[PATCH] D66413: [ARM} Add support for MVE vmaxv

[Sam Tebbs via Phabricator via llvm-commits]

samtebbs marked 2 inline comments as done.
samtebbs added inline comments.


================
Comment at: llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp:1064
     return false;
+  case Instruction::ICmp:
   case Instruction::Add:
----------------
dmgreen wrote:
> Is there a VMINV patch too? Am I correct that this would enable both min and max reduction intrinsics?
It would enable both. There isn't a separate patch for vminv so I could add it to this one if you'd like.


================
Comment at: llvm/test/CodeGen/Thumb2/mve-vmaxv.ll:14
+; CHECK-NEXT:    mov.w r0, #-128
+; CHECK-NEXT:    vmaxv.s8 r0, q0
+; CHECK-NEXT:    bx lr
----------------
dmgreen wrote:
> The old code for these was quite large, right? Much larger than needing to materialise the constant?
Yep, in this case it was 78 instructions long! And the others varied from 15 to 61.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D66413/new/

https://reviews.llvm.org/D66413