[
https://issues.apache.org/jira/browse/LUCENE-10297?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Feng Guo updated LUCENE-10297:
------------------------------
Description:
We already have a bitset optimization for low cardinality fields, but the
optimization only works on extremly low cardinality fields (doc count > 1/16
total doc), for medium cardinality case like 32/128 can rarely get this
optimization.
In [https://github.com/apache/lucene-solr/pull/1538], we made some effort to
use readLELongs to speed up BKD id blocks, but did not get a obvious gain on
this approach. Maybe this is because we are trying to optimize the unsorted
situation (typically happens for high cardinality fields) and the bottleneck of
queries on high cardinality fields is {{visitDocValues}} but not {{readDocIds}}
?
IMO medium cardinality fields may be tempted for this optimization because they
need to read lots of ids for one term. The basic idea is that we can compute
the delta of the sorted ids and encode/decode them like what we do in
{{StoredFieldsInts}}. I benchmarked the optimization by mocking some random
longPoint and querying them with {{PointInSetQuery}}. As expected, the medium
cardinality fields got spped up and high cardinality fields get even results.
*Benchmark Result*
|doc count|field cardinality|field term count|baseline(ms)|candidate(ms)|diff
percentage|
|100000000|32|1|19|16|-15.79%|
|100000000|32|2|34|14|-58.82%|
|100000000|32|4|76|22|-71.05%|
|100000000|32|8|139|42|-69.78%|
|100000000|32|16|279|82|-70.61%|
|100000000|128|1|17|11|-35.29%|
|100000000|128|8|75|23|-69.33%|
|100000000|128|16|126|25|-80.16%|
|100000000|128|32|245|50|-79.59%|
|100000000|128|64|528|97|-81.63%|
|100000000|1024|1|3|2|-33.33%|
|100000000|1024|8|13|8|-38.46%|
|100000000|1024|32|31|19|-38.71%|
|100000000|1024|128|120|67|-44.17%|
|100000000|1024|512|480|133|-72.29%|
|100000000|8192|1|3|3|0.00%|
|100000000|8192|16|18|15|-16.67%|
|100000000|8192|64|19|14|-26.32%|
|100000000|8192|512|69|43|-37.68%|
|100000000|8192|2048|236|134|-43.22%|
|100000000|1048576|1|3|2|-33.33%|
|100000000|1048576|16|18|19|5.56%|
|100000000|1048576|64|17|17|0.00%|
|100000000|1048576|512|34|32|-5.88%|
|100000000|1048576|2048|89|93|4.49%|
was:
We already have a bitset optimization for low cardinality fields, but the
optimization only works on extremly low cardinality fields (doc count > 1/16
total doc), for medium cardinality case like 32/128 can rarely get this
optimization.
In [https://github.com/apache/lucene-solr/pull/1538], we made some effort to
use readLELongs to speed up BKD id blocks, but did not get a obvious gain on
this approach. Maybe this is because we are trying to optimize the unsorted
situation (typically happens for high cardinality fields) and the bottleneck of
queries on high cardinality fields is {{visitDocValues}} but not {{readDocIds}}
?
IMO medium cardinality fields may be tempted for this optimization because they
need to read lots of ids. The basic idea is that we can compute the delta of
the sorted ids and encode/decode them like what we do in {{StoredFieldsInts}}.
I benchmarked the optimization by mocking some random longPoint and querying
them with {{PointInSetQuery}}. As expected, the medium cardinality fields got
spped up and high cardinality fields get even results.
