[jdk16] RFR: 8259641: C2: assert(early->dominates(LCA)) failed: early is high enough

[Roland Westrelin]

On Thu, 14 Jan 2021 15:37:44 GMT, Tobias Hartmann <thartmann at openjdk.org> wrote:

>> For the fix of 8258393 that causes this regression, I wrote that the
>> goal was:
>> 
>> "I propose to be less conservative in anti-dependence analysis for
>> Phis. For a Phi, when computing the LCA, I think it's sufficient to
>> only consider region's inputs that we actually reach by following the
>> memory edges and that's what I propose here."
>> 
>> But that's not what the logic I added does. It checks whether a Phi
>> input should be taken into account by checking whether it's in the
>> worklist but the worklist doesn't include mem, the memory state from
>> which we started.
>
> Looks good to me.

Thanks for the reviews, Christian & Tobias.

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

PR: https://git.openjdk.java.net/jdk16/pull/121