%3 = "tfl.reshape"(%2, %cst_0) {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<1x1x1x2xf32>, tensor<1xi32>) -> tensor<2xf32>
    %4 = "tfl.reshape"(%3, %cst_1) {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<2xf32>, tensor<2xi32>) -> tensor<2x1xf32>
    return %4 : tensor<2x1xf32>
  }
```

#### Compute Costs Pass
In the compute cost pass, we will essentially compute the cost of each op within

  // CHECK-SAME: %[[ARG3:[^:[:space:]]+]]
  func.func private @main(%arg0: tensor<1x4xf32>, %arg1: tensor<4x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module} {
    %0 = stablehlo.dot_general %arg0, %arg1, contracting_dims = [1] x [0] : (tensor<1x4xf32>, tensor<4x3xf32>) -> tensor<1x3xf32>
    return %0 : tensor<1x3xf32>
    // CHECK: %[[DOT:.+]] = stablehlo.dot_general %[[ARG2]], %[[ARG3]]
    // CHECK: return %[[DOT]]
  }

  func.return %24 : tensor<1x4xf32>

  func.return %24 : tensor<1x4xf32>

  %1 = "tfl.quantize"(%0) {qtype = tensor<2x1x!quant.uniform<i8:f32, 0.024986599940879671:92>>} : (tensor<2x1xf32>) -> tensor<2x1x!quant.uniform<i8:f32, 0.024986599940879671:92>>
  %2 = "tfl.dequantize"(%1) : (tensor<2x1x!quant.uniform<i8:f32, 0.024986599940879671:92>>) -> tensor<2x1xf32>

module {
  // CHECK-LABEL: main
  func.func @main(%arg0: tensor<4xf32>, %arg1: tensor<4xf32>) -> tensor<4xf32> {
    %0 = "tfl.squared_difference"(%arg0, %arg1) : (tensor<4xf32>, tensor<4xf32>) -> tensor<4xf32>
    func.return %0 : tensor<4xf32>
    // CHECK:  [[VAL_0:%.*]] = tfl.sub %arg0, %arg1 {fused_activation_function = "NONE"} : tensor<4xf32>

func.func @transpose_simple_1d() -> tensor<2xf32> {
  %0 = stablehlo.constant dense<[0.000000e+0, 1.000000e+0]> : tensor<2xf32>
  %1 = stablehlo.transpose %0, dims = [0] : (tensor<2xf32>) -> tensor<2xf32>
  return %1 : tensor<2xf32>
}
// CHECK-DAG: %[[CONST_0:.+]] = stablehlo.constant dense<[0.000000e+00, 1.000000e+00]> : tensor<2xf32>
// CHECK-NOT: transpose
// CHECK: return %[[CONST_0]] : tensor<2xf32>

// -----

  func.return
}

// CHECK-LABEL: func @no_embedding_ops
func.func @no_embedding_ops(%arg0: tensor<2x2xf32>) -> (tensor<2x2xf32>) {
  // CHECK: tf.Add
  %0 = "tf.Add"(%arg0, %arg0) : (tensor<2x2xf32>, tensor<2x2xf32>) -> tensor<2x2xf32>
  func.return %0 : tensor<2x2xf32>
}

// CHECK-LABEL: func @nested_embedding_op

  %3 = "tfl.pack"(%1, %2) {axis = 0 : i32, values_count = 2 : i32} : (tensor<1xf32>, tensor<1xf32>) -> tensor<2x1xf32>
  func.return %3 : tensor<2x1xf32>
}

// CHECK: %[[CST:.*]] = arith.constant dense<1> : tensor<4xi32>

    return %2 : tensor<1x3xf32>
  }

  func.func private @composite_dot_general_fn(%arg0: tensor<1x2xf32>, %arg1: tensor<2x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module} {
    %0 = stablehlo.dot_general %arg0, %arg1, contracting_dims = [1] x [0] : (tensor<1x2xf32>, tensor<2x3xf32>) -> tensor<1x3xf32>
    return %0 : tensor<1x3xf32>
  }
}