// CHECK: %[[dq_w:.*]] = "tfl.dequantize"(%[[q_w]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>) -> tensor<64x3x3x3xf32>
// CHECK: %[[emb:.*]] = "tfl.gather"(%[[dq_w]], %arg0)
// CHECK: return %[[emb:.*]]

// PerTensor: %[[q_w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>

	_ = complex(f64, f32) // ERROR "complex"
	_ = complex(f32, F32) // ERROR "complex"
	_ = complex(F32, f32) // ERROR "complex"
	_ = complex(f64, F64) // ERROR "complex"
	_ = complex(F64, f64) // ERROR "complex"

	_ = complex(F1()) // ERROR "not enough arguments"
	_ = complex(F3()) // ERROR "too many arguments"

	_ = complex() // ERROR "not enough arguments"

	c128 = complex(f32, f32) // ERROR "cannot use"

!tf_res_f32 = tensor<*x!tf_type.resource<tensor<f32>>>

// CHECK-LABEL: func @hoist_tail_assign
// CHECK-SAME:  ([[ARG0:%.*]]: tensor<*x!tf_type.resource<tensor<f32>>>)
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"

  %1:4 = "tf.CustomAggregator"(%arg0) <{id = "1", calibration_method = 1 : i32, num_bins = 0 : i32, max_percentile = 0.000000e+00 : f32, min_percentile = 0.000000e+00 : f32}> {min = 7.547870e-07 : f32, max = 0.999992311 : f32} : (tensor<1x1024xf32>) -> (tensor<1x1024xf32>, tensor<f32>, tensor<f32>, tensor<*xi64>)

// CHECK: %[[q1:.*]] = "quantfork.qcast"(%[[maxpool]])
// CHECK-SAME: quant.uniform<i8:f32, 5.000000e-02:-10>
// CHECK: %[[dq1:.*]] = "quantfork.dcast"(%[[q1]])
// CHECK-SAME: quant.uniform<i8:f32, 5.000000e-02:-10>
// CHECK: %[[reshape:.*]] = "tf.Reshape"(%[[dq1]]
// CHECK: %[[q2:.*]] = "quantfork.qcast"(%[[reshape]])
// CHECK-SAME: quant.uniform<i8:f32, 5.000000e-02:-10>
// CHECK: %[[dq2:.*]] = "quantfork.dcast"(%[[q2]])

// -----

// CHECK-LABEL: @clip
func.func @clip(%arg0 : tensor<f32>, %arg1 : tensor<f32>, %arg2 : tensor<f32>) -> tensor<f32> {
  // CHECK: [[VAL:%.+]] = mhlo.clamp %arg1, %arg0, %arg2

  %0 = "tf.ClipByValue"(%arg0, %arg1, %arg2) : (tensor<f32>, tensor<f32>, tensor<f32>) -> tensor<f32>
  // CHECK: return [[VAL]]
  func.return %0 : tensor<f32>
}

// -----

// CHECK-LABEL: @clip_dynamic

  %input_zp = "tf.Const"() { value = dense<-10> : tensor<i32> } : () -> tensor<i32>
  %filter_scale = "tf.Const"() { value = dense<0.5> : tensor<f32> } : () -> tensor<f32>
  %filter_zp = "tf.Const"() { value = dense<0> : tensor<i32> } : () -> tensor<i32>
  %accum_scale = "tf.Const"() { value = dense<1.0> : tensor<f32> } : () -> tensor<f32>

// CHECK-LABEL: main
func.func @main(%arg0: tensor<f32>, %arg1: tensor<f32>) -> (tensor<f32>) {
  // CHECK: %[[INPUT0:.*]] = "tf.Placeholder.input"
  %0 = "tf.Placeholder.input"(%arg0) : (tensor<f32>) -> tensor<f32>
  // CHECK: %[[INPUT1:.*]] = "tf.Placeholder.input"
  %1 = "tf.Placeholder.input"(%arg1) : (tensor<f32>) -> tensor<f32>
  %2 = arith.constant dense<true> : tensor<i1>

  %0 = "tf.Const"() {value = dense<2.000000e+00> : tensor<f32>} : () -> tensor<f32>
  %1 = "tf.AddV2"(%0, %0) {device = "/job:worker/replica:0/task:0/device:CPU:0"} : (tensor<f32>, tensor<f32>) -> tensor<f32>
  %2 = "tf.AddV2"(%0, %0) {device = "/job:worker/replica:0/task:0/device:CPU:1"} : (tensor<f32>, tensor<f32>) -> tensor<f32>
  func.return
}

// CHECK: func @no_change_test
func.func @no_change_test() -> ()  {

  func.func @f(
    %arg0: tensor<f32> {tf.resource_name = "resource"}
  ) attributes { tf_saved_model.exported_names = ["foo.some_func"] } {
    func.return
  }

}

// -----

module attributes {tf_saved_model.semantics, tf_saved_model.under_construction} {
  "tf_saved_model.global_tensor"() { is_mutable, sym_name = "v", type = tensor<f32>, value = dense<42.0> : tensor<f32> } : () -> ()
  // CHECK: func @f
  func.func @f(