// match: (Mul64 x y)
	// result: (Int64Make (Add32 <typ.UInt32> (Mul32 <typ.UInt32> (Int64Lo x) (Int64Hi y)) (Add32 <typ.UInt32> (Mul32 <typ.UInt32> (Int64Hi x) (Int64Lo y)) (Select0 <typ.UInt32> (Mul32uhilo (Int64Lo x) (Int64Lo y))))) (Select1 <typ.UInt32> (Mul32uhilo (Int64Lo x) (Int64Lo y))))
	for {
		x := v_0
		y := v_1
		v.reset(OpInt64Make)
		v0 := b.NewValue0(v.Pos, OpAdd32, typ.UInt32)

		map1, map2 = fake, fake
	}
	return
}

func TestConvertCPUProfile(t *testing.T) {
	addr1, addr2, map1, map2 := testPCs(t)

	b := []uint64{
		3, 0, 500, // hz = 500
		5, 0, 10, uint64(addr1 + 1), uint64(addr1 + 2), // 10 samples in addr1
		5, 0, 40, uint64(addr2 + 1), uint64(addr2 + 2), // 40 samples in addr2
		5, 0, 10, uint64(addr1 + 1), uint64(addr1 + 2), // 10 samples in addr1
	}

  // CHECK-NEXT: "tf.AddV2"({{.*}}, [[v]]) {device = "/CPU:0"} : (tensor<i32>, tensor<i32>) -> tensor<i32>
  // CHECK-NEXT: return
  %const = "tf.Const"() {device = "/CPU:0", value = dense<0> : tensor<i32>} : () -> tensor<i32>
  %x = "tf.AddV2"(%const, %const) {device = "/CPU:0"} : (tensor<i32>, tensor<i32>) -> tensor<i32>
  %r = "tf.AddV2"(%arg, %x) {device = "/CPU:0"} : (tensor<i32>, tensor<i32>) -> tensor<i32>

  func.func @add_and_pack(%arg0: tensor<i1>) -> tensor<2xf32> {
    // CHECK-NOT: "tf.VarHandleOp"
    // CHECK-NOT: "tf.ReadVariableOp"
    // CHECK: %[[CONST:.*]] = "tf.Const"()
    // CHECK: %[[ADD:[0-9]*]] = "tf.AddV2"(%arg1, %[[CONST]])
    // CHECK: %[[PACK:[0-9]*]] = "tf.Pack"(%[[CONST]], %[[ADD]])
    // CHECK: return %[[PACK]]
    %0 = "tf.Const"() {value = dense<4.200000e+01> : tensor<f32>} : () -> tensor<f32>

    %0 = tf_executor.graph {
      %outputs, %control = tf_executor.island wraps "tf.Const"() {device = "", value = dense<1> : tensor<i32>} : () -> tensor<i32>
      %outputs_0, %control_1 = tf_executor.island wraps "tf.AddV2"(%arg0, %outputs) {device = ""} : (tensor<i32>, tensor<i32>) -> tensor<*xi32>
      tf_executor.fetch %outputs_0 : tensor<*xi32>
    }
    func.return %0 : tensor<*xi32>
  }

	"DIVWUCC",
	"DIVWUVCC",
	"DIVWUV",
	"MODUD",
	"MODUW",
	"MODSD",
	"MODSW",
	"EQV",
	"EQVCC",
	"EXTSB",
	"EXTSBCC",
	"EXTSH",
	"EXTSHCC",
	"FABS",
	"FABSCC",
	"FADD",
	"FADDCC",
	"FADDS",
	"FADDSCC",
	"FCMPO",
	"FCMPU",
	"FCTIW",
	"FCTIWCC",
	"FCTIWZ",
	"FCTIWZCC",
	"FDIV",
	"FDIVCC",
	"FDIVS",
	"FDIVSCC",
	"FMADD",
	"FMADDCC",
	"FMADDS",
	"FMADDSCC",

// CHECK-LABEL: func @otherVariantWhileBody
// CHECK:       [[CST:%.*]] = "tf.Const"()
// CHECK-NEXT:  [[ADD:%.*]] = "tf.AddV2"(%arg2, [[CST]])
// CHECK-NEXT:  [[TENSOR_MAP_INSERT_RESULT:%.*]] = "tf.TensorMapInsert"(%arg3, %arg2, %arg2)
// CHECK-NEXT:  [[ADD_2:%.*]] = "tf.AddV2"(%arg0, [[CST]])
// CHECK-NEXT:  return [[ADD_2]], %arg1, [[ADD]], [[TENSOR_MAP_INSERT_RESULT]]

  // CHECK-NEXT: %[[ADD2:.*]] = "tf.AddV2"(%[[BARG0]], %[[DELTA]])
  %add2 = "tf.AddV2"(%arg0, %constant) : (tensor<i32>, tensor<i32>) -> tensor<i32>
  // CHECK-NEXT: return %[[ADD2]], %[[ADD1]]
  %id = "tf.Identity"(%arg2) : (tensor<*x!tf_type.resource<tensor<f32>>>) -> tensor<*x!tf_type.resource<tensor<f32>>>

// CHECK: return %[[identity]] : tensor<1x1x2xf32>

// Tests that an AddV2 op accepting two bf16 operands is transformed into
// an AddV2 op that accepts two fp32 operands.
func.func @cast_bf16_add_v2_to_fp32(%arg0: tensor<2xbf16>, %arg1: tensor<2xbf16>) -> tensor<2xf32> {
  %0 = "tf.AddV2"(%arg0, %arg1) : (tensor<2xbf16>, tensor<2xbf16>) -> tensor<2xbf16>

	hi, lo1b := bits.Mul32(uint32(x>>32), n)
	lo1, c := bits.Add32(lo1a, lo1b, 0)
	hi += c
	if lo1 == 0 && lo0 < uint32(n) {
		n64 := uint64(n)
		thresh := uint32(-n64 % n64)
		for lo1 == 0 && lo0 < thresh {
			x := r.Uint64()
			lo1a, lo0 = bits.Mul32(uint32(x), n)
			hi, lo1b = bits.Mul32(uint32(x>>32), n)
			lo1, c = bits.Add32(lo1a, lo1b, 0)
			hi += c
		}
	}
	return hi
}