// CHECK: tac.device = "GPU", tac.inference_type = "FLOAT"
  %0 = "tfl.depthwise_conv_2d"(%arg0, %arg1, %arg2) {depth_multiplier = 1 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<1x112x112x32xf32>, tensor<1x3x3x32xf32>, tensor<32xf32>) -> tensor<1x112x112x32xf32>
  func.return %0 : tensor<1x112x112x32xf32>
}

// -----

    %0:2 = "tf_device.launch"() ({
	  %1 = "tf.Cast"(%arg0) {Truncate = false} : (tensor<2048xi64>) -> tensor<2048xi32>
	  %2 = "tf.Cast"(%arg1) {Truncate = false} : (tensor<2048xi64>) -> tensor<2048xi32>

  }) : (tensor<i32>, tensor<1xf32>, tensor<i32>) -> (tensor<i32>, tensor<1xf32>, tensor<i32>)
  func.return %0#1, %0#2 : tensor<1xf32>, tensor<i32>

    %2 = "tf.Const"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>
    %3 = "tf.Const"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>
    %4 = "tf.Add"(%2, %3) : (tensor<i32>, tensor<i32>) -> tensor<i32>
    "tf.AssignAddVariableOp"(%0, %4) {dtype = "tfdtype$DT_INT32"} : (tensor<*x!tf_type.resource<tensor<2x8xi32>>>, tensor<i32>) -> ()

// CHECK-LABEL: import_stats_skip
func.func @import_stats_skip(%arg0: tensor<4xf32>, %cst: tensor<i32>) -> (tensor<2xf32>,tensor<2xf32>) {
  %0:2 = "tfl.split"(%cst, %arg0) {num_splits = 2 : i32} : (tensor<i32>, tensor<4xf32>) -> (tensor<2xf32>, tensor<2xf32>)
    loc(fused["skip1", "skip2.cc":10:8, callsite("op" at "skip3.cc":10:8)])
  func.return %0#0, %0#1 : tensor<2xf32>, tensor<2xf32>

    // CHECK: name: "fo.o"
    %1:2 = tf_executor.island wraps "tf.Const"() {dtype = "tfdtype$DT_INT32", value = dense<1> : tensor<i32>} : () -> (tensor<i32>) loc("fo{o")
    // CHECK: name: "foo"
    %2:2 = tf_executor.island wraps "tf.Const"() {dtype = "tfdtype$DT_INT32", value = dense<2> : tensor<i32>} : () -> (tensor<i32>) loc("foo@1")
    // CHECK: name: "ba.r"

	use(a1k[ui16])
	use(a100k[ui16]) // ERROR "index bounds check elided"
	use(p1[ui16])
	use(p1k[ui16])
	use(p100k[ui16]) // ERROR "index bounds check elided"

	use(s[i32])
	use(a1[i32])
	use(a1k[i32])
	use(a100k[i32])
	use(p1[i32])
	use(p1k[i32])
	use(p100k[i32])

	use(s[ui32])
	use(a1[ui32])
	use(a1k[ui32])
	use(a100k[ui32])
	use(p1[ui32])
	use(p1k[ui32])
	use(p100k[ui32])

	use(s[i64])

  %us = tfrt.constant.i32 1000
  %ch0 = "tfrt_test.blocking.usleep"(%us) : (i32) -> (!tfrt.chain)

  %ch1 = tfrt.merge.chains %ch, %ch0 : !tfrt.chain, !tfrt.chain

  %ch2 = tfrt_fallback_async.createop(%ch1) key(0) device("/CPU:0") "tf.MatMul"() {T = i32} num_args(2)

  %ch3, %result = tfrt_fallback_async.executeop.seq(%ch2) key(0) cost(100) device("/CPU:0") "tf.MatMul"(%a, %b) {T = i32}  : 1

  %next_index = "tf.AddV2"(%index, %one) {device = "/job:localhost/replica:0/task:0/device:CPU:0"} : (tensor<i32>, tensor<i32>) -> tensor<i32>

        tf_device.return %5 : tensor<i1>
      }, {
        %5 = "tf.opC"(%1#0) : (tensor<i1>) -> tensor<i32>
        tf_device.return %5 : tensor<i32>
      }) {} : () -> (tensor<i1>, tensor<i32>)
      tf_executor.yield %4#0, %4#1 : tensor<i1>, tensor<i32>
    }
    tf_executor.fetch %3#0, %3#1 : tensor<i1>, tensor<i32>
  }
  func.return
}

// CHECK:      [[INPUT_A:%.+]], {{%.+}} = tf_executor.island {