#loc3])\0Amodule {\0A  func.func @host_func(%arg0: tensor<3360x?xi32> loc(fused[#loc1, #loc2, #loc3]), %arg1: tensor<3xi32> loc(fused[#loc1, #loc2, #loc3]), %arg2: tensor<i32> loc(fused[#loc1, #loc2, #loc3]), %arg3: tensor<2xi32> loc(fused[#loc1, #loc2, #loc3])) -> (tensor<1x1120x?xi32>, tensor<1x1120x?xi32>, tensor<1120x?xi32>, tensor<2xi32>) {\0A    %0 = \22tf.Reshape\22(%arg0, %arg1) {_xla_outside_compilation = \220\22} : (tensor<3360x?xi32>, tensor<3xi32>) -> tensor<3x1120x?xi32> loc(#loc9)\0A    %1:3 = ...

#loc3])\0Amodule {\0A  func.func @host_func(%arg0: tensor<3360x?xi32> loc(fused[#loc1, #loc2, #loc3]), %arg1: tensor<3xi32> loc(fused[#loc1, #loc2, #loc3]), %arg2: tensor<i32> loc(fused[#loc1, #loc2, #loc3]), %arg3: tensor<2xi32> loc(fused[#loc1, #loc2, #loc3])) -> (tensor<1x1120x?xi32>, tensor<1x1120x?xi32>, tensor<1120x?xi32>, tensor<2xi32>) {\0A    %0 = \22tf.Reshape\22(%arg0, %arg1) {_xla_outside_compilation = \220\22} : (tensor<3360x?xi32>, tensor<3xi32>) -> tensor<3x1120x?xi32> loc(#loc9)\0A    %1:3 = ...

  %261 = "tf.Shape"(%254) {device = ""} : (tensor<?x?xi64>) -> tensor<2xi32>
  %262 = "tf.StridedSlice"(%261, %6, %15, %16) {begin_mask = 0 : i64, device = "", ellipsis_mask = 0 : i64, end_mask = 1 : i64, new_axis_mask = 0 : i64, shrink_axis_mask = 0 : i64} : (tensor<2xi32>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<0xi32>

  // CHECK-DAG: %[[INDICES_3:.*]] = "tf.Mul"(%[[INDICES_2]], %arg0) : (tensor<5xi32>, tensor<5xi32>) -> tensor<5xi32>
  // CHECK-DAG: %[[INDICES_4:.*]] = "tf.TensorScatterAdd"(%[[cst_3]], %0, %[[UPDATES]]) : (tensor<5xi32>, tensor<5x1xi32>, tensor<5xi32>) -> tensor<5xi32>
  // CHECK-DAG: %[[INDICES_5:.*]] = "tf.AddV2"(%[[INDICES_3]], %[[INDICES_4]]) : (tensor<5xi32>, tensor<5xi32>) -> tensor<5xi32>

  %shape1 = arith.constant dense<[64]> : tensor<1xi32>
  %0 = "tfl.reshape"(%arg0, %shape0) : (tensor<4x4x4xf32>, tensor<2xi32>) -> tensor<16x4xf32>
  %1 = "tfl.reshape"(%0, %shape1) : (tensor<16x4xf32>, tensor<1xi32>) -> tensor<64xf32>
  %2 = "tfl.reshape"(%0, %shape1) : (tensor<16x4xf32>, tensor<1xi32>) -> tensor<64xf32>
  %3 = arith.addf %1, %2 : tensor<64xf32>
  func.return %3 : tensor<64xf32>

func.func @modified(%arg0: tensor<1x224x224x3xf32>) -> tensor<1x401408xf32> attributes {tf.entry_function = {control_outputs = "", inputs = "input", outputs = "output"}} {
  %cst = arith.constant dense<[1, 401408]> : tensor<2xi32>
  %0 = "tfl.quantize"(%arg0) {qtype = tensor<1x224x224x3x!quant.uniform<i8:f32, 7.812500e-03>>} : (tensor<1x224x224x3xf32>) -> tensor<1x224x224x3x!quant.uniform<i8:f32, 7.812500e-03>>

func.func @foldDivI32() -> tensor<2xi32> {
  %0 = stablehlo.constant dense<[9, 3]> : tensor<2xi32>
  %1 = stablehlo.constant dense<[4, 6]> : tensor<2xi32>
  %2 = stablehlo.divide %0, %1 : tensor<2xi32>
  return %2 : tensor<2xi32>
}

// CHECK-LABEL: foldDivI32
// CHECK: stablehlo.constant dense<[2, 0]> : tensor<2xi32>

// -----

func.func @divideToMulReciprocalSplat(%arg0: tensor<2xf32>) -> tensor<2xf32> {

  return %0 : tensor<1x1600x1x1xf32>
}

//CHECK:func.func private @custom_tf_op(%arg0: tensor<1x320x1x1xf32>, %arg1: tensor<2xi32>) -> tensor<1x1600x1x1xf32> {
//CHECK: %0 = "vhlo.custom_call_v1"(%arg0, %arg1) <{

// CHECK-SAME: (tensor<1x2x2x3xi8>, tensor<2x2x3x2xi8>, tensor<2xi32>, tensor<f32>, tensor<i32>, tensor<2xf32>, tensor<2xi32>, tensor<2xf32>, tensor<2xi32>, tensor<f32>, tensor<i32>) -> tensor<*xf32>
// CHECK: return %[[conv_quant]]

// CHECK-LABEL: func private @quantized_conv2d_with_bias_and_relu6_float_output_fn_0

  %cst = arith.constant dense<[1, 2]> : tensor<2xi32>
  %0 = "tfl.mean"(%arg0, %cst) {keep_dims = false} : (tensor<1x48x48x512xf32>, tensor<2xi32>) -> tensor<1x512xf32>
  func.return %0 : tensor<1x512xf32>
}

// CHECK:       func @mean_4d_no_keepdim([[VAL_0:%.*]]: tensor<1x48x48x512xf32>) -> tensor<1x512xf32> {
// CHECK:           [[VAL_1:%.*]] = "tfl.pseudo_const"(){{.*}}dense<[1, 512]> : tensor<2xi32>