iverase commented on pull request #1866:
URL: https://github.com/apache/lucene-solr/pull/1866#issuecomment-693389673
> Do you think we could always use this new code path?
I run the Lucene benchmark for points using always the new code path, there
is a small decrease of performance but not very significative:
```
||Approach||Shape||M hits/sec ||QPS ||Hit count ||
||Dev||Base ||Diff||Dev||Base||Diff||Dev||Base||Diff||
|shapes|polyRussia|9.33|9.47|-2%|2.66|2.70|-2%|3508846|3508846| 0%|
|shapes|polyMedium|3.05|3.17|-4%|37.34|38.87|-4%|2693559|2693559| 0%|
|shapes|poly 10|40.82|41.41|-1%|25.81|26.18|-1%|355809475|355809475| 0%|
|shapes|box|41.48|41.62|-0%|42.20|42.35|-0%|221118844|221118844| 0%|
|shapes|distance|43.62|44.34|-2%|25.63|26.05|-2%|382961957|382961957| 0%|
```
If I use a `SparseFixedBitSet` instead of a `FixedBitSet` then the decrease
of QPS is quite notable:
```
|shapes|polyRussia|8.80|9.51|-7%|2.51|2.71|-7%|3508846|3508846| 0%|
|shapes|polyMedium|3.04|3.15|-4%|37.24|38.64|-4%|2693559|2693559| 0%|
|shapes|poly 10|32.62|40.88|-20%|20.63|25.85|-20%|355809475|355809475| 0%|
|shapes|box|31.05|42.37|-27%|31.59|43.12|-27%|221118844|221118844| 0%|
|shapes|distance|34.06|43.94|-22%|20.01|25.82|-22%|382961957|382961957| 0%|
```
I am not concern for testability as one code path is tested when only
indexing points and the other one should be trigger when indexing only polygons.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
