RFR: 8316180: Thread-local backoff for secondary_super_cache updates

[Aleksey Shipilev]

On Wed, 13 Sep 2023 14:02:19 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

> Work in progress, submitting for broader attention.
> 
> 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 the mitigation for AArch64 and x86_64. More platforms can be implemented in this PR, or deferred to later PRs. Port maintainers, feel free to suggest the patches for your arches, I'll be happy to fold them in.
> 
> 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.
> 
> Current PR deliberately sets backoff at `1000` to simplify testing. The actual value should be chosen by broader experiments.
> 
> Additional testing:
>  - [x] Linux x86_64 fastdebug, `tier1 tier2 tier3`
>  - [x] Linux AArch64 fastdebug, `tier1 tier2 tier3`

Motivational improvements for new benchmark shows that `contended` case improves a lot the larger backoff values we set, and `uncontended` does not regress.


$ make test TEST=micro:SecondarySuperCache TEST_VM_OPTS="-XX:+UnlockExperimentalVMOptions -XX:SecondarySuperMissBackoff=..."

== AArch64 (Mac M1, 10 cores)

# SecondarySuperMissBackoff=0 (current JDK)
SecondarySuperCache.contended    avgt   15 956,819 ± 26,373  ns/op
SecondarySuperCache.uncontended  avgt   15   2,797 ±  0,111  ns/op

# SecondarySuperMissBackoff=1
SecondarySuperCache.contended    avgt   15 401,065 ± 144,172 ns/op
SecondarySuperCache.uncontended  avgt   15   2,763 ±   0,077 ns/op

# SecondarySuperMissBackoff=10
SecondarySuperCache.contended    avgt   15  45,953 ± 24,371  ns/op
SecondarySuperCache.uncontended  avgt   15   2,787 ±  0,096  ns/op

# SecondarySuperMissBackoff=100 
SecondarySuperCache.contended    avgt   15  17,581 ± 2,910  ns/op
SecondarySuperCache.uncontended  avgt   15   2,771 ± 0,084  ns/op

# SecondarySuperMissBackoff=1000
SecondarySuperCache.contended    avgt   15   7,841 ± 0,413  ns/op
SecondarySuperCache.uncontended  avgt   15   2,739 ± 0,010  ns/op

# SecondarySuperMissBackoff=10000
SecondarySuperCache.contended    avgt   15   6,780 ± 0,082  ns/op
SecondarySuperCache.uncontended  avgt   15   2,781 ± 0,045  ns/op


=== x86_64 (Xeon, 18 cores)

Benchmark                        Mode  Cnt     Score     Error  Units

# SecondarySuperMissBackoff=0 (current JDK)
SecondarySuperCache.contended    avgt   15  2380.915 ± 159.350  ns/op
SecondarySuperCache.uncontended  avgt   15     9.165 ±   0.017  ns/op

# SecondarySuperMissBackoff=1
SecondarySuperCache.contended    avgt   15  1523.914 ±  19.694  ns/op
SecondarySuperCache.uncontended  avgt   15     9.173 ±   0.026  ns/op

# SecondarySuperMissBackoff=10
SecondarySuperCache.contended    avgt   15   736.271 ±   6.554  ns/op
SecondarySuperCache.uncontended  avgt   15     9.169 ±   0.039  ns/op

# SecondarySuperMissBackoff=100 
SecondarySuperCache.contended    avgt   15   254.527 ±   3.167  ns/op
SecondarySuperCache.uncontended  avgt   15     9.175 ±   0.053  ns/op

# SecondarySuperMissBackoff=1000
SecondarySuperCache.contended    avgt   15   91.392 ±   1.006   ns/op
SecondarySuperCache.uncontended  avgt   15    9.172 ±   0.025   ns/op

# SecondarySuperMissBackoff=10000
SecondarySuperCache.contended    avgt   15   67.798 ±   0.575   ns/op
SecondarySuperCache.uncontended  avgt   15    9.173 ±   0.047   ns/op

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

PR Comment: https://git.openjdk.org/jdk/pull/15718#issuecomment-1721574298