[llvm] [AMDGPU] Don't DEALLOC_VGPRS from callable functions (PR #72245)

[Matt Arsenault via llvm-commits]

================
@@ -556,3 +558,27 @@ body:             |
     S_WAITCNT_VSCNT undef $sgpr_null, 0
     S_ENDPGM 0
 ...
+
+---
+name:            gfx_function
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: gfx_function
+    ; CHECK-NOT: S_SENDMSG 3
+    ; CHECK: S_ENDPGM 0
+    GLOBAL_STORE_DWORD undef renamable $vgpr0_vgpr1, killed renamable $vgpr1, 0, 4, implicit $exec
+    S_WAITCNT_VSCNT undef $sgpr_null, 0
+    S_ENDPGM 0
+...
+
+---
+name:            ccc_function
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: ccc_function
+    ; CHECK-NOT: S_SENDMSG 3
+    ; CHECK: S_ENDPGM 0
+    GLOBAL_STORE_DWORD undef renamable $vgpr0_vgpr1, killed renamable $vgpr1, 0, 4, implicit $exec
+    S_WAITCNT_VSCNT undef $sgpr_null, 0
+    S_ENDPGM 0
+...
----------------
arsenm wrote:

Tangentially related but do we avoid inserting waits before unreachable in noreturn functions?

https://github.com/llvm/llvm-project/pull/72245