[
https://issues.apache.org/jira/browse/LUCENE-1410?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12801918#action_12801918
]
Renaud Delbru commented on LUCENE-1410:
---------------------------------------
{quote}
Did you see any effect of the JIT? This is possible by making a loop that runs
for about 1 second, and use that loop 3 times, see the output posting of 3 Oct
2008. When the 2nd and 3rd time are just about equal one can assume that the
JIT is done.
{quote}
I haven't seen the effect of the JIT on FOR.
I have re-performed the benchmark, repeating the decompression until the test
time is superior to 300 ms or 1s (as it is done in
TestFrameOfRef#doDecompPerfTestUsing() method), but I haven't observed a
difference with the previous benchmark.
In fact, it seems that it happens only for block size of 32, in the other
cases, it seems imperceptible (in the previous benchmark, the number looks
unstable, in the new benchmark however it looks more stable, but no significant
increase between the first loop and the other ones).
In the first benchmark, I was not repeating the loop until 1 second is reached
since there is no easy way to "reset" the reader. In the new benchmark, I am
recreating the reader in the loop (which causes some overhead). Do you think it
can have a consequence on the JIT ?
{quote}
For the blocks that you've been measuring one might also see an effect of L1/L2
data cache size.
{quote}
Yes, it should be an effect of the cache size. It was to check the increase of
performance and find the optimal block size. This optimal block size may be
dependent on my hardware (Intel(R) Core(TM)2 Duo CPU T7300 @ 2.00GHz, cache
size : 4096 KB).
> PFOR implementation
> -------------------
>
> Key: LUCENE-1410
> URL: https://issues.apache.org/jira/browse/LUCENE-1410
> Project: Lucene - Java
> Issue Type: New Feature
> Components: Other
> Reporter: Paul Elschot
> Priority: Minor
> Attachments: autogen.tgz, LUCENE-1410-codecs.tar.bz2,
> LUCENE-1410b.patch, LUCENE-1410c.patch, LUCENE-1410d.patch,
> LUCENE-1410e.patch, TermQueryTests.tgz, TestPFor2.java, TestPFor2.java,
> TestPFor2.java
>
> Original Estimate: 21840h
> Remaining Estimate: 21840h
>
> Implementation of Patched Frame of Reference.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: java-dev-h...@lucene.apache.org
