[llvm] [AArch64][GlobalISel] Lower unmerge to extract_subvector (PR #195046)

[David Green via llvm-commits]

================
@@ -317,7 +317,7 @@ define i16 @mulv_v8i16(<8 x i16> %a) {
 ;
 ; CHECK-GI-LABEL: mulv_v8i16:
 ; CHECK-GI:       // %bb.0: // %entry
-; CHECK-GI-NEXT:    mov d1, v0.d[1]
+; CHECK-GI-NEXT:    ext v1.16b, v0.16b, v0.16b, #8
----------------
davemgreen wrote:

It matches what we do for SDAG, and is (almost always) the same performance-wise. See #195010 which will change it in both places.

https://github.com/llvm/llvm-project/pull/195046