[llvm] [AMDGPU] Add support for preloading implicit kernel arguments (PR #83817)

[Matt Arsenault via llvm-commits]

================
@@ -138,10 +138,17 @@ let TargetPrefix = "amdgcn" in {
 // ABI Special Intrinsics
 //===----------------------------------------------------------------------===//
 
-defm int_amdgcn_workitem_id : AMDGPUReadPreloadRegisterIntrinsic_xyz;
+defm int_amdgcn_workitem_id : AMDGPUReadPreloadRegisterIntrinsic_xyz<llvm_i32_ty>;
 defm int_amdgcn_workgroup_id : AMDGPUReadPreloadRegisterIntrinsic_xyz_named
                                <"__builtin_amdgcn_workgroup_id">;
 
+// This intrinsic is used to track the user SGPRs that hold implicit kernel
----------------
arsenm wrote:

This should go down in the backend-only intrinsic place? 

https://github.com/llvm/llvm-project/pull/83817