// =============================================================================

func.func @sigmoidGrad(%arg0: tensor<?x32xf32>, %arg1: tensor<?x32xf32>) -> tensor<?x32xf32> {
  %0 = "tf.SigmoidGrad"(%arg0, %arg1) : (tensor<?x32xf32>, tensor<?x32xf32>) -> tensor<?x32xf32>
  func.return %0 : tensor<?x32xf32>
// CHECK-LABEL: sigmoidGrad
// CHECK-NEXT: [[ONE:%.+]] = arith.constant dense<1.000000e+00> : tensor<f32>

            "ReciprocalGrad", "RsqrtGrad", "SqrtGrad", "TruncateDiv",
            "TruncateMod", "Equal", "NotEqual", "Greater", "GreaterEqual",
            "Less", "LessEqual", "SigmoidGrad", "SoftplusGrad", "SoftsignGrad",
            "TanhGrad", "Pow", "SquaredDifference", "ApproximateEqual",
            // Others
            "AddN", "Bitcast", "Cast", "ClipByValue", "Const", "Empty",

  // CHECK-DAG: [[SUB:%.+]] =  mhlo.subtract [[ONE]], %arg0 : tensor<2xf32>
  // CHECK-DAG: [[MUL1:%.+]] =  mhlo.multiply [[MUL0]], [[SUB]] : tensor<2xf32>
  // CHECK: return [[MUL1]]
  %0 = "tf.SigmoidGrad"(%arg0, %arg1) : (tensor<2xf32>, tensor<2xf32>) -> tensor<2xf32>
  func.return %0 : tensor<2xf32>
}

// -----

// CHECK-LABEL: @sigmoid_grad_complex

    static bool isCompatibleReturnTypes(TypeRange inferred, TypeRange actual) {
      return ArraysAreCastCompatible(inferred, actual);
    }
  }];
}

def TF_SigmoidGradOp : TF_Op<"SigmoidGrad", [Pure, TF_SameOperandsAndResultTypeResolveRef]> {
  let summary = "Computes the gradient of the sigmoid of `x` wrt its input.";

  let description = [{