[vectorIntrinsics] RFR: 8274533: [Vector API] Incorrect computation by VectorMask.toLong operation after JDK-8273949

[Paul Sandoz]

On Thu, 30 Sep 2021 11:43:51 GMT, Jie Fu <jiefu at openjdk.org> wrote:

>> Hi Jei, 
>> I am working on an alternate patch which fixes instruction sequence in legacy storeMask patterns.
>> Will share it shortly, I think it will also address this issue, kindly review the patch once its posted.
>> 
>> Best Regards,
>> Jatin
>
>> Hi Jei, I am working on an alternate patch which fixes instruction sequence in legacy storeMask patterns. Will share it shortly, I think it will also address this issue, kindly review the patch once its posted.
>> 
>> Best Regards, Jatin
> 
> Hi @jatin-bhateja ,
> 
> This is just a small bug fix, which follows the code style of current implementation.
> Maybe you can still go on with your enhancement after this fix.
> 
> It took me hours to narrow down and fix the problem.
> Otherwise, it would be completely a waste of time to do this kind of things.
> Thanks.

@DamonFool I think Jatin is saying there is a broader architectural issue that needs to be fixed, encompassing the part you fixed. In that case i would prefer to wait and fix more broadly.

It's easy to get into these situations when we are all working diligently across various TZs. We can avoid them by better communication on the panama-dev list e.g. let's try and send quick emails identifying issues and intent of work. Alas, because of TZs, this can introduce some delay, and sometimes we need to make judgements on how fast to proceed. I will send a quick email reporting an issue Jatin is aware of and possibly related to this one.

So in this case, although i realize and appreciate your efforts, can you please place this PR on hold and give some time for Jatin submit his PR?

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

PR: https://git.openjdk.java.net/panama-vector/pull/138