On Mon, 18 May 2026 09:42:32 GMT, Aleksey Shipilev <[email protected]> wrote:
>>> Agreed. Looks like doing save-in-vector is more hassle than it is worth. I >>> would be happy if we just not yank the register for tier1 code. tier2/3 >>> code can suffer with one fewer GP register. >> >> Mm, that's an interesting idea. Could I do it in a follow-up? > > Yes. As a middle ground, I propose ergonomically setting > `ProfileCaptureRatio=1` when user supplies `TieredStopAtLevel=1` right in > this PR. AFAICS, this already disables register reservation on C1 side, at > the cost of disabling scalable counters on interpreter side. This would cover > 99.9% of C1 throughput cases we care about. Then you can mop up the > interpreter scalability in that C1-only case and C1 throughput in > interpreter+C1+C2 edge cases, as the follow-up. This is an experimental feature patch, so it will be disabled by default. (Or, to put it another way, default `ProfileCaptureRatio=1`. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/28541#discussion_r3350677908
