<?xml version="1.0" encoding="UTF-8"?>
<commit>
  <added type="array">
    <added>
      <filename>doc/experiments/hash/code/chained_rev1.rb</filename>
    </added>
    <added>
      <filename>doc/experiments/hash/code/chained_rev2.rb</filename>
    </added>
    <added>
      <filename>doc/experiments/hash/code/chained_rev3.rb</filename>
    </added>
  </added>
  <modified type="array">
    <modified>
      <diff></diff>
      <filename>doc/experiments/hash/hash-benchmarks.xls</filename>
    </modified>
  </modified>
  <removed type="array"/>
  <parents type="array">
    <parent>
      <id>a4d4d458cb62bbe8cc242f7fd30e7fba60a55e2e</id>
    </parent>
  </parents>
  <author>
    <name>Brian Ford</name>
    <email>bford@engineyard.com</email>
  </author>
  <url>http://github.com/evanphx/rubinius/commit/1e33851f730af4ad257071a44104822c1b4cd6dd</url>
  <id>1e33851f730af4ad257071a44104822c1b4cd6dd</id>
  <committed-date>2009-06-15T15:49:44-07:00</committed-date>
  <authored-date>2009-06-15T14:49:07-07:00</authored-date>
  <message>Improvements to current Hash algorthm.

These experiments are located at doc/experiments/hash/code/chained*.rb.
The code in chained_rev2.rb uses the cleanest code for the Iterator.
Using this code would enable using iteration internally for several
Hash methods thereby removing a layer of blocks. The improvement can
be seen in the spreadsheet on the bm_each benchmark:

  http://spreadsheets.google.com/ccc?key=rdiNYWh0xTGShVHikWVmnUA

However, the Iterator code quality has a tradeoff in performance at
the moment because of the extra read and write to an ivar. This can
be improved with improvements to CompactLookupTable.</message>
  <tree>e4cdb85c25cfba2b97f495a22d044bfd6721370f</tree>
  <committer>
    <name>Brian Ford</name>
    <email>bford@engineyard.com</email>
  </committer>
</commit>
