[GitHub] [lucene-solr] iverase commented on issue #747: LUCENE-8888: Improve distribution of points with data dimension in BKD tree leaves

iverase commented on issue #747: LUCENE-: Improve distribution of points with data dimension in BKD tree leaves URL: https://github.com/apache/lucene-solr/pull/747#issuecomment-508390720 I have run the benchmarks and all and all it seems there is little effect in indexing throughput:

[GitHub] [lucene-solr] iverase commented on issue #747: LUCENE-8888: Improve distribution of points with data dimension in BKD tree leaves

iverase commented on issue #747: LUCENE-: Improve distribution of points with data dimension in BKD tree leaves URL: https://github.com/apache/lucene-solr/pull/747#issuecomment-507694176 Ok, so finally I modify the HeapPointWriter/Reader to hold all the bytes in one array.

[GitHub] [lucene-solr] iverase commented on issue #747: LUCENE-8888: Improve distribution of points with data dimension in BKD tree leaves

iverase commented on issue #747: LUCENE-: Improve distribution of points with data dimension in BKD tree leaves URL: https://github.com/apache/lucene-solr/pull/747#issuecomment-507604565 > Do you think we could go even further and use a similar idea in HeapPointWriter instead of

[GitHub] [lucene-solr] iverase commented on issue #747: LUCENE-8888: Improve distribution of points with data dimension in BKD tree leaves

iverase commented on issue #747: LUCENE-: Improve distribution of points with data dimension in BKD tree leaves URL: https://github.com/apache/lucene-solr/pull/747#issuecomment-506317120 Thanks @jpountz, I can indeed refactor PointValue and it seems it makes a lot of sense. It