We've added Jankstats, Macrobenchmark startup tests, Baseline profile, and some custom Trace metrics to the media-sample in Horologist. If you want to test out the real impact of these on performance or testing, please ping me, happy to discuss.
https://github.com/google/horologist/pull/565
StartupBenchmark_startup[compilation=BaselineProfile]
timeToFullDisplayMs min 1,468.5, median 1,522.9, max 1,585.2
timeToInitialDisplayMs min 650.4, median 678.5, max 762.3
StartupBenchmark_startup[compilation=None]
timeToFullDisplayMs min 3,618.2, median 3,845.0, max 4,448.7
timeToInitialDisplayMs min 806.8, median 901.9, max 1,069.0
