[clang] [llvm] [AMDGPU][GFX12] Add 16 bit atomic fadd instructions (PR #75917)

[Matt Arsenault via cfe-commits]

================
@@ -1368,6 +1391,28 @@ def int_amdgcn_struct_ptr_buffer_atomic_cmpswap : Intrinsic<
 // gfx908 intrinsic
 def int_amdgcn_struct_buffer_atomic_fadd : AMDGPUStructBufferAtomic<llvm_anyfloat_ty>;
 def int_amdgcn_struct_ptr_buffer_atomic_fadd : AMDGPUStructPtrBufferAtomic<llvm_anyfloat_ty>;
+// gfx12 intrinsic
+def int_amdgcn_struct_buffer_atomic_fadd_v2bf16 : Intrinsic <
+  [llvm_v2i16_ty],
----------------
arsenm wrote:

Probably should start using v2bf16 for these, although we have the bad precedent of not using it for the existing operations (I'm currently working on fixing this for some other intrinsics)

https://github.com/llvm/llvm-project/pull/75917