").getType().cast<ShapedType>().getRank() >= " # m>]>>;

class TFL_OperandHasRankRange<int n, int x, int y> :
  PredOpTrait<"operand " # n # " has rank range [" # x # ", " # y # "]",
    Or<[TFL_OperandIsUnrankedPred<n>,
      CPred<"$_op.getOperand(" # n # ").getType().cast<ShapedType>().getRank() "
      ">= " # x # " && $_op.getOperand(" # n # ").getType().cast<ShapedType>()."
      "getRank() <= " # y>]>>;

// Converts the condition for an IfOp/WhileOp to a boolean value.
Value ConvertConditionToBoolean(Operation* op, Value cond) {
  if (auto ranked_type = mlir::dyn_cast<RankedTensorType>(cond.getType()))
    if (ranked_type.getRank() == 0 &&
        ranked_type.getElementType().isSignlessInteger(1))
      return cond;

  OpBuilder builder(op);
  Value to_bool = builder.create<TF::ToBoolOp>(op->getLoc(), cond);

  "GetHLOAxisFromTFAxis("
  "$0, $1.getType().cast<RankedTensorType>().getRank(), &$_builder)">;

// Same as the above but with $1 of type operand_range from variadic TensorFlow
// input.
def GetHLOAxisFromTFAxisVariadic : NativeCodeCall<
  "GetHLOAxisFromTFAxis("
  "$0, (*$1.begin()).getType().cast<RankedTensorType>().getRank(), "
  "&$_builder)">;

def CastElementsToI64Elements : NativeCodeCall<

                         (TFL_TopKV2Op $input, $k)>;

def ReductionDimensionIsLastDim : Constraint<CPred<"($0.cast<IntegerAttr>().getInt() == "
  "$1.getType().cast<ShapedType>().getRank() - 1 || $0.cast<IntegerAttr>().getInt() == -1)">>;

// Legalizes TF_ApproxTopKOp to TFL_TopKV2Op with the following constraints:
//    1. It computes max k
//    2. The reduction dimension is the last dim of the input.

// `ShapedType` or its rank is unknown.
inline bool HasRankOf(Value value, const int64_t rank) {
  auto shaped_type = mlir::dyn_cast_or_null<ShapedType>(value.getType());
  return shaped_type && shaped_type.hasRank() && shaped_type.getRank() == rank;
}

// Creates a new type that has the shape from the `old_type` and the element
// type from the `element_type`.
Type CloneTypeWithNewElementType(Type old_type, Type element_type);

        return (constant.getType().getRank() == 2);
      };

  auto op = cast<BatchMatMulOpType>(bmm_op);

  // Create a tfl.transpose op that performs ZX transpose on `input`.
  auto create_z_x_transpose_op = [&](Value input) -> Value {
    RankedTensorType input_type = mlir::cast<RankedTensorType>(input.getType());
    const int input_rank = input_type.getRank();

    return op.emitError("lhs must have static shape.");
  }
  if (!rhs_shape.hasStaticShape()) {
    return op.emitError("rhs must have static shape.");
  }

  const int64_t padding_nums_size = 2 * (rhs_shape.getRank() - 2);
  padding_nums.reserve(padding_nums_size);
  if (conv_padding.strref() == "EXPLICIT") {
    for (auto padding_elem :
         op.getExplicitPaddingAttr().template getAsRange<IntegerAttr>()) {

  }
  return success();
}

inline ShapedType MergeType(ShapedType a, ShