RFR: 8139457: Array bases are aligned at HeapWord granularity [v47]

[Fei Yang]

On Thu, 6 Jul 2023 13:21:30 GMT, Roman Kennke <rkennke at openjdk.org> wrote:

>> Roman Kennke has updated the pull request incrementally with three additional commits since the last revision:
>> 
>>  - Use BytesPerWord
>>  - Revert unnecessary change in s390
>>  - Revert unnecessary change in PPC
>
> I've simplified the PR significantly:
> - The gap is now usually cleared as part of the header initialization (as is already done for instance oops). This allows to use simple and fast word-sized clearing of the rest of the array.
> - Calculations of min and max array and tlab sizes are all reverted, because they are still conservatively correct. Optimizing those for a few bytes extra seems to be a rather complex task and should be done as separate PR.
> - The ARM parts could be reverted wholesale (sorry, @tstuefe) because 32 bit arch doesn't require any changes anymore.
> 
> @RealFYang can you please check the RISCV port and bring it into the same shape as aarch64/x86?
> @TheRealMDoerr Can you ack PPC and s390 ports? I've only done very minor cleanups there, compared to early version.
> @coleenp please review again? Maybe bring others in as you see fit and/or run Mach5 testing (preferably also with -XX:-UseCompressedClassPointers, because that is what this PR is about)

@rkennke : Here you go. Small update for RISC-V by referencing AArch64.

[11044-riscv-fix.txt](https://github.com/openjdk/jdk/files/12026158/11044-riscv-fix.txt)

Run non-trivial benchmark workloads, will perform more tests.

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

PR Comment: https://git.openjdk.org/jdk/pull/11044#issuecomment-1632158856