Status Update
Comments
la...@google.com <la...@google.com> #2
So just took a look at the trace and it doesn't seem to have any errors in the "Info And Stats" page. My suspicion is that given the unfinished slices are all happening at just the start of the trace, the problem you're seeing is
You should be able to solve this by either increasing the buffer size ftrace events are written into or by using compact sched. To do that you should change the ftrace datasource to look like this:
data_sources {
config {
name: "linux.ftrace"
... all the stuff you have right now
ftrace_config {
... all the stuff you have right now
compact_sched {
enabled: true
}
}
}
}
la...@google.com <la...@google.com> #3
Note that
ap...@google.com <ap...@google.com> #4
Branch: androidx-main
commit 17d1f80b6b6f8040363fbbb1d728d8e4fbe8a4c0
Author: Chris Craik <ccraik@google.com>
Date: Thu Oct 28 10:38:41 2021
Use more standard Perfetto buffer behavior
Bug: 199324831
Bug: 204448861
Test: TrivialStartupBenchmark # metrics still work, trace size consistent
Relnote: "Reduce chances of dropped data, and lower memory overhead from tracing"
As recommended by Perfetto team, switch to more standard buffer behavior:
- Smaller per-cpu ftrace ring buffers, which dump into single larger buffer
- Dump data to disk when central buffer full, so shouldn't drop data in long running (e.g. scrolling) benchmarks
- Add flushing to main buffer, so trace processor works in batches on very long traces
Change-Id: Id254422b83525f9d87a6bd359e2723601b510ab5
M benchmark/benchmark-common/src/main/java/androidx/benchmark/perfetto/PerfettoConfig.kt
cc...@google.com <cc...@google.com> #5
Assuming fixed with the above, will reopen if we still have trace section termination issues.
Description
See sample trace:https://ui.perfetto.dev/#!/?s=c733d216b22e916988fd8a9e3511dc4f383161fca4d6eb1d225bf386e747b
Has several errors, Lalit said it's likely a configuration error, and recommended using compact sched on Q+.