RFR: 8342393: Promote commutative vector IR node sharing [v10]

[Jatin Bhateja]

On Thu, 23 Jan 2025 18:39:50 GMT, Jatin Bhateja <jbhateja at openjdk.org> wrote:

>> I suggest to go first with changes as they are - vector specific.  (We have to submit our testing to make sure they are safe first).
>> Scalar implementation will have to be investigated more carefully to make sure we will not get performance regression.
>> We can rename the flag and its accessors to general name, refactor some code as @eme64 suggested later when we do scalar work.
>
> Hi @vnkozlov , @iwanowww , @eme64 Please let me know if this looks good.

> Hi @jatin-bhateja , the new change is more clear to me. Thanks for updating the predicate IR handling. Looks fine!
> 
> The only concern to me now is the test. I saw a series of IR checks in the added test. It seems some ones check the similar scenario. I'v added the comment in the test file. Could you please check? Thanks!

Hi @XiaohongGong 
The intention here is to exercise all possible input combinations and IR ordering during GVN hashing, so there are explicit test points for (A1 + B1) + (B1 + A1) and (B1 + A1) + (A1 + B1) if that is what you are referring to.

Hi @eme64 , Can you kindly run the current patch through your test framework and share back the results

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

PR Comment: https://git.openjdk.org/jdk/pull/22863#issuecomment-2612146587