[llvm-branch-commits] [llvm] AMDGPU: Handle legal v2f16/v2bf16 atomicrmw fadd for global/flat (PR #95394)

[Stanislav Mekhanoshin via llvm-branch-commits]

================
@@ -1669,13 +1670,16 @@ defm : FlatSignedAtomicPatWithAddrSpace <"FLAT_ATOMIC_ADD_F32", "int_amdgcn_flat
 }
 
 let OtherPredicates = [HasAtomicFlatPkAdd16Insts] in {
+// FIXME: These do not have signed offsets
----------------
rampitec wrote:

Can you just use FlatAtomicPat?

https://github.com/llvm/llvm-project/pull/95394