(TFL_RangeOp $start, $limit, $delta)>;
def LegalizeRelu6 : Pat<(TF_Relu6Op $arg), (TFL_Relu6Op $arg)>;
def LegalizeRelu : Pat<(TF_ReluOp $arg), (TFL_ReluOp $arg)>;
// TFL Relu doesn't support I32/I64 type, so legalizes TF Relu to TFL Maximum.
def LegalizeReluI32 :
  Pat<(TF_ReluOp TensorOf<[I32]>:$arg),
      (TFL_MaximumOp $arg,
        (Arith_ConstantOp ConstantAttr<RankedI32ElementsAttr<[]>,"0">))>;

  let hasFolder = 1;
}

def TFL_ReluOp: TFL_Op<"relu", [
    PredOpTrait<"x and y must have same element type",
      TFL_TCresVTEtIsSameAsOp<0, 0>>,
    Pure,
    QuantizableResult,
    SameOperandsAndResultShape]> {
  let summary = "Relu operator";

  let description = [{
    Element-wise Relu operator
      x -> max(0, x)
  }];

  // CHECK: "NONE"
  %0 = tfl.add %arg0, %arg1 {fused_activation_function = "NONE"} : tensor<4xi32>
  // CHECK: "RELU"
  %1 = tfl.add %arg0, %arg1 {fused_activation_function = "RELU"} : tensor<4xi32>
  // CHECK: "RELU_N1_TO_1"
  %2 = tfl.add %arg0, %arg1 {fused_activation_function = "RELU_N1_TO_1"} : tensor<4xi32>
  // CHECK: "RELU6"

// CHECK: "tfl.dynamic_update_slice"(%arg0, %arg1, %arg2) : (tensor<4x5xi32>, tensor<1x5xi32>, tensor<2xi64>) -> tensor<4x5xi32>
}

func.func @testReluI32(%arg0: tensor<1xi32>) -> tensor<1xi32> {
  %0 = "tf.Relu"(%arg0) : (tensor<1xi32>) -> tensor<1xi32>
  func.return %0: tensor<1xi32>

// CHECK-LABEL: testReluI32
// CHECK:  %[[CONST_0:.*]] = arith.constant dense<0> : tensor<i32>

  let summary = "Computes rectified linear gradients for a Relu operation.";

  let arguments = (ins
    Arg<TF_IntOrFpTensor, [{The backpropagated gradients to the corresponding Relu operation.}]>:$gradients,
    Arg<TF_IntOrFpTensor, [{The features passed as input to the corresponding Relu operation, OR
the outputs of that operation (both work equivalently).}]>:$features
  );

    to matrix multiplication and convolution, these building blocks include:
    Direct batched convolution Pooling: maximum, minimum, average Normalization:
    LRN, batch normalization Activation: rectified linear unit (ReLU) Data
    manipulation: multi-dimensional transposition (conversion), split, concat,
    sum and scale.

*   TensorForest Estimator now supports SavedModel export for serving.