[lworld] RFR: 8359345: [lworld] C2 crashes with -XX:ForceNonTearable=* [v2]

[Tobias Hartmann]

On Tue, 1 Jul 2025 10:21:02 GMT, Tobias Hartmann <thartmann at openjdk.org> wrote:

>> For test method `testFieldLoad1` the graph looks like this after parsing:
>> ![Screenshot from 2025-06-30 13-42-44](https://github.com/user-attachments/assets/94726f40-8260-47ab-86e1-a47b074a6929)
>> 
>> Now since [JDK-8354068](https://bugs.openjdk.org/browse/JDK-8354068) `MemNode::optimize_simple_memory_chain` rewires the memory input of strict field load `46 LoadL` to before the membar:
>> ![Screenshot from 2025-06-30 15-06-13](https://github.com/user-attachments/assets/b8ac826e-1426-4212-8567-0aa1c8da5ad9)
>> 
>> Later in `PhaseCFG::insert_anti_dependences`, a precedence edge from the membar to the load (now named `15 LoadL`) is added, because the type of the `CastI2N` is `TypeNarrowOop::BOTTOM` and thus `AliasType::is_rewritable` is true (it should be false). The resulting graph is unschedulable:
>> ![Screenshot from 2025-06-30 13-37-59](https://github.com/user-attachments/assets/ea1e5361-8863-4ebc-b733-155bc53bce72)
>> 
>> [JDK-8354981](https://bugs.openjdk.org/browse/JDK-8354981) added these `MemBarCPUOrders` around flat, atomic stores and then [JDK-8357474](https://bugs.openjdk.org/browse/JDK-8357474) also added them around loads. The same problem exists for `MemBarCPUOrders` emitted by stores. The memory input of following loads will be wired around them and an anti-dependency will be added.
>> 
>> The fix is to change `CastI2NNode` to a `TypeNode` that carries the exact type. This ways we can also avoid the `CastPPNode` after the decode.
>> 
>> Thanks,
>> Tobias
>
> Tobias Hartmann has updated the pull request incrementally with one additional commit since the last revision:
> 
>   New fix

Testing is all green. I'm integrating this to get the CI clean. Thanks again!

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

PR Comment: https://git.openjdk.org/valhalla/pull/1494#issuecomment-3023832618