RFR: 8370947: Mitigate Neoverse-N1 erratum 1542419 negative impact on GCs and JIT performance [v25]

[Evgeny Astigeevich]

On Wed, 4 Feb 2026 14:23:08 GMT, Andrew Haley <aph at openjdk.org> wrote:

> What's the test plan for this? That is, how do we ensure the erratum-specific code path gets tested?

`test/hotspot/jtreg/gc/TestDeferredICacheInvalidation.java` is a minimal test checking no assert is triggered.
Other possible tests to write:
- If `NeoverseN1Errata1542419` is set then `UseDeferredICacheInvalidation` is set. There can be a set of tests checking those options are set correctly.
- There could be tests that involve multiple threads executing methods while GC is patching those methods.

Tests to run on Neoverse-N1:
- jcstress
- tier 1 - tier 4
- benchmarks: specjvm, specjbb, dacapo, renaissance

Does this plan look reasonable?

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

PR Comment: https://git.openjdk.org/jdk/pull/28328#issuecomment-3854622153