[llvm] [AArch64] Keep floating-point conversion in SIMD (PR #147707)

[Paul Walker via llvm-commits]

================
@@ -6627,6 +6627,17 @@ def : Pat<(f16 (any_uint_to_fp (i32 (any_fp_to_uint f16:$Rn)))),
           (UCVTFv1i16 (f16 (FCVTZUv1f16 f16:$Rn)))>;
 }
 
+let Predicates = [HasNEONandIsSME2p2StreamingSafe] in {
----------------
paulwalker-arm wrote:

Is this necessary, it looks like you're already inside a `HasNEONandIsSME2p2StreamingSafe` block?

https://github.com/llvm/llvm-project/pull/147707