RFR: 8269725: AArch64: Add VectorMask query implementation for NEON [v2]

Fri Jul 9 06:56:55 UTC 2021

On Fri, 9 Jul 2021 06:52:53 GMT, Xiaohong Gong <xgong at openjdk.org> wrote:

>> Yes, but you reverse the bits in $dst, which reverses the byte order and changes each 0x01 byte to 0x80. You then count the leading zeroes. What, then, is the point of negating each 0x01 byte to 0xff first?
>
> Hi @theRealAph , the new commit has removed the redundant `negr` for `firsttrue, lasttrue`, and all tests pass. Would you mind having look at it again? Thanks so much!

> > Seems `negr` is also needless for `lasttrue`, since `7 >> 3 == 0` ?
> 
> Yeah. And the in-reg mask representation of these VectorMaskOp nodes (0/1) are different from anytrue/alltrue (VectorTest, 0/-1), which looks a bit confusing.

Yes, to make it clear, I'v added some comments for it. Thanks!

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

PR: https://git.openjdk.java.net/jdk/pull/4699