RFR: 8218565: HandleMark cleanup

[coleen.phillimore at oracle.com]

Claes,

Very nice!  Thank you!

- HandleArea* area = _area; // help compilers with poor alias analysis

Maybe less dubious 20 years ago. :)

Coleen

On 2/6/19 10:45 AM, Claes Redestad wrote:
> Hi,
>
> after JDK-8218145 I noticed HandleMark::pop_and_restore became outlined
> in places where it was previously inlined. Analyzing the code it seems
> there are a few calls in that method that are rarely taken, so inlining
> could be helped by outlining these. Doing so marginally helps static
> footprint and a small reduction in instructions retired on some startup
> tests on linux-x64.
>
> When in the neighborhood I noticed some code duplication between
> ~HandleMark and pop_and_restore, along with a now unused debugging
> counter (_nof_handlemarks - use was removed by JDK-8146850).
>
> Webrev: http://cr.openjdk.java.net/~redestad/8218565/open.00/
> Bug:    https://bugs.openjdk.java.net/browse/JDK-8218565
>
> Testing: tier1+2, verified pop_and_restore gets inlined in more places
> with a smaller footprint cost
>
> Thanks!
>
> /Claes