[
https://issues.apache.org/jira/browse/LUCENE-8781?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
David Smiley reopened LUCENE-8781:
----------------------------------
I set out to do some benchmarking today with toggling this setting on the
"FST50" PostingsFormat in org.apache.lucene.codecs.memory. In fst.Builder I
toggled the default setting of the shorter constructor, and then I ran the
tests with {{-Dtests.postingsformat=FST50}}. I got a couple assertion errors
from TestTermsEnum and TestSpanMultiTermQueryWrapper.
{noformat}
java.lang.AssertionError
at
__randomizedtesting.SeedInfo.seed([608D7E34838A15F1:84E00F70BA72F54D]:0)
at org.apache.lucene.util.fst.FST.readNextRealArc(FST.java:983)
at org.apache.lucene.util.fst.Util.readCeilArc(Util.java:992)
{noformat}
and
{noformat}
ava.lang.ArrayIndexOutOfBoundsException: Index -2147483405 out of bounds for
length 256
at
__randomizedtesting.SeedInfo.seed([2AA29F5F1414EE75:B06C6AD7E182571E]:0)
at
org.apache.lucene.util.fst.ReverseBytesReader.readByte(ReverseBytesReader.java:31)
at org.apache.lucene.util.fst.FST.readLabel(FST.java:507)
at org.apache.lucene.util.fst.Util.readCeilArc(Util.java:973)
{noformat}
Due to the complexities of the FST internals and that some callers make
internal assumptions, it's not clear if this problem here might also occur for
the places where we enable this direct arc encoding. So I consider this a bug.
> Explore FST direct array arc encoding
> --------------------------------------
>
> Key: LUCENE-8781
> URL: https://issues.apache.org/jira/browse/LUCENE-8781
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Mike Sokolov
> Assignee: Dawid Weiss
> Priority: Major
> Fix For: master (9.0), 8.2
>
> Attachments: FST-2-4.png, FST-6-9.png, FST-size.png
>
> Time Spent: 3h 20m
> Remaining Estimate: 0h
>
> This issue is for exploring an alternate FST encoding of Arcs as full-sized
> arrays so Arcs are addressed directly by label, avoiding binary search that
> we use today for arrays of Arcs. PR:
> https://github.com/apache/lucene-solr/pull/657
> h3. Testing
> ant test passes. I added some unit tests that were helpful in uncovering bugs
> while
> implementing which are more difficult to chase down when uncovered by the
> randomized testing we already do. They don't really test anything new;
> they're just more focused.
> I'm not sure why, but ant precommit failed for me with:
> {noformat}
> ...lucene-solr/solr/common-build.xml:536: Check for forbidden API calls
> failed while scanning class
> 'org.apache.solr.metrics.reporters.SolrGangliaReporterTest'
> (SolrGangliaReporterTest.java): java.lang.ClassNotFoundException:
> info.ganglia.gmetric4j.gmetric.GMetric (while looking up details about
> referenced class 'info.ganglia.gmetric4j.gmetric.GMetric')
> {noformat}
> I also got Test2BFST running (it was originally timing out due to excessive
> calls to ramBytesUsage(), which seems to have gotten slow), and it passed;
> that change isn't include here.
> h4. Micro-benchmark
> I timed lookups in FST via FSTEnum.seekExact in a unit test under various
> conditions.
> h5. English words
> A test of looking up existing words in a dictionary of ~170000 English words
> shows improvements; the numbers listed are % change in FST size, time to look
> up (FSTEnum.seekExact) words that are in the dict, and time to look up random
> strings that are not in the dict. The comparison is against the current
> codebase with the optimization disabled. A separate comparison of showed no
> significant change of the baseline (no opto applied) vs the current master
> FST impl with no code changes applied.
> || load=2 || load=4 || load=16 ||
> | +4, -6, -7 | +18, -11, -8 | +22, -11.5, -7 |
> The "load factor" used for those measurements controls when direct array arc
> encoding is used;
> namely when the number of outgoing arcs was > load * (max label - min label).
> h5. sequential and random terms
> The same test, with terms being a sequence of integers as strings shows a
> larger improvement, around 20% (load=4). This is presumably the best case for
> this delta, where every Arc is encoded as a direct lookup.
> When random lowercase ASCII strings are used, a smaller improvement of around
> 4% is seen.
> h4. luceneutil
> Testing w/luceneutil (wikimediumall) we see improvements mostly in the
> PKLookup case. Other results seem noisy, with perhaps a small improvment in
> some of the queries.
> {noformat}
> Task QPS base StdDev QPS opto StdDev
> Pct diff
> OrHighHigh 6.93 (3.0%) 6.89 (3.1%)
> -0.5% ( -6% - 5%)
> OrHighMed 45.15 (3.9%) 44.92 (3.5%)
> -0.5% ( -7% - 7%)
> Wildcard 8.72 (4.7%) 8.69 (4.6%)
> -0.4% ( -9% - 9%)
> AndHighLow 274.11 (2.6%) 273.58 (3.1%)
> -0.2% ( -5% - 5%)
> OrHighLow 241.41 (1.9%) 241.11 (3.5%)
> -0.1% ( -5% - 5%)
> AndHighMed 52.23 (4.1%) 52.41 (5.3%)
> 0.3% ( -8% - 10%)
> MedTerm 1026.24 (3.1%) 1030.52 (4.3%)
> 0.4% ( -6% - 8%)
> HighTerm 1111.10 (3.4%) 1116.70 (4.0%)
> 0.5% ( -6% - 8%)
> HighTermDayOfYearSort 14.59 (8.2%) 14.73 (9.3%)
> 1.0% ( -15% - 20%)
> AndHighHigh 13.45 (6.2%) 13.61 (4.4%)
> 1.2% ( -8% - 12%)
> HighTermMonthSort 63.09 (12.5%) 64.13 (10.9%)
> 1.6% ( -19% - 28%)
> LowTerm 1338.94 (3.3%) 1383.90 (5.5%)
> 3.4% ( -5% - 12%)
> PKLookup 120.45 (2.5%) 130.91 (3.5%)
> 8.7% ( 2% - 15%)
> {noformat}
> h4.FST perf tests
> I ran LookupBenchmarkTest to see the impact on the suggesters which make
> heavy use of FSTs. Some show little or no improvement, but in some cases
> there are substantial gains.
> !FST-2-4.png!
> !FST-6-9.png!
> !FST-size.png!
> chart TK
> h3. API change / implementation notes
> The only change in the public FST API is that the Builder constructor now
> takes an additional
> boolean ("useDirectArcAddressing") controlling whether or not this new
> optimization is
> applied. However, FST's internal details are not really hidden, so in
> practice any change to its
> encoding can have ripple effects in other classes.
> This is because the FST decoding is repeated in several places in the code
> base, sometimes with
> subtle variations: eg FST, FSTEnum, and fst.Util have very similar, but not
> shared, traversal code,
> and there are a few other places this same or similar decoding algorithm
> appears as well (eg
> blocktreeords codec). I do think it would be useful to make this code more
> DRY, and to hide the
> traversal implementation within a single class (or at least in the package),
> but it seemed better to
> do that separately. I also think there are possibilities for improving this
> encoding further, maybe
> by encoding Arcs in contiguous blocks with gaps in between, but before
> attempting anything like
> that, I think it would good to do that refactor to make the code in its
> current form easier to
> understand and work with.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
