// CHECK:       tf.OpA
// CHECK:       ^bb0
// CHECK:         (%[[BARG:[a-zA-Z0-9_]+]]: tensor<*xf32>)
// CHECK-NOT:   tf.OpB
// CHECK:       tf.OpC
// CHECK-SAME:    %[[BARG]]
func.func @remove_first_result(%arg0: tensor<*xf32>, %arg1: tensor<*xf32>) -> tensor<*xf32> {
  %0:2 = "tf.WhileRegion"(%arg0, %arg1) ({
  ^bb0(%arg2: tensor<*xf32>, %arg3: tensor<*xf32>):
    %1 = "tf.OpA"() {is_stateless = true} : () -> tensor<i1>

		*a = oa
		return nil
	}

	*a = oa
	return nil
}

// Opa - implements opa policy agent calls.
type Opa struct {
	args   Args
	client *http.Client
}

// Enabled returns if opa is enabled.
func Enabled(kvs config.KVS) bool {
	return kvs.Get(URL) != ""
}

// LookupConfig lookup Opa from config, override with any ENVs.

	Compression compress.Config `json:"compress"`

	// OpenID configuration
	OpenID openid.Config `json:"openid"`

	// External policy enforcements.
	Policy struct {
		// OPA configuration.
		OPA opa.Args `json:"opa"`

		// Add new external policy enforcements here.
	} `json:"policy"`

	LDAPServerConfig xldap.LegacyConfig `json:"ldapserverconfig"`

  // CHECK-NOT: "tf_device.launch"
  // CHECK: "tf.opA"
  "tf.opA"() : () -> tensor<i1>
  func.return
}

// Tests that single TensorFlow op with empty device attribute is not wrapped in `tf_device.launch`.
// CHECK-LABEL: func @empty_device_attribute
func.func @empty_device_attribute() {
  // CHECK-NOT: "tf_device.launch"
  // CHECK: "tf.opA"
  "tf.opA"() {device = ""} : () -> tensor<i1>
  func.return
}

      %0 = "tf.opA"(%arg0) : (tensor<i1>) -> tensor<i1>
      func.return %0 : tensor<i1>
    }
    func.func nested @_tpu_v1_compat_outlined_func1(%arg0: tensor<i1>, %arg1: tensor<f32>) -> (tensor<i1>, tensor<i32>) {
      %0 = "tf.opA"(%arg0) : (tensor<i1>) -> tensor<i1>
      %1 = "tf.opA"(%0) : (tensor<i1>) -> tensor<i1>

  func.return
}

// CHECK: [[CLUSTER:%.*]] = "tf_device.cluster"() ({
// CHECK:    [[OPA:%.*]] = "tf.opA"() {is_stateless = true} : () -> tensor<i32>
// CHECK:    "tf.opB"() ({
// CHECK:      "tf.opC"([[OPA]])
// CHECK:    tf_device.return [[OPA]]
// CHECK:    "tf.opD"([[CLUSTER]])

// Preceding user is using resource updated by a nested op.

func.func @hoist_tail_assign(%arg0: !tf_res_f32) {
  // CHECK: [[REPLICATE:%.*]]:4 = tf_device.replicate {n = 2 : i32}
  %replicate:2 = tf_device.replicate {n = 2 : i32} {
    // CHECK: [[OP_A:%.*]] = "tf.OpA"
    %op_a = "tf.OpA"() : () -> tensor<f32>
    // CHECK: [[OP_B:%.*]] = "tf.OpB"
    %op_b = "tf.OpB"() : () -> tensor<i32>
    // CHECK-NOT: tf.AssignVariableOp
    "tf.AssignVariableOp"(%arg0, %op_a) : (!tf_res_f32, tensor<f32>) -> ()

        "tf_device.launch"() ({
          %1 = "tf.opA"() : () -> tensor<!tf_type.string>
          "tf.EnqueueTPUEmbeddingArbitraryTensorBatch"(%1){_xla_replica_id = 0 : i64, device_ordinal = -1 : i64} : (tensor<!tf_type.string>) -> ()
          tf_device.return
        }) {device = "/job:worker/replica:0/task:0/device:CPU:0"} : () -> ()
        "tf_device.launch"() ({
          %1 = "tf.opA"() : () -> tensor<!tf_type.string>

        "tf_device.launch"() ({
          "tf.opA"() : () -> ()
          tf_device.return
        }) {device = "CORE_0"} : () -> ()
        tf_device.return
      }
      tf_executor.yield
    }
    tf_executor.fetch
  }
  func.return
}

// CHECK: "tf_device.launch"
// CHECK: device = "CORE_0"
// CHECK: "tf.opA"
// CHECK: "tf_device.launch"
// CHECK: device = "CORE_0"

// CHECK-LABEL: does_basic_reordering
func.func @does_basic_reordering() -> () {
   // CHECK: _TPUCompileMlir
   // CHECK-SAME: X
   // CHECK: _TPUCompileMlir
   // CHECK-SAME: Y
   // CHECK: OpA
   // CHECK: OpB
   // CHECK: OpC
   "tf.OpA"() : () -> ()
   %status_x, %program_x = "tf._TPUCompileMlir"() { metadata = "X", mlir_module = "..." } : () -> (tensor<!tf_type.string>, tensor<!tf_type.string>)
   "tf.OpB"() : () -> ()