llvm/test/CodeGen/AMDGPU/atomic-oversize.ll - third_party/llvm-project - Git at Google

 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck %s

 define void @test(ptr %a) nounwind {
 ; CHECK-LABEL: test:
 ; CHECK: __atomic_load_16
 ; CHECK: __atomic_store_16
   %1 = load atomic i128, ptr %a seq_cst, align 16
   store atomic i128 %1, ptr %a seq_cst, align 16
   ret void
 }
	; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s \| FileCheck %s

	define void @test(ptr %a) nounwind {
	; CHECK-LABEL: test:
	; CHECK: __atomic_load_16
	; CHECK: __atomic_store_16
	%1 = load atomic i128, ptr %a seq_cst, align 16
	store atomic i128 %1, ptr %a seq_cst, align 16
	ret void
	}