[
https://issues.apache.org/jira/browse/LUCENE-5752?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14035693#comment-14035693
]
Michael McCandless commented on LUCENE-5752:
--------------------------------------------
I ran the normal luceneutil bench on wikimediumall:
{noformat}
Task QPS base StdDev QPS comp StdDev
Pct diff
HighSloppyPhrase 3.60 (7.3%) 3.51 (7.4%)
-2.3% ( -15% - 13%)
HighPhrase 4.59 (5.5%) 4.54 (6.2%)
-1.1% ( -12% - 11%)
MedSloppyPhrase 3.59 (3.8%) 3.55 (4.8%)
-1.0% ( -9% - 7%)
HighTerm 63.28 (3.7%) 62.65 (4.3%)
-1.0% ( -8% - 7%)
MedTerm 99.13 (3.0%) 98.15 (3.6%)
-1.0% ( -7% - 5%)
MedPhrase 231.08 (5.4%) 229.10 (6.3%)
-0.9% ( -11% - 11%)
PKLookup 160.27 (2.4%) 159.32 (2.9%)
-0.6% ( -5% - 4%)
LowTerm 323.40 (1.9%) 321.68 (2.3%)
-0.5% ( -4% - 3%)
LowSloppyPhrase 45.04 (1.5%) 44.81 (2.2%)
-0.5% ( -4% - 3%)
AndHighLow 413.85 (1.6%) 412.50 (2.4%)
-0.3% ( -4% - 3%)
LowSpanNear 11.23 (3.6%) 11.20 (3.2%)
-0.2% ( -6% - 6%)
HighSpanNear 10.36 (5.3%) 10.33 (4.7%)
-0.2% ( -9% - 10%)
MedSpanNear 34.23 (3.0%) 34.16 (3.0%)
-0.2% ( -6% - 5%)
AndHighHigh 28.81 (0.6%) 28.79 (0.8%)
-0.1% ( -1% - 1%)
LowPhrase 13.51 (2.2%) 13.50 (1.6%)
-0.1% ( -3% - 3%)
AndHighMed 34.92 (0.5%) 34.90 (0.9%)
-0.1% ( -1% - 1%)
IntNRQ 3.45 (6.6%) 3.45 (6.3%)
-0.1% ( -12% - 13%)
Prefix3 93.31 (4.3%) 93.26 (3.5%)
-0.0% ( -7% - 8%)
Wildcard 20.15 (4.1%) 20.15 (2.8%)
-0.0% ( -6% - 7%)
Respell 49.36 (3.1%) 49.52 (2.8%)
0.3% ( -5% - 6%)
OrNotHighHigh 10.73 (6.3%) 10.81 (6.4%)
0.8% ( -11% - 14%)
OrHighHigh 9.89 (6.4%) 9.97 (6.3%)
0.8% ( -11% - 14%)
OrHighMed 31.86 (6.3%) 32.12 (6.4%)
0.8% ( -11% - 14%)
OrHighNotHigh 13.39 (6.1%) 13.51 (6.1%)
0.8% ( -10% - 13%)
OrHighNotMed 36.55 (5.9%) 36.88 (6.1%)
0.9% ( -10% - 13%)
OrNotHighMed 23.44 (6.4%) 23.65 (6.6%)
0.9% ( -11% - 14%)
OrHighLow 22.74 (6.6%) 22.96 (6.9%)
1.0% ( -11% - 15%)
OrNotHighLow 24.38 (6.7%) 24.62 (6.8%)
1.0% ( -11% - 15%)
OrHighNotLow 29.97 (6.5%) 30.29 (6.9%)
1.0% ( -11% - 15%)
Fuzzy2 45.50 (3.2%) 46.05 (3.1%)
1.2% ( -5% - 7%)
Fuzzy1 60.83 (4.0%) 61.71 (3.9%)
1.4% ( -6% - 9%)
{noformat}
Net/net I think this is noise ... but Rob pointed out the Fuzzy1/2
tasks here don't do a prefix, so I'll fixup luceneutil to support that
and test it too.
> Explore light weight Automaton replacement
> ------------------------------------------
>
> Key: LUCENE-5752
> URL: https://issues.apache.org/jira/browse/LUCENE-5752
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 5.0
>
> Attachments: LUCENE-5752.patch
>
>
> This effort started with the patch on LUCENE-4556, to create a "light
> weight" replacement for the current object-heavy Automaton class
> (which creates separate State and Transition objects).
> I took that initial patch much further, and cutover most places in
> Lucene that use Automaton to LightAutomaton. Tests pass.
> The core idea of LightAutomaton is all states are ints, and you build
> up the automaton under the restriction that you add all outgoing
> transitions one state at a time. This worked well for most
> operations, but for some (e.g. UTF32ToUTF8!!) it was harder, so I also
> added a separate builder to add transitions in any order and then in
> the end they are sorted and added to the real automaton.
> If this is successful I think we should just replace the current
> Automaton with LightAutomaton; right now they both exist in my current
> patch...
> This is very much a work in progress, and I'm not sure the
> restrictions the API imposes are "reasonable" (some algos got uglier).
> But I think it's at least worth exploring/iterating... I'll make a branch and
> commit my current state.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
