Michael Sokolov

Vector similarity

2

These changes are designed to test different vector similarity functions. In particular I want to test DOT_PRODUCT8 being added in LUCENE-10577

I wanted to test LUCENE-10577: https://github.com/apache/lucene/pull/913 with luceneutil. This 1. Adds a reference to a scaled vector document file that I uploaded to a shared space in my apache homedir...

rename master branch in git

4

What would it take to rename this branch? Does anything rely on the name, aside from people's existing checkouts? I think we could just create a new branch named say...

Currently this feature (retrieving a vector value for each search result) is disabled, and the only way to enable it is by changing this line https://github.com/mikemccand/luceneutil/blob/master/src/main/perf/TaskParser.java#L150 and passing `vectorField` to...

Address resource contention when search threads > 1

1

I noticed that when running multiple tasks with multiple search threads (ie the IndexSearcher uses an executor to search multiple segment slices concurrently) that the performance results from one task...

float performance tester, for demo only

1

This PR has a standalone test program that demonstrates some interesting performance characteristics of different data access patterns. It's not intended to be merged; I'm just putting it up here...

Precompile vector tasks

2

We have been reading the entire GloVe dictionary (400K entries, 332M) in order to convert our vector tasks from text to vectors, but we have already precomputed these in the...

Luke Webapp

4

See #11851 for an overview of what this is

Luke web interface

4

### Description I threw together a demo for ApacheCon to show off vector search and I wanted a scrappy UI I could hack on. Luke seemed like a good place...

NeighborArray is now fixed size

4

Remove the ability for `NeighborArray` to grow. Always allocates it one larger than requested, to reserve space for a temporary neighbor.