RFR: 8308994: C2: Re-implement experimental post loop vectorization

[Pengfei Li]

On Thu, 29 Jun 2023 10:54:29 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

>> Hi @eme64,
>> 
>> I guess you have done your first round of review. @fg1417 and I really appreciate all your constructive inputs. By reading your comments, I believe you have reviewed this patch in very detail. Thanks again!
>> 
>> What I am doing now:
>> 
>> - I'm trying to fix the issues which I think can be fixed immediately.
>> - I'm trying to answer all your simple questions ASAP.
>> 
>> For your request of big refactoring work, I feel like I personally may not have enough time and capability to complete it in a short time. We may need some discussion about it. But it's great to know more about your "hybrid vectorizer" plan from your feedback. It looks like a grand plan, and requires significant effort and cooperation. I strongly agree that we need some conversation to discuss where we should move forward and what we can cooperate. Could you give us a moment to digest your idea before we schedule a conversation?
>> 
>> BTW: What's your preferred time for a conversation? We are based in Shanghai (GMT+8)
>
> Hi @pfustc !
> 
> I'm grad you appreciate my review.
> 
>> For your request of big refactoring work, I feel like I personally may not have enough time and capability to complete it in a short time.
> 
> Are you under some time constraint? No pressure from my side, take the time you need.
> 
> I would very much love to have a conversation over a video call with you. I think that would be beneficial for all of us. The problem from our side (Oracle) are intellectual property concerns. OpenJDK emails and PR's are all under the Oracle Contributor Agreement. So there I'm free to have conversations. I'm trying to figure out if we can have a similar frame for a video call, sadly it may take a few weeks or months to get that sorted, as many people are on summer vacation.
> 
> Please take some time to digest the feedback. This is a big change set, it will take a while to be ready for integration at any rate. And again, I would really urge you to consider some refactoring of SuperWord in a separate RFE before this change here.
> 
> I'm looking forward to more collaboration - over PR comments, emails, and hopefully eventually video calls as well 😃 
> Emanuel

Hi @eme64, I just experimented some initial SuperWord refactoring work but found the refactoring process may cause more crashes/bugs with `PostLoopMultiversioning`. It seems that nobody is currently using this experimental feature and/or has interest to maintain it. If we have already reached a consensus that we will abandon it eventually, shall we propose a PR to remove it first before doing the refactoring? I think this way may speedup our refactoring process. An alternative approach is keeping the legacy code in SuperWord for now but tolerating new bugs of `PostLoopMultiversiong`, which already has many bugs. What's your opinion on this?

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

PR Comment: https://git.openjdk.org/jdk/pull/14581#issuecomment-1624936112