RFR: 8344583: Make ArchiveWorkers lifecycle robust [v2]

Aleksey Shipilev shade at openjdk.org
Thu Nov 21 07:43:13 UTC 2024


On Thu, 21 Nov 2024 02:14:07 GMT, David Holmes <dholmes at openjdk.org> wrote:

> I'm struggling to see the complete picture with this code (and I still have concerns about the impact of this new thread pool on startup!).

We _are_ doing this to optimize startup, because we spend a lot of time relocating CDS archive, see the original https://github.com/openjdk/jdk/pull/21302#issuecomment-2388349450. 

I have now re-measured it to see if moving the mark exactly near the only current use still helps startup, and it does, at the same level. So we can make the lifecycle even tighter, and never start the pool if we do not need relocations, or `-XX:-AOTCacheParallelRelocation` is deliberately specified. This insulates us from issues even further. Done this in new commit.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/22276#issuecomment-2490283517


More information about the hotspot-runtime-dev mailing list