#7143 is going to add benchmarks to arrow-reader-writer-benchmarks, but these only cover dictionary encoded cases. It is hard to force non-dictionary encoding so we should try to write some benchmarks that give a better sense of how that path changes with any algorithm changes.
It isn't clear how much this code path is exercised in practice so I'm not sure on the priority.
Reporter: Micah Kornfield / @emkornfield
Note: This issue was originally created as ARROW-8760. Please see the migration documentation for further details.