[llvm] [AMDGPU]: Fall back to default mutations when iglp is not applied (PR #93418)

[Matt Arsenault via llvm-commits]

================
@@ -18,3 +18,23 @@ body:             |
     %2 = V_ADD_F32_e64 0, killed %1, 0, 1, 0, 0, implicit $mode, implicit $exec
     %3 = FLAT_LOAD_DWORD %0, 4, 0, implicit $exec, implicit $flat_scr :: (load (s32))
 ...
+---
+# GCN-LABEL: name: cluster_flat_loads_iglp_opt
+# GCN:      FLAT_LOAD_DWORD %0, 0
+# GCN-NEXT: FLAT_LOAD_DWORD %0, 4
+# GCN-NEXT: V_ADD_F32_e64
+name: cluster_flat_loads_iglp_opt
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: vreg_64 }
+  - { id: 1, class: vgpr_32 }
+  - { id: 2, class: vgpr_32 }
+  - { id: 3, class: vgpr_32 }
----------------
arsenm wrote:

Should use inline register class syntax and drop the registers section 

https://github.com/llvm/llvm-project/pull/93418