RFR: 8283257: x86: Clean up invocation/branch counter updates code

[Claes Redestad]

On Wed, 16 Mar 2022 12:07:55 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

> I looked briefly at optimizing `InterpreterMacroAssembler::increment_mask_and_jump` a bit, but it looks that current code is the best we can do. This improvement does a few related cleanups without semantic changes. 
> 
> Additional testing:
>  - [x] Linux x86_64 fastdebug `tier1`
>  - [x] Eyeballing interpreter generated code

(I thought I reviewed this, but my comments got lost to the void.)

Code changes look good IMO.

There are known issue that there can be heavy contention on some profiling counters, especially in synthetic, heavily multi-threaded benchmarks (SPECjvm2008), see [JDK-8134940](https://bugs.openjdk.java.net/browse/JDK-8134940). I recall @veresov did some experiments several years ago to attempt to address that, but AFAIK nothing ever came of that. This is a tricky area where I think it'd be good if also failed experiments were documented in detail.

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

Marked as reviewed by redestad (Reviewer).

PR: https://git.openjdk.java.net/jdk/pull/7838