On Tue, 11 Oct 2022 11:31:51 GMT, Alan Bateman <al...@openjdk.org> wrote:
> I wouldn't expect it to change the execution time of tier1_part1 I have *seen* the change in total CPU time when I tried to do the same thing this PR does. This is why I bailed on doing the same thing PR proposes. Cutting out the tests into parallel `@test` blocks is not a straightforward thing to do: you need to have significantly large tests to get a parallelism benefit that covers the additional overhead of doing more JVMs. This is why my PRs that do this kind of transformation show `time`-s before/after the change, where one can see that `wall` time gets reduced (parallelism benefit), while `usr+sys` times stay more or less the same (overhead penalties). I insist doing this for any kind of test parallelism changes. ------------- PR: https://git.openjdk.org/jdk/pull/10641
