RFR: 8373697: AArch64: Remove deoptimization stub code for nmethods with small stack frames
Ruben
duke at openjdk.org
Thu Feb 26 10:55:34 UTC 2026
On Thu, 26 Feb 2026 10:36:23 GMT, Andrew Haley <aph at openjdk.org> wrote:
> Sure, but you can pass a pointer to anything, and that includes a pointer to an array containing all the original PCs.
I understand, and this way we don't have to allocate a permanent field in the TLS for storing this pointer to the array. I will spend some time thinking through the implementation using array of original PCs.
Would it be acceptable to dynamically allocate memory upon deoptimization event? I am unsure because that can result in out-of-memory, and so make a particular deoptimization event impossible to process - while it is an unlikely corner case, the constraint does not currently exist as far as I know.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/28857#issuecomment-3965811722
More information about the hotspot-dev
mailing list