*Benchmark Result*
|doc count|field cardinality|field term count|baseline(ms)|candidate(ms)|diff
percentage|
|100000000|32|1|19|16|-15.79%|
|100000000|32|2|34|14|-58.82%|
|100000000|32|4|76|22|-71.05%|
|100000000|32|8|139|42|-69.78%|
|100000000|32|16|279|82|-70.61%|
|100000000|128|1|17|11|-35.29%|
|100000000|128|8|75|23|-69.33%|
|100000000|128|16|126|25|-80.16%|
|100000000|128|32|245|50|-79.59%|
|100000000|128|64|528|97|-81.63%|
|100000000|1024|1|3|2|-33.33%|
|100000000|1024|8|13|8|-38.46%|
|100000000|1024|32|31|19|-38.71%|
|100000000|1024|128|120|67|-44.17%|
|100000000|1024|512|480|133|-72.29%|
|100000000|8192|1|3|3|0.00%|
|100000000|8192|16|18|15|-16.67%|
|100000000|8192|64|19|14|-26.32%|
|100000000|8192|512|69|43|-37.68%|
|100000000|8192|2048|236|134|-43.22%|
|100000000|1048576|1|3|2|-33.33%|
|100000000|1048576|16|18|19|5.56%|
|100000000|1048576|64|17|17|0.00%|
|100000000|1048576|512|34|32|-5.88%|
|100000000|1048576|2048|89|93|4.49%|
> Speed up medium cardinality fields with readLELongs and SIMD
> ------------------------------------------------------------
>
> Key: LUCENE-10297
> URL: https://issues.apache.org/jira/browse/LUCENE-10297
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/codecs
> Reporter: Feng Guo
> Priority: Major
> Time Spent: 10m
> Remaining Estimate: 0h
>
> We already have a bitset optimization for low cardinality fields, but the
> optimization only works on extremly low cardinality fields (doc count > 1/16
> total doc), for medium cardinality case like 32/128 can rarely get this
> optimization.
> In [https://github.com/apache/lucene-solr/pull/1538], we made some effort to
> use readLELongs to speed up BKD id blocks, but did not get a obvious gain on
> this approach. Maybe this is because we are trying to optimize the unsorted
> situation (typically happens for high cardinality fields) and the bottleneck
> of queries on high cardinality fields is {{visitDocValues}} but not
> {{readDocIds}} ?
> IMO medium cardinality fields may be tempted for this optimization because
> they need to read lots of ids for one term. The basic idea is that we can
> compute the delta of the sorted ids and encode/decode them like what we do in
> {{StoredFieldsInts}}. I benchmarked the optimization by mocking some random
> longPoint and querying them with {{PointInSetQuery}}. As expected, the medium
> cardinality fields got spped up and high cardinality fields get even results.
> *Benchmark Result*
> |doc count|field cardinality|field term count|baseline(ms)|candidate(ms)|diff
> percentage|
> |100000000|32|1|19|16|-15.79%|
> |100000000|32|2|34|14|-58.82%|
> |100000000|32|4|76|22|-71.05%|
> |100000000|32|8|139|42|-69.78%|
> |100000000|32|16|279|82|-70.61%|
> |100000000|128|1|17|11|-35.29%|
> |100000000|128|8|75|23|-69.33%|
> |100000000|128|16|126|25|-80.16%|
> |100000000|128|32|245|50|-79.59%|
> |100000000|128|64|528|97|-81.63%|
> |100000000|1024|1|3|2|-33.33%|
> |100000000|1024|8|13|8|-38.46%|
> |100000000|1024|32|31|19|-38.71%|
> |100000000|1024|128|120|67|-44.17%|
> |100000000|1024|512|480|133|-72.29%|
> |100000000|8192|1|3|3|0.00%|
> |100000000|8192|16|18|15|-16.67%|
> |100000000|8192|64|19|14|-26.32%|
> |100000000|8192|512|69|43|-37.68%|
> |100000000|8192|2048|236|134|-43.22%|
> |100000000|1048576|1|3|2|-33.33%|
> |100000000|1048576|16|18|19|5.56%|
> |100000000|1048576|64|17|17|0.00%|
> |100000000|1048576|512|34|32|-5.88%|
> |100000000|1048576|2048|89|93|4.49%|
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
