[
https://issues.apache.org/jira/browse/HBASE-20716?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16523681#comment-16523681
]
Sahil Aggarwal commented on HBASE-20716:
----------------------------------------
[~stack]
What I got from description and some code reading is that we want to remove
these booleans:
UNSAFE_UNALIGNED
LITTLE_ENDIAN
probably because these checks happen in hot code path and can be avoided by
checking the same at startup and assigned the relevant handler. Tried running
jmh on 2 versions: one with flag and another without flag, after making it
final there seem to be almost no difference.
+non-final flag LITTLE_ENDIAN:+
Result "org.sample.UnsafeAccessBenchmark.testUnsafeWithFlag": 2178237030.784
ops/s
Result "org.sample.UnsafeAccessBenchmark.testUnsafeWithoutFlag": 3255613805.960
ops/s
+final flag LITTLE_ENDIAN:+
Result "org.sample.UnsafeAccessBenchmark.testUnsafeWithFlag": 3243770223.012
ops/s
Result "org.sample.UnsafeAccessBenchmark.testUnsafeWithoutFlag": 3256975533.152
ops/s
> Unsafe access cleanup
> ---------------------
>
> Key: HBASE-20716
> URL: https://issues.apache.org/jira/browse/HBASE-20716
> Project: HBase
> Issue Type: Improvement
> Components: Performance
> Reporter: stack
> Priority: Critical
> Labels: beginner
>
> We have two means of getting at unsafe; UnsafeAccess and then internal to the
> Bytes class. They are effectively doing the same thing. We should have one
> avenue to Unsafe only.
> Many of our paths to Unsafe via UnsafeAccess traverse flags to check if
> access is available, if it is aligned and the order in which words are
> written on the machine. Each check costs -- especially if done millions of
> times a second -- and on occasion adds bloat in hot code paths. The unsafe
> access inside Bytes checks on startup what the machine is capable off and
> then does a static assign of the appropriate class-to-use from there on out.
> UnsafeAccess does not do this running the checks everytime. Would be good to
> have the Bytes behavior pervasive.
> The benefit of one access to Unsafe only is plain. The benefits we gain
> removing checks will be harder to measure though should be plain when you
> disassemble a hot-path; in a (very) rare case, the saved byte codes could be
> the difference between inlining or not.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
