[llvm] [AMDGPU] Extend llvm.amdgcn.update.dpp intrinsic to support f64 (PR #91190)

[Matt Arsenault via llvm-commits]

================
@@ -3338,13 +3338,15 @@ def : GCNPat <
                         (as_i1timm $bound_ctrl))
 >;
 
+foreach vt = [i64, f64] in {
----------------
arsenm wrote:

For the 16/32 cases, we currently have Reg32Types. Should create a new Reg64Types and start populating it for this 

https://github.com/llvm/llvm-project/pull/91190