RFR: 8333393: PhaseCFG::insert_anti_dependences can fail to raise LCAs and to add necessary anti-dependence edges [v4]

[Daniel Lundén]

On Fri, 17 Jan 2025 08:09:59 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

>> Daniel Lundén has updated the pull request incrementally with two additional commits since the last revision:
>> 
>>  - Fix comma splice in comment
>>  - Update after comments
>
> src/hotspot/share/opto/gcm.cpp line 781:
> 
>> 779:     assert(initial_mem_block->dominates(early), "invariant");
>> 780:     // If the load has an explicit control input, walk up the dominator tree
>> 781:     // from the early block (inclusive) to the initial memory block
> 
> Would be nice to give a motivation for this: why is that all of the relevant blocks?

I have updated the comment now, please check if it is clearer now. `initial_mem` and relevant Phi memory aliases that we must add as search roots must all be in blocks that dominate the early block (where the load can at earliest be scheduled).

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

PR Review Comment: https://git.openjdk.org/jdk/pull/22852#discussion_r1920686652