisations, the parser will behave similarly on
   the JVM:
   
   ```
   $ bloop run pine-bench-jvm -- slow
   [...]
   Benchmark: Parse HTML w/o attributes
   - depth=2:
     units: 7
     iterations: 991955
     run time: 2048 μs/it ± 51
   - depth=6:
     units: 127
     iterations: 45471
     run time: 43403 μs/it ± 412
   - depth=10:
     units: 2047
     iterations: 2523
     run time: 777550 μs/it ± 10750
   - depth=14:
     units: 32767
     iterations: 147
     run time: 13759510 μs/it ± 160258
   
   Summary:
     Unit growth: 18.1x, 16.1x, 16.0x
     Run time growth: 21.2x, 17.9x, 17.7x
   ```

   

   tindzk committed Oct 3, 2020
   Configuration menu

   • View commit details
   • Copy full SHA for 8e70cc9
   • Browse repository at this point

   Copy the full SHA

   8e70cc9 View commit details

   Browse the repository at this point in the history