[vectorIntrinsics+mask] RFR: 8264563: Add masked vector intrinsics for binary/store operations [v5]

[Paul Sandoz]

On Thu, 22 Apr 2021 09:42:12 GMT, Xiaohong Gong <xgong at openjdk.org> wrote:

>>> > > I am tempted to have a template method for both non-mask and mask. Thus specialized code (that before calling the intrinsic) may be somewhat duplicated before calling `binaryMaskOp`, using a constant for the op -> lambda function, and passing in null or the mask value.
>>> > 
>>> > 
>>> > Yeah, that's why I added the `lanewise0` and `lanewise0Template` methods.
>>> 
>>> I am suggesting to separate out templates for non-masked and masked, similar to the current pattern. That would result in some duplication for the special cases, but i think makes it easier to reason about in one place, rather than being somewhat spread out. Perhaps easier to explain as code. We should just iterate on the branch.
>> 
>> Thanks for your more details! I think I misunderstood your point before. I will think about your suggestion. Thanks so much!
>
>> > > > I am tempted to have a template method for both non-mask and mask. Thus specialized code (that before calling the intrinsic) may be somewhat duplicated before calling `binaryMaskOp`, using a constant for the op -> lambda function, and passing in null or the mask value.
>> > > 
>> > > 
>> > > Yeah, that's why I added the `lanewise0` and `lanewise0Template` methods.
>> > 
>> > 
>> > I am suggesting to separate out templates for non-masked and masked, similar to the current pattern. That would result in some duplication for the special cases, but i think makes it easier to reason about in one place, rather than being somewhat spread out. Perhaps easier to explain as code. We should just iterate on the branch.
>> 
>> Thanks for your more details! I think I misunderstood your point before. I will think about your suggestion. Thanks so much!
> 
> Hi @PaulSandoz , the latest commit changes something according to your suggestion. It totally separateS the special handles for the special cases for non-masked and masked version, although finally they still use one template `lanewise0Template`. Could you please take a look again? If it matches what you suggested above, do you think it need to separate out the template methods further? Thanks!

@XiaohongGong See the following diff which more accurately depicts what i was suggesting:
https://gist.github.com/PaulSandoz/d93fa2a78c3c7be5fdb2d714371639ce
i.e. we follow the exiting template pattern more directly in both the non-mask and mask.
Perhaps we find later on we can consolidate, but right now i find that pattern easier to follow and it is very clear on the null contract for no mask.

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

PR: https://git.openjdk.java.net/panama-vector/pull/57