// CHECK: %[[BROADCAST:.*]] = "mhlo.broadcast_in_dim"(%arg0) <{broadcast_dimensions = dense<1> : tensor<1xi64>}> : (tensor<4xui8>) -> tensor<2x4xui8>
  // CHECK:  mhlo.shift_right_logical %[[BROADCAST]], %arg1 : tensor<2x4xui8>
  %0 = "tf.RightShift"(%arg0, %arg1) : (tensor<4xui8>, tensor<2x4xui8>) -> tensor<2x4xui8>
  func.return %0 : tensor<2x4xui8>
}

}

// CHECK-LABEL: EliminateRedundantLogicalOr
func.func @EliminateRedundantLogicalOr(%arg0: tensor<1xi1>, %arg1: tensor<2x3xi1>) -> (tensor<1xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<1xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<2x3xi1>, tensor<2x3xi1>) {
  %cst_false0 = arith.constant dense<false> : tensor<1xi1>

func.func @testBroadcastGradientArgsInvalidS0Rank() -> (tensor<2x2xi32>, tensor<0xi32>) {
  %s0 = "tf.Const"() {value = dense<[[4, 1], [2, 3]]> : tensor<2x2xi32>} : () -> tensor<2x2xi32>
  %s1 = "tf.Const"() {value = dense<[2, 4]> : tensor<2xi32>} : () -> tensor<2xi32>
  // expected-error @+1 {{failed to verify that operand 0 is 1-D}}
  %r0, %r1 = "tf.BroadcastGradientArgs"(%s0, %s1) : (tensor<2x2xi32>, tensor<2xi32>) -> (tensor<1xi32>, tensor<0xi32>)

func.func @DynamicStitch_uint8(%arg0: tensor<2x2xui8>) -> tensor<2x2xui8> {
  // CHECK-NOT: tf.DynamicStitch

  %indices = "tf.Const"() {value = dense<[1, 0]> : tensor<2xi32>} : () -> tensor<2xi32>
  %0 = "tf.DynamicStitch"(%indices, %arg0) : (tensor<2xi32>, tensor<2x2xui8>) -> tensor<2x2xui8>
  func.return %0 : tensor<2x2xui8>
}

// CHECK-LABEL: func @DynamicStitch_scalar_item

  func.func @xla_spmd_full_to_shard_shape(%arg0: tensor<2x2xi64>) -> (tensor<1x2xi64>) {
    // CHECK: %[[SHARDING:.*]] = mhlo.custom_call @Sharding(%arg0) {backend_config = "", mhlo.sharding = "{devices=[2,1]0,1}"} : (tensor<2x2xi64>) -> tensor<2x2xi64>
    // CHECK: %[[MANUAL:.*]] = mhlo.custom_call @SPMDFullToShardShape(%[[SHARDING]]) {backend_config = "", mhlo.sharding = "{manual}"} : (tensor<2x2xi64>) -> tensor<1x2xi64>
    // CHECK: return %[[MANUAL]]

// PerTensor-SAME: f = @quantized_matmul_fn_0}> : (tensor<2x12xf32>, tensor<12x2xi8>, tensor<f32>, tensor<i32>) -> tensor<*xf32>
// PerTensor: return %[[out]]

// PerChannel-LABEL: func @matmul
// PerChannel-DAG: %[[q_w:.*]] = "tf.Const"() <{value = dense<0> : tensor<12x2xi8>}> : () -> tensor<12x2xi8>
// PerChannel-DAG: %[[scale:.*]] = "tf.Const"() <{value = dense<3.93700805E-9> : tensor<f32>}> : () -> tensor<f32>

  %ch0 = tfrt.new.chain

  %a1 = tfrt_fallback_async.const_dense_tensor dense<[[2, 2], [2, 2]]> : tensor<2x2xi32>
  %a2 = tfrt_fallback_async.const_dense_tensor dense<[[3, 3], [3, 3]]> : tensor<2x2xi32>
  %b = tfrt_fallback_async.const_dense_tensor dense<[[1, 1], [1, 1]]> : tensor<2x2xi32>

  // Two batch_size=2 batches get concatenated.

// -----

func.func @transpose(%arg0 : tensor<2x2xi32>, %arg1 : tensor<2xi32>) -> tensor<2x2xi32> {
  %0 = "tfl.transpose"(%arg0, %arg1) : (tensor<2x2xi32>, tensor<2xi32>) -> tensor<2x2xi32>
  func.return %0 : tensor<2x2xi32>
}

// -----

// CHECK-LABEL: transpose_with_output_that_has_dynamic_sizes
func.func @transpose_with_output_that_has_dynamic_sizes(%arg0 : tensor<2x2xi32>, %arg1 : tensor<2xi32>) -> tensor<?x?xi32> {

%0 = "mhlo.dot_general"(%arg0, %arg1) {
  dot_dimension_numbers = #mhlo.dot<
    lhs_batching_dimensions = [0],
    rhs_batching_dimensions = [0],
    lhs_contracting_dimensions = [3],
    rhs_contracting_dimensions = [3]
  >} : (tensor<2x2x?x3xf32>, tensor<2x4x?x3xf32>) -> tensor<2x2x?x4x?xf32>
func.return %0 : tensor<2x2x?x4x?xf32>

  %cst_0 = arith.constant dense<4> : tensor<2x2xi32>
  %0 = "tf.SpaceToBatchND"(%arg0, %cst, %cst_0) : (tensor<1x128x128x3xf32>, tensor<2xi32>, tensor<2x2xi32>) -> tensor<4x68x68x3xf32>
  %1 = "tf.Conv2D"(%0, %arg1) {padding = "VALID", strides = [1, 1, 1, 1]} : (tensor<4x68x68x3xf32>, tensor<5x5x3x8xf32>) -> tensor<4x64x64x8xf32>
  %2 = "tf.BatchToSpaceND"(%1, %cst, %cst_0) : (tensor<4x64x64x8xf32>, tensor<2xi32>, tensor<2x2xi32>) -> tensor<1x120x120x8xf32>