RFR: 8287567: AArch64: Implement post-call NOPs

[Andrew Haley]

On Wed, 1 Jun 2022 20:29:21 GMT, Dean Long <dlong at openjdk.org> wrote:

> > @dean-long , are you happy with that explanation?
> 
> Yes, and thanks for the interesting performance data.

It probably makes no difference, to be honest. I'm just a bit more confident right now to match an instruction that I am 100% certain never gets emitted by the compiler.

> I had an ulterior motive for my question. I was hoping there would be a platform-independent pattern that new ports could use initially, then optimize later. I think the
> 
> ```
>  b done
>  (raw data)
>  done:
> ```
> 
> pattern will work as long as call sites without a post-call NOP disambiguate by using a regular NOP. Then NativePostCallNop::check() will only see either a NOP or a branch.

True, although you'd have to find them all.

We could perhaps argue that  JMP *+ 2 (or however the offset works out) will never be emitted by a compiler, and I think it won't be with current C1 and C2, but it feels rather icky to me.

-------------

PR: https://git.openjdk.java.net/jdk/pull/8955