[PATCH] [ARMv8] New NEON instructions vmaxnm/vminnm.

Tim Northover t.p.northover at gmail.com
Tue Jul 16 03:42:28 PDT 2013


  Hi Joey,

  > This adds MC support for the NEON instructions vmaxnm and vminm. It also adds IR level intrinsics.

  I think there's one problem here, and a few things I don't understand
  at the moment:

  The problem: doesn't this enable these instructions even on v7 cores?
  I can't remember what exact constraints the v8 documents place on
  their accessibility, but that's probably not right.

  And my questions:
  1. These seem to be the first NEON instructions to have separate
  ARM/Thumb variants like that. Why didn't the conventional mechanism
  (whatever that is) work in this case?
  2. Why the "np" variants. Aren't all NEON instructions unpredicated?
  3. Why the "NEONDup" namespace? If shares a common feature with DUP,
  the namespace should probably be renamed to reflect that.

  Cheers.

  Tim

http://llvm-reviews.chandlerc.com/D1153



More information about the llvm-commits mailing list