Open
Description
Summary
Today it's easy to test build/evaluation timing changes across the configured PerfStar scenarios. However, changes like #12033 may also introduce regressions or improvements in the memory usage of the build. It would be great to be able to collect that information across PerfStar runs and compare it similar to the duration information.
Background and Motivation
Make sure we don't regress memory usage with features/fixes.
Proposed Feature
- Collect memory usage across the nodes involved in the build and report it individually and in aggregate for each test case in PerfStar.
- Be able to pivot/slice to a memory-use-centric view of selected runs, getting absolute and percent-changed information about the memory usage of the build.
Alternative Designs
No response