RFR: 8287567: AArch64: Implement post-call NOPs

[Dean Long]

On Tue, 31 May 2022 13:34:24 GMT, Andrew Haley <aph at openjdk.org> wrote:

> Post-call NOPs (introduced with Loom virtual threads) are not implemented in the AArch64 port. Given that these are a significant optimization on x86, we should have them on AArch64 too. 
> 
> I looked at a variety of possible instruction sequences, and ended up with `nop; movk zr, lo; movk zr, hi`. `movk` is typically one of the fastest AArch64 instructions, so it should be suitable. It's not possible to patch the `movk` instructions concurrently with this code executing, but as far as I can tell the offsets in post-call NOPs are only ever patched while the code is being installed.

Is there any advantage to having the first instruction be "nop" instead of a branch that skips over the movks? 

b done
movk zr, lo
movk zr, hi
done:

 I was going to suggest:

 b done
 (raw data)
 done:

but then it might be hard for NativePostCallNop::check() to prevent false positives.

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

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