Reference.reachabilityFence

[Vitaly Davidovich]

Right, conceptually it seems it shouldn't be difficult to do that given
this method is intrinsically known to the JVM.

I don't think safepoint poll would need retention either.  Given the method
isn't called and is just a pseudo marker for liveness, skipping its
would-be safepoint seems perfectly fine; safepoint poll
placement/optimization should work as-if this method didn't exist at all.

sent from my phone
On Nov 24, 2015 4:39 AM, "Andrew Haley" <aph at redhat.com> wrote:

> On 24/11/15 08:36, Paul Sandoz wrote:
> > AFAIK there are no plans to do such expansion.
>
> It would be trivial for a back end simply to fail to emit a call.
> We'd still have to emit a safepoint check.
>
> Andrew.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20151124/5503e6d2/attachment.html>