(SelectOp $condition, $input2, $input1)>;

  // Fuse select(broadcast_to(input, shape), x, y) -> selectV2(input, x, y)
  // Also, fuse selectv2(broadcast_to(input, shape), x, y) -> selectV2(input, x, y)
  // It is safe to perform this transform here because-
  // the shapes of `pre_broadcast` and `dim` must be broadcast
  // compatible for the `broadcast_to` op to be valid.

    pass_manager.addPass(mlir::odml::CreateLegalizeHloToTfLitePass());
  }
  // TF dialect passes
  pass_manager.addPass(mlir::odml::CreateLegalizeHloToTfPass());

  // folds tf.BroadcastTo ops with subsequent ops if they have built in
  // broadcasting support. This needs to be run immediately after HLO->TF
  // legalization; otherwise other passes like `ConvertTFBroadcastTo` will

  SCATTER_ND = 122,
  SELECT_V2 = 123,
  DENSIFY = 124,
  SEGMENT_SUM = 125,
  BATCH_MATMUL = 126,
  PLACEHOLDER_FOR_GREATER_OP_CODES = 127,
  CUMSUM = 128,
  CALL_ONCE = 129,
  BROADCAST_TO = 130,
  RFFT2D = 131,
  CONV_3D = 132,
  IMAG=133,
  REAL=134,
  COMPLEX_ABS=135,
  HASHTABLE = 136,
  HASHTABLE_FIND = 137,
  HASHTABLE_IMPORT = 138,
  HASHTABLE_SIZE = 139,

// Create a pass that convert ops that copy tensors between devices, e.g.
// tf.Identity.
std::unique_ptr<OperationPass<mlir::func::FuncOp>>
CreateTensorDeviceCopyConversionPass();

// Returns a pass that folds tf.BroadcastTo nodes with subsequent nodes if they
// have built in broadcasting support.
std::unique_ptr<OperationPass<func::FuncOp>> CreateBroadcastFoldPass();

void populateTfControlFlowToScfPatterns(MLIRContext* context,

by a scalar, broadcasting (usually) confers some time or space
benefit, as the broadcasted tensor is never materialized.

However, `broadcast_to` does not carry with it any such benefits.
The newly-created tensor takes the full memory of the broadcasted
shape. (In a graph context, `broadcast_to` might be fused to
subsequent operation and then be optimized away, however.)
  }];

  let arguments = (ins

- Cel: converting a quantity value into a quantity value failed. ([#123669](https://github.com/kubernetes/kubernetes/pull/123669), [@pohly](https://github.com/pohly)) [SIG API Machinery]
- Client-go/tools/record.Broadcaster: fixed automatic shutdown on WithContext cancellation ([#124635](https://github.com/kubernetes/kubernetes/pull/124635), [@pohly](https://github.com/pohly)) [SIG API Machinery]

  %emptyf = "tf.Cast"(%emptyi) : (tensor<i32>) -> tensor<f32>
  %size_shape = "tf.Const"() {value = dense<[10, 8, 4]> : tensor<3xi32>} : () -> tensor<3xi32>
  %tl = "tf.BroadcastTo"(%emptyf, %size_shape) : (tensor<f32>, tensor<3xi32>) -> tensor<10x8x4xf32>
  // TensorListPushBack lowering
  %index_in_list = "tf.Const"() {value = dense<0> : tensor<1xi32>} : () -> tensor<1xi32>

example:

>>> x = tf.constant([1, 2, 3])   # Shape (3,)
>>> y = tf.broadcast_to(x, [2, 3])


When doing broadcasted operations such as multiplying a tensor
by a scalar, broadcasting (usually) confers some time or space
benefit, as the broadcasted tensor is never materialized.

However, `broadcast_to` does not carry with it any such benefits.

- K8s.io/client-go/tools/record.EventBroadcaster: after Shutdown() is called, the broadcaster now gives up immediately after a failure to write an event to a sink. Previously it tried multiple times for 12 seconds in a goroutine. ([#115514](https://github.com/kubernetes/kubernetes/pull/115514), [@pohly](https://github.com/pohly)) [SIG API...

    "SCATTER_ND",
    "SELECT_V2",
    "DENSIFY",
    "SEGMENT_SUM",
    "BATCH_MATMUL",
    "PLACEHOLDER_FOR_GREATER_OP_CODES",
    "CUMSUM",
    "CALL_ONCE",
    "BROADCAST_TO",
    "RFFT2D",
    "CONV_3D",
    "IMAG",
    "REAL",
    "COMPLEX_ABS",
    "HASHTABLE",
    "HASHTABLE_FIND",
    "HASHTABLE_IMPORT",
    "HASHTABLE_SIZE",
    "REDUCE_ALL",