RFR JDK-8223347 Integration of Vector API (Incubator): Java API, implementation, and tests

[John Rose]

Thanks, Paul!  Talking with you about it helped me formulate my thoughts better.

> On May 6, 2020, at 9:02 AM, Paul Sandoz <paul.sandoz at oracle.com> wrote:
> 
> Hi John,
> 
> Thanks.  For the benefit of others, John and I had a long chat about this and Joe’s CSR comments.
> 
> I have a better appreciation of your approach to the design and some of the more subjective aspects to guide developers to API points, and to make code more readable (that’s creative API design :-) ).
> 
> I agree with your assessment on size, lane count, and Mask/Shuffle.vectorSpecies.
> 
> Re: VectorSpecies.fromByteArray, I now see the method Vector.reinterpretShape appeals to VectorSpecies.fromByteArray for its specification.  Removal would result in a less elegant specification of the behavior (making harder to understand).  In that sense I think it’s worth its weight.  However, I would suggest keeping in sync with a proposed change (on panama-dev) to the related load/store byte[]/ByteBuffer methods, requiring they all accept a ByteOrder.
> I think this does bring up the wider point you raised about where factory methods reside, and I agree about waiting for specialized generics, as that might allow us to make better moves.
> 
> Paul.