[llvm] AMDGPU: Add codegen for atomicrmw operations usub_cond and usub_sat (PR #141068)

[via llvm-commits]

================
@@ -1808,10 +1814,10 @@ defm : GlobalFLATAtomicPats <"GLOBAL_ATOMIC_CMPSWAP_X2", "AMDGPUatomic_cmp_swap_
 defm : GlobalFLATAtomicPats <"GLOBAL_ATOMIC_XOR_X2", "atomic_load_xor_global", i64>;
 
 let SubtargetPredicate = isGFX12Plus in {
-  defm : GlobalFLATAtomicPatsRtnWithAddrSpace <"GLOBAL_ATOMIC_COND_SUB_U32", "int_amdgcn_atomic_cond_sub_u32", "global_addrspace", i32>;
+  defm : GlobalFLATAtomicPatsRtn <"GLOBAL_ATOMIC_COND_SUB_U32", "atomic_load_usub_cond_global", i32>;
 
   let OtherPredicates = [HasAtomicCSubNoRtnInsts] in
----------------
anjenner wrote:

Fixed in latest version of patch.

https://github.com/llvm/llvm-project/pull/141068