David Smiley

My suspicion is that batch & disable tiered compilation is trying to reduce the number of ancillary activities the JVM is doing that might interfere with the measurements (introduce sporadic...

If I recall (at Berlin Buzzwords "spatial chat") did you say the runs were using separate JVMs? (I was surprised to hear that). If it's the same JVM and then...

I didn't know Lucene offered sorting on a BinaryDocValuesField; I thought it was only value retrieval? It's not obvious to me from the diff how the sorting is happening; can...

> I'm a little confused on what the bug was here. > > E.g. were you seeing all 0 for the GC statistics at the end of all benchmark runs?...

Might you look at the PR please?

IDs are taken at random: https://github.com/mikemccand/luceneutil/blob/master/src/main/perf/PKLookupTask.java#L60 I agree that a realistic tests should have _some_ IDs not match; some before & after the range of actual IDs. Thanks for pointing...

Am I incorrect that it chooses IDs to query at random and thus is not likely to choose any IDs that are actually indexed? Up above, I referenced a line...

Doh! Thanks for clarifying what's going on; I was very confused by this. That line of code looked different to me. I temporarily added a bit more logging in printResults...

It's not clear to me if we need to concern ourselves with the use-case of Solr's `/export`. I believe the goal is to speed up DocValues access assuming certain access...

As mentioned in the associated JIRA issue https://issues.apache.org/jira/browse/SOLR-14619 , this PR is blocked up a bit on the Filter score matter.