[ARROW-11877] [C++] Add initial microbenchmarks for Dataset internals

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.0.0
• Component/s: C++
• Labels: benchmark benchmarks dataset datasets pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/27719
• Assignee: David Li
• Reporter: David Li
• Created: 05/Mar/21 18:46
• Updated: 11/Jan/23 08:22
• Resolved: 10/Mar/21 15:53

Description

A quick investigation of ARROW-11781 showed much of the overhead lies in evaluating partition expressions against the filter. While much of this is just kernel evaluation, we should have benchmarks of key Datasets internals like SimplifyWithGuarantee.

Issue Links

• relates to: ARROW-11781 [Python] Reading small amount of files from a partitioned dataset is unexpectedly slow
• links to: GitHub Pull Request #9638