[llvm] [AMDGPU] Follow up for Split OPERAND_REG_IMM_INT64 into signed and unsigned variants (PR #219755)

[Matt Arsenault via llvm-commits]

================
@@ -3213,6 +3213,20 @@ def VOP_I64_I64_I32 : VOPProfile <[i64, i64, i32, untyped]>;
 def VOP_I64_I32_I64 : VOPProfile <[i64, i32, i64, untyped]>;
 def VOP_I64_I64_I64 : VOPProfile <[i64, i64, i64, untyped]>;
 
+// Signed 64-bit operand variants for arithmetic shift operations
+def VOP_I64_I64_I32_Signed : VOPProfile <[i64, i64, i32, untyped]> {
----------------
arsenm wrote:

The 'Signed' is the I part, and applies to individual operands not the operation as a whole 

https://github.com/llvm/llvm-project/pull/219755