[PATCH] D44585: [AMDGPU] Scalarize when scalar code cheaper than vector code.

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: test/CodeGen/AMDGPU/scalarize.ll:97
+entry:
+ %rdx.shuf1 = shufflevector <2 x i16> %a, <2 x i16> %b, <2 x i32> <i32 1, i32 2>
+ %bin.rdx2 = add <2 x i16> %a, %rdx.shuf1
----------------
arsenm wrote:
> This out of bounds vector index is undefined
Nevermind, I forgot how shuffle vector works


https://reviews.llvm.org/D44585