(x.Args[0] == y.Args[0] && x.Args[1] == y.Args[1] ||
			x.Args[0] == y.Args[1] && x.Args[1] == y.Args[0]) {
		return true
	}
	if x.Op == Op386LEAL1 && y.Op == Op386LEAL1 && y.AuxInt == x.AuxInt+n && x.Aux == y.Aux &&
		(x.Args[0] == y.Args[0] && x.Args[1] == y.Args[1] ||
			x.Args[0] == y.Args[1] && x.Args[1] == y.Args[0]) {
		return true
	}

//   %4 = mhlo.constant dense<0.000000e+00> : tensor<f32>
//   %5 = "mhlo.broadcast"(%4)
//   %6 = "mhlo.select"(%3, %input, %5)
//   %7 = "mhlo.reduce"(%6, %4) ({
//   ^bb0(%arg1: tensor<f32>, %arg2: tensor<f32>):
//     %9 = mhlo.add %arg1, %arg2 : tensor<f32>
//     "mhlo.return"(%9) : (tensor<f32>) -> ()
//   }) {dimensions = dense<0> : tensor<1xi64>}
//

  }];

  let arguments = (ins
    Arg<TF_Float32Tensor, [{A 4D Tensor for input data.}]>:$x,
    Arg<TF_Float32Tensor, [{A 1D Tensor for scaling factor, to scale the normalized x.}]>:$scale,
    Arg<TF_Float32Tensor, [{A 1D Tensor for offset, to shift to the normalized x.}]>:$offset,
    Arg<TF_Float32Tensor, [{A 1D Tensor for population mean. Used for inference only;