RFR: 8298935: fix cyclic dependency bug in create_pack logic in SuperWord::find_adjacent_refs [v7]

[Emanuel Peter]

On Mon, 20 Feb 2023 19:08:56 GMT, Jatin Bhateja <jbhateja at openjdk.org> wrote:

>> Thanks @jatin-bhateja . In essence, I am trusting `DepPreds` to give me the correct dependency information. Is that not correct? Do you have an example where `DepPreds` would have a `Load` depend on a `Load`, without a store in between?
>> 
>> Also: it seems that `SuperWord::independent_path` does nothing special about RAR, it just follows all dependencies of `DepPreds` recursively. This is what is used in `SuperWord::independent`. So if `DepPreds` could not be trusted, then we have the same issue here.
>> 
>> In the end, `find_dependence` should do nothing else but `SuperWord::independent`, except for a whole pack at once, and not only pairwise for two memops.
>
> You are correct Load-Load(RAR) is not captured in dependence graph anyways.
> 
> My above comment was in context of scenario b),  I verified that newly added find_dependence method will not find any circular intra-pack dependence in this case.

@jatin-bhateja Ah, I misunderstood. Thanks for clarifying. So you verified that WAR is not an issue? Thanks!

I agree that it should not be an issue. For a `Load -> Store` to block vectorization, we would have to find a dependency path from a `Load -> Store -> Load` (LSL) or a `Store -> Load -> Store` (SLS), since my verification only checks if any of the memops in the pack depend on another memop of the pack (both loads or both stores).
But both LSL and SLS have a `Store -> Load` (RAW) and so we cannot vectorize.
Do you agree with this argument?

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

PR: https://git.openjdk.org/jdk/pull/12350