On Wed, 1 Apr 2026 07:38:10 GMT, Eric Fang <[email protected]> wrote:

>> @erifan Yes, please do it separately, probably before this PR, right?
>> 
>> Right, there are two parts to this - they can be separate tasks:
>> - After boxing elimination, we should make sure that all vector nodes 
>> (better all nodes) are put on the worklist. This has some cost, but I think 
>> it is acceptable. We could also make this conditional: only do it if you 
>> find any vectors in the graph.
>> - We need to improve notification during IGVN, so that IGVN optimizations 
>> that traverse more than a single hop always get triggered. We can verify 
>> that with `VerifyIterativeGVN`.
>
> Hi @eme64 I created a PR for the missing identity optimizations, see 
> https://github.com/openjdk/jdk/pull/30529. The PR #30529 is based on this PR 
> #28313, because in this way I can write a stress test for the test failure we 
> encountered in this PR.
> 
> Of course, #30529 doesn't have to be based on #28313, so that the problem 
> encountered in this PR wouldn't exist in #30529, and then we could enhance 
> the IGVN notification for `VectorMaskCastNode` in this PR. It depends on 
> which PR we want to be merged first. What do you think?

Hi @erifan , I'm going on vacation for about 2 weeks, feel free to wait or find 
someone else to review.

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

PR Comment: https://git.openjdk.org/jdk/pull/28313#issuecomment-4183263262

Reply via email to