On Wed, 29 Apr 2026 03:52:49 GMT, Serguei Spitsyn <[email protected]> wrote:
>> Regarding initialization, does allocate_instance() return a fully >> initialized object (all nulls and 0's). I assume it does. >> >> Regarding BufferedValuePayload doing deep cloning, it is likely best if it >> does not, so I just wanted to confirm that this it does not. Also, does it >> honor GC memory barriers? > >> Also, does it honor GC memory barriers? > > There is no need in an extra GC barrier after `src.copy_to(dst);`. The > implementation of the BufferedValuePayload::copy_to()` which makes a call to > the `HeapAccess<>::value_copy(src, dst)` for a buffered layout which should > have all needed barriers in place. Ok. thanks for looking into it. ------------- PR Review Comment: https://git.openjdk.org/valhalla/pull/2363#discussion_r3158496319
