[lworld] RFR: 8227588: [lworld] Better optimize non-flattened inline type array accesses

[Tobias Hartmann]

On Thu, 25 Dec 2025 13:35:43 GMT, Quan Anh Mai <qamai at openjdk.org> wrote:

> Hi,
> 
> This PR removes all the `// TODO 8227588`. Either the problem has already been solved, or I analysed and changed the `TODO` to a more detailed issue.
> 
> Please kindly review, thanks a lot.

test/hotspot/jtreg/compiler/valhalla/inlinetypes/TestBasicFunctionality.java line 382:

> 380:     @Test
> 381:     @IR(applyIf = {"UseArrayFlattening", "true"},
> 382:         counts = {SCOPE_OBJECT, ">= 1"}) // LOAD_OF_ANY_KLASS, "<= 12"}) // TODO 8372332, 8374331 (loads should be removed)

What about setting `test12` to non-(OSR)-compilable before triggering the deopt, i.e. once `!info.isWarmUp()`? This would prevent re-compilation. You can get the method via the `info` object.

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

PR Review Comment: https://git.openjdk.org/valhalla/pull/1855#discussion_r2661346390