RFR: 8361699: C2: assert(can_reduce_phi(n->as_Phi())) failed: Sanity: previous reducible Phi is no longer reducible before SUT

[Roberto Castañeda Lozano]

On Thu, 11 Sep 2025 07:41:55 GMT, Roberto Castañeda Lozano <rcastanedalo at openjdk.org> wrote:

>>> @robcasloz - are you thinking that the "fixed point" loops on `find_scalar_replaceable_allocs` aren't sufficient?
>> 
>> You're right, that should do.
>> 
>>> At first glance yes, I think that the code would be more cleaned up if done that way. If the code had been written like that in the first place we wouldn't have seen the current issue. (...)
>> 
>> Agree, a single fixed point loop combining NSR detection and propagation would be ideal for clarity and maintainability.
>> 
>>>  I propose that we move forward with the current patch and work on this refactoring as a separate issue.
>> 
>> Sounds good, please file a RFE for that. I would suggest then to postpone the clean-up in `revisit_reducible_phi_status` to that RFE.
>
>> @robcasloz - I pushed some changes addressing yours and @eme64 comments. Could you please re-run your internal tests?
> 
> Thanks, I will report back within a couple of days.

> Thank you @robcasloz ; I'll start working on that early next week.

@JohnTortugo thanks.

Please, keep in mind that [HotSpot requires two approvals for non-trivial changes like this](https://openjdk.org/guide/#hotspot-development) (apologies if my previous comment somehow could be interpreted as an invitation to integrate, that was not the intention).

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

PR Comment: https://git.openjdk.org/jdk/pull/27063#issuecomment-3305759603