RFR: 8333334: C2: Make result of `Node::dominates` more precise to enhance scalar replacement [v3]

[Tobias Hartmann]

On Thu, 13 Jun 2024 09:42:03 GMT, MaxXing <duke at openjdk.org> wrote:

> For the latter case, the control node chain should be replaced by top and then eliminated (otherwise there must be something suboptimal).

Right, my concern was that not all control nodes would be folded but we have asserts that would catch such an endless loop during IGVN.

> I have already added a standalone IR test that does not rely on ConcurrentHashMap to this patch.

Great! Please have a look at the failure that I reported above.

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

PR Comment: https://git.openjdk.org/jdk/pull/19496#issuecomment-2167681307