[
https://issues.apache.org/jira/browse/LUCENE-3892?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13411517#comment-13411517
]
Han Jiang commented on LUCENE-3892:
-----------------------------------
bq. The Pulsing parts in last patch is not included here, because they doesn't
improve performance significantly.
Here are some tests between For vs PulsingFor, PFor vs PulsingPFor. Run on the
1M docs with wikimediumhard.tasks
It is strange that PKLookup still doesn't benefit for FixedBlockInt:
{noformat}
Task QPS For StdDev ForQPS PulsingForStdDev PulsingFor
Pct diff
AndHighHigh 23.01 0.33 22.94 0.66 -4% -
4%
AndHighMed 56.41 0.76 57.41 1.74 -2% -
6%
Fuzzy1 86.74 0.85 82.22 2.39 -8% -
-1%
Fuzzy2 28.23 0.38 26.15 0.97 -11% -
-2%
IntNRQ 41.78 1.65 40.78 3.53 -14% -
10%
OrHighHigh 14.44 0.34 14.50 0.92 -8% -
9%
OrHighMed 30.59 0.77 31.12 1.93 -6% -
10%
PKLookup 110.31 2.03 109.22 2.43 -4% -
3%
Phrase 8.18 0.44 7.97 0.40 -12% -
8%
Prefix3 99.64 2.38 97.09 3.46 -8% -
3%
Respell 99.66 0.45 92.76 2.81 -10% -
-3%
SloppyPhrase 4.28 0.16 4.08 0.13 -11% -
2%
SpanNear 4.08 0.13 3.93 0.06 -7% -
0%
Term 33.63 1.25 34.06 1.71 -7% -
10%
TermBGroup1M 15.54 0.46 15.78 0.56 -4% -
8%
TermBGroup1M1P 20.34 0.73 20.62 0.62 -5% -
8%
TermGroup1M 19.18 0.52 19.72 0.49 -2% -
8%
Wildcard 34.86 0.88 34.27 1.77 -9% -
6%
{noformat}
{noformat}
AndHighHigh 19.98 0.31 19.92 0.26 -3% -
2%
AndHighMed 58.21 1.51 57.86 1.18 -5% -
4%
Fuzzy1 91.86 1.17 85.86 1.18 -8% -
-4%
Fuzzy2 32.66 0.58 30.08 0.57 -11% -
-4%
IntNRQ 33.89 0.82 32.66 1.10 -9% -
2%
OrHighHigh 15.79 1.29 14.96 0.67 -16% -
7%
OrHighMed 30.31 2.09 28.91 1.67 -15% -
8%
PKLookup 112.80 0.81 111.82 2.90 -4% -
2%
Phrase 6.14 0.11 6.23 0.10 -1% -
5%
Prefix3 147.80 2.88 138.35 2.11 -9% -
-3%
Respell 118.57 1.18 108.30 1.86 -11% -
-6%
SloppyPhrase 5.78 0.15 5.66 0.29 -9% -
5%
SpanNear 6.32 0.14 6.40 0.16 -3% -
6%
Term 41.60 2.44 38.12 0.33 -14% -
-1%
TermBGroup1M 14.40 0.48 13.73 0.19 -8% -
0%
TermBGroup1M1P 23.68 0.44 22.82 0.44 -7% -
0%
TermGroup1M 15.25 0.48 14.51 0.20 -9% -
0%
Wildcard 32.76 0.53 31.76 0.62 -6% -
0%
{noformat}
> Add a useful intblock postings format (eg, FOR, PFOR, PFORDelta,
> Simple9/16/64, etc.)
> -------------------------------------------------------------------------------------
>
> Key: LUCENE-3892
> URL: https://issues.apache.org/jira/browse/LUCENE-3892
> Project: Lucene - Java
> Issue Type: Improvement
> Reporter: Michael McCandless
> Labels: gsoc2012, lucene-gsoc-12
> Fix For: 4.1
>
> Attachments: LUCENE-3892-BlockTermScorer.patch,
> LUCENE-3892-direct-IntBuffer.patch, LUCENE-3892-for&pfor-with-javadoc.patch,
> LUCENE-3892-for&pfor.patch, LUCENE-3892-handle_open_files.patch,
> LUCENE-3892_for.patch, LUCENE-3892_for_byte[].patch,
> LUCENE-3892_for_int[].patch, LUCENE-3892_for_unfold_method.patch,
> LUCENE-3892_pfor.patch, LUCENE-3892_pfor.patch, LUCENE-3892_pfor.patch,
> LUCENE-3892_pfor_unfold_method.patch, LUCENE-3892_pulsing_support.patch,
> LUCENE-3892_settings.patch, LUCENE-3892_settings.patch
>
>
> On the flex branch we explored a number of possible intblock
> encodings, but for whatever reason never brought them to completion.
> There are still a number of issues opened with patches in different
> states.
> Initial results (based on prototype) were excellent (see
> http://blog.mikemccandless.com/2010/08/lucene-performance-with-pfordelta-codec.html
> ).
> I think this would make a good GSoC project.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
