Index: include/clang/Basic/BuiltinsAMDGPU.def
===================================================================
--- include/clang/Basic/BuiltinsAMDGPU.def
+++ include/clang/Basic/BuiltinsAMDGPU.def
@@ -98,6 +98,8 @@
 BUILTIN(__builtin_amdgcn_ds_faddf, "ff*3fIiIiIb", "n")
 BUILTIN(__builtin_amdgcn_ds_fminf, "ff*3fIiIiIb", "n")
 BUILTIN(__builtin_amdgcn_ds_fmaxf, "ff*3fIiIiIb", "n")
+BUILTIN(__builtin_amdgcn_ds_append, "ii*3", "n")
+BUILTIN(__builtin_amdgcn_ds_consume, "ii*3", "n")
 
 //===----------------------------------------------------------------------===//
 // CI+ only builtins.
Index: lib/CodeGen/CGBuiltin.cpp
===================================================================
--- lib/CodeGen/CGBuiltin.cpp
+++ lib/CodeGen/CGBuiltin.cpp
@@ -12478,6 +12478,14 @@
   case AMDGPU::BI__builtin_amdgcn_fmed3f:
   case AMDGPU::BI__builtin_amdgcn_fmed3h:
     return emitTernaryBuiltin(*this, E, Intrinsic::amdgcn_fmed3);
+  case AMDGPU::BI__builtin_amdgcn_ds_append:
+  case AMDGPU::BI__builtin_amdgcn_ds_consume: {
+    Intrinsic::ID Intrin = BuiltinID == AMDGPU::BI__builtin_amdgcn_ds_append ?
+      Intrinsic::amdgcn_ds_append : Intrinsic::amdgcn_ds_consume;
+    Value *Src0 = EmitScalarExpr(E->getArg(0));
+    Function *F = CGM.getIntrinsic(Intrin, { Src0->getType() });
+    return Builder.CreateCall(F, { Src0, Builder.getFalse() });
+  }
   case AMDGPU::BI__builtin_amdgcn_read_exec: {
     CallInst *CI = cast<CallInst>(
       EmitSpecialRegisterBuiltin(*this, E, Int64Ty, Int64Ty, true, "exec"));
Index: test/CodeGenOpenCL/builtins-amdgcn.cl
===================================================================
--- test/CodeGenOpenCL/builtins-amdgcn.cl
+++ test/CodeGenOpenCL/builtins-amdgcn.cl
@@ -536,6 +536,18 @@
   *out = __builtin_amdgcn_s_getpc();
 }
 
+// CHECK-LABEL: @test_ds_append_lds(
+// CHECK: call i32 @llvm.amdgcn.ds.append.p3i32(i32 addrspace(3)* %ptr, i1 false)
+kernel void test_ds_append_lds(global int* out, local int* ptr) {
+  *out = __builtin_amdgcn_ds_append(ptr);
+}
+
+// CHECK-LABEL: @test_ds_consume_lds(
+// CHECK: call i32 @llvm.amdgcn.ds.consume.p3i32(i32 addrspace(3)* %ptr, i1 false)
+kernel void test_ds_consume_lds(global int* out, local int* ptr) {
+  *out = __builtin_amdgcn_ds_consume(ptr);
+}
+
 // CHECK-DAG: [[$WI_RANGE]] = !{i32 0, i32 1024}
 // CHECK-DAG: attributes #[[$NOUNWIND_READONLY:[0-9]+]] = { nounwind readonly }
 // CHECK-DAG: attributes #[[$READ_EXEC_ATTRS]] = { convergent }