[aarch64-port-dev ] Pending approvals

[Hui Shi]

Thanks Felix and Andrew!

Are you concern tbz instruction has smaller offset range  compared with
original b.cond instruction?

For C1 changes in
http://mail.openjdk.java.net/pipermail/aarch64-port-dev/2016-March/003280.html,
branch
target label is placed closely with branch instruction (distance is 4
bytes), so it's safe to use shorter offset branch instruction here.


        __ tbnz(t, exact_log2(JVM_ACC_HAS_FINALIZER), register_finalizer);
        __ ret(lr);

        __ bind(register_finalizer);

Regards
Hui

On 29 May 2016 at 17:04, Andrew Haley <aph at redhat.com> wrote:

> Hi,
>
> On 28/05/16 12:27, Felix Yang wrote:
>
> >     Seems that the following two are pending for some time:
> >
> >
> http://mail.openjdk.java.net/pipermail/aarch64-port-dev/2016-April/003329.html
>
> That one is fine.
>
> >
> http://mail.openjdk.java.net/pipermail/aarch64-port-dev/2016-March/003280.html
>
> That one might cause problems due to branch range overflows
> in C1.  The interpreter part looks OK, though.  I have a test
> case which stresses long branches if you're interested to try
> it.
>
> Andrew.
>
>