RFR: 8322630: Remove ICStubs and related safepoints [v3]

[Erik Österlund]

On Fri, 26 Jan 2024 15:59:35 GMT, Axel Boldt-Christmas <aboldtch at openjdk.org> wrote:

>> Erik Österlund has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Whitespace fix
>>   
>>   Co-authored-by: Thomas Schatzl <59967451+tschatzl at users.noreply.github.com>
>
> Just did an initial read through of the PR. Just added some cleanup suggestion. Also noticed something I though looked wrong in the ARM32 port. 
> 
> I also went through and tried to find the handful of places in the codebase where the term `ICHolder` (or its derivatives) were still used.  Put them in a separate branch to not clutter this PR. Would be nice to take this all the way and not have stale comments or naming lurking about.  (Also nuked the `DECC` copy-paste-typo)
> Comment cleanups:
> f1bb02ea472eb314c93d80b830c59bd03e280116
> 
> All platforms use `data` as a register alias for the `CompileICData*` register in the `ic_check`. But c2i and itable stubs still use `holder`.  Maybe go all the way here?
> 5422ed32def491bd1e145959b7f3c49c88cfc50e
> 
> Also for PPC and s390 I think the code is easier to understand if the global inline cache register aliases these platforms have are used. But maybe that is just me.
> 39c0a7ede5187cba52d6fcf48c0852213c48c899
> 
> As for the implementation I could not see anything wrong (except the ARM32 port). But I'll leave it people with more expertise in this area.

Thanks for the reviews! I applied the cleanups from @xmas92 and @RealFYang. I added the JVMCI hook for @dougxc and started bulk allocating and freeing CompiledICData to deal with the situation reported by @tschatzl. I haven't touched the ARM32 code though - waiting for @voitylov there.

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

PR Comment: https://git.openjdk.org/jdk/pull/17495#issuecomment-1914687744