RFR: 8316180: Thread-local backoff for secondary_super_cache updates [v12]

[Aleksey Shipilev]

On Wed, 18 Oct 2023 08:34:04 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

>> See more details in the bug and related issues.
>> 
>> This is the attempt to mitigate [JDK-8180450](https://bugs.openjdk.org/browse/JDK-8180450), while the more complex fix that would obviate the need for `secondary_super_cache` is being worked out. The goal for this fix is to improve performance in pathological cases, while keeping non-pathological cases out of extra risk, *and* staying simple enough and reliable for backports to currently supported JDK releases.
>> 
>> This implements mitigation on most current architectures:
>>   - ✅ x86_64: implemented
>>   - 🔴 x86_32: considered, abandoned; cannot be easily done without blowing up code size
>>   - ✅ AArch64: implemented
>>   - 🔴 ARM32: considered, abandoned; needs cleanups and testing; see [JDK-8318414](https://bugs.openjdk.org/browse/JDK-8318414)
>>   - ✅ PPC64: implemented, thanks @TheRealMDoerr 
>>   - ✅ S390: implemented, thanks @offamitkumar 
>>   - ✅ RISC-V: implemented, thanks @RealFYang 
>>   - ✅ Zero: does not need implementation
>> 
>> Note that the code is supposed to be rather compact, because it is inlined in generated code. That is why, for example, we cannot easily do x86_32 version: we need a thread, so the easiest way would be to call into VM. But we cannot that easily: the code blowout would make some forward branches in external code non-short. I think we we cannot implement this mitigation on some architectures, so be it, it would be a sensible tradeoff for simplicity.
>> 
>> Setting backoff at `0` effectively disables the mitigation, and gives us safety hatch if something goes wrong.
>> 
>> I believe we can go in with `1000` as the default, given the experimental results mentioned in this PR.
>> 
>> Additional testing:
>>  - [x] Linux x86_64 fastdebug, `tier1 tier2 tier3`
>>  - [x] Linux AArch64 fastdebug, `tier1 tier2 tier3`
>
> Aleksey Shipilev has updated the pull request incrementally with four additional commits since the last revision:
> 
>  - Editorial cleanups
>  - RISC-V implementation
>  - Mention ARM32 bug
>  - Make sure benchmark runs with C1

Deeper performance evaluation shows that Dacapo:pmd has regressions, that get linearly worse as we get into larger backoffs. They reach 8% at backoff=10000. I am currently investigating the cause.

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

PR Comment: https://git.openjdk.org/jdk/pull/15718#issuecomment-1770271456