On Thu, 28 May 2026 11:00:00 GMT, Marc Chevalier <[email protected]> wrote:
> The point is not really to limit the amount of nodes, but rather to limit how > fast they come. > > As suggested on JBS, I've made `make_scalar_in_safepoints` able to process a > single safepoint, and `SafePointNode::Ideal` is just taking care of itself. I > have added an overload for compatibility. The name is a bit awkward, > especially the plural, since if the new last argument is provided, it doesn't > work on multiple safepoints. Yet it does the same, and we can argue it still > works on all the safepoints contained in a singleton set. Besides, > `make_scalar_in_safepoint` exists, and is a helper of > `make_scalar_in_safepoints`, so I'd rather not mess with names too much. > Similar semantics get the same name. > > Conceptually, it also sounds nicer to me that idealizing a safepoint is not > touching all the other cousin safepoints at the same time. > > Thanks, > Marc > > --------- > - [x] I confirm that I make this contribution in accordance with the [OpenJDK > Interim AI Policy](https://openjdk.org/legal/ai). This pull request has now been integrated. Changeset: 10ca1d34 Author: Marc Chevalier <[email protected]> URL: https://git.openjdk.org/valhalla/commit/10ca1d34bb13f5a101ac88bb5d64356ff2b8da78 Stats: 136 lines in 5 files changed: 122 ins; 2 del; 12 mod 8373598: [lworld] C2 asserts with "excessive live node increase in single iteration of IGVN" Reviewed-by: thartmann ------------- PR: https://git.openjdk.org/valhalla/pull/2480
