On Fri, 21 Aug 2026 21:09:20 GMT, Aleksey Shipilev <[email protected]> wrote:

> Yeah, but thinking that uncommon traps are maybe about cold branches, the 
> branch data that lost 0->1 update due to subsampling looks to be a plausible 
> way to get extra uncommon trap? Something along those lines...

That could be. I'm going to create a test mode that runs both standard and 
randomized counter in parallel. That should end the need for speculation about 
what the cause my be.

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

PR Comment: https://git.openjdk.org/jdk/pull/28541#issuecomment-5379637278

Reply via email to