[llvm] AMDGPU GlobalISel Add64 support (PR #124763)

[Matt Arsenault via llvm-commits]

================
@@ -56,9 +79,14 @@ define amdgpu_kernel void @lshl_add_u64_v2s(i64 %a) {
 
 define amdgpu_kernel void @lshl_add_u64_s2s(i64 %v, i64 %a) {
----------------
arsenm wrote:

This should really use a scalar return, I think this pattern could break pure scalar arithmetic situations 

https://github.com/llvm/llvm-project/pull/124763