%0 = arith.constant dense<1.0> : tensor<1x2xf32>
  %1 = "tfl.concatenation"(%arg0, %0) {axis = 0 : i32, fused_activation_function = "NONE"} : (tensor<1x2xf32>, tensor<1x2xf32>) -> tensor<2x2xf32>
  func.return %1 : tensor<2x2xf32>

// CHECK-NEXT: %[[cst:.*]] = arith.constant dense<1.000000e+00> : tensor<1x2xf32>
// CHECK-NEXT: %[[cc:.*]] = "tfl.concatenation"(%arg0, %[[cst]])
// CHECK-NEXT: return %[[cc]]
}

    %22 = "tfl.concatenation"(%cst_6, %21) {axis = 0 : i32, fused_activation_function = "NONE"} : (tensor<1xi32>, tensor<?xi32>) -> tensor<?xi32>
    %23 = "tfl.reshape"(%arg2, %cst_12) : (tensor<i32>, tensor<1xi32>) -> tensor<1xi32>
    %24 = "tfl.fill"(%cst_11, %cst_9) : (tensor<1xi32>, tensor<i32>) -> tensor<?xi32>

monitoring::GaugeCell<std::string>& SavedModelWritePath();

// Returns "/tensorflow/core/saved_model/write/path_and_fingerprint" cell, which
// contains the path (saved_model_path) and fingerprint (concatenation of
// graph_def_program_hash, signature_def_hash, saved_object_graph_hash,
// and checkpoint_hash) of the SavedModel when it is exported.
monitoring::GaugeCell<std::string>& SavedModelWritePathAndSingleprint();

 %0 = "tfl.concatenation"(%arg0, %arg1, %arg2, %arg3) {axis = 0 : i32, fused_activation_function = "NONE"} : (tensor<384x384xf32>, tensor<384x384xf32>, tensor<384x384xf32>, tensor<384x384xf32>) -> tensor<?x?xf32>
  func.return %0: tensor<?x?xf32>
}

// CHECK-NOT: "tfl.reshape"

// (concatenation of graph_def_program_hash, signature_def_hash,
// saved_object_graph_hash, and checkpoint_hash) of the newly written
// SavedModel.
auto* saved_model_write_path_and_singleprint =
    monitoring::Gauge<std::string, 0>::New(
        "/tensorflow/core/saved_model/write/path_and_singleprint",
        "The path (saved_model_path) and singleprint (concatenation of "

// CHECK:  [[VAL_1:%.*]] = "tfl.reshape"(%2, %[[CST]]) {tac.device = "GPU",  tac.inference_type = "FLOAT"} : (tensor<1xf32>, tensor<4xi32>) -> tensor<1x1x1x1xf32>
// CHECK:  [[VAL_2:%.*]] = "tfl.concatenation"([[VAL_0]], [[VAL_1]]) <{axis = 3 : i32, fused_activation_function = "NONE"}> {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<1x1x1x1xf32>, tensor<1x1x1x1xf32>) -> tensor<1x1x1x2xf32>

  %2 = "tfl.var_handle"() {container = "", shared_name = "read_assign2/states"} : () -> tensor<!tf_type.resource>
  %3 = "tfl.read_variable"(%2) : (tensor<!tf_type.resource>) -> tensor<1x2x1x3xf32>
  %4 = "tfl.concatenation"(%3, %1) {axis = 1 : i32, fused_activation_function = "NONE"} : (tensor<1x2x1x3xf32>, tensor<1x32x1x3xf32>) -> tensor<1x34x1x3xf32>

// CHECK-NEXT:      %10 = "tfl.pseudo_const"() <{value = dense<4> : tensor<1xi32>}> : () -> tensor<1xi32>
// CHECK-NEXT:      %11 = "tfl.concatenation"(%10, %9, %8) <{axis = 0 : i32, fused_activation_function = "NONE"}> : (tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<3xi32>

    func.return %0 : tensor<2x1xf32>
    // CHECK: %[[VAL_0:.*]] = arith.constant dense<[2, 1]> : tensor<2xi32>
    // CHECK: %[[CONCAT:.*]] = "tfl.concatenation"(%arg0, %arg1) <{axis = 0 : i32, fused_activation_function = "NONE"}> : (tensor<1xf32>, tensor<1xf32>) -> tensor<2xf32>

// and increments the counter. The name of the counter is
// the concatenation of prefix and the flag name.
func CountFlags(prefix string, flagSet flag.FlagSet) {
	counter.CountFlags(prefix, flagSet)
}

// CountFlagValue creates a counter for the flag value
// if it is set and increments the counter. The name of the
// counter is the concatenation of prefix, the flagName, ":",
// and value.String() for the flag's value.