Freeze/thaw at safepoint stub?

Ron Pressler ron.pressler at oracle.com
Wed Jun 8 15:01:08 UTC 2022


Hi.

The code to support freeze at safepoints reaches deep, including into GC code, and serves as an anchor for the design. Because we didn’t deliver it in JEP 425 we removed the entry points but have kept some internal code in place to keep ongoing work mergeable, and ensure the code doesn’t start drifting away. So please don’t remove it.

— Ron

> On 8 Jun 2022, at 14:30, Aleksey Shipilev <shade at redhat.com> wrote:
> 
> Hi,
> 
> I see the code that allows freeze/thaw from safepoint stub, in addition for "normal" yield stub. But I also suspect that code is effectively dead? That is:
> 
> - FreezeBase::_preempt is always false;
> - FreezeBase::freeze_start_frame_safepoint_stub() is effectively not called;
> - FreezeBase::recurse_freeze_stub_frame() is effectively not called;
> - is_pinned0 is always called with safepoint = false;
> - ...and I suspect the thawing paths that deal with stubs are also not called?
> 
> Is this a leftover from some earlier implementation? If so, can I remove those parts to avoid unnecessary clutter?
> 
> -- 
> Thanks,
> -Aleksey
> 



More information about the loom-dev mailing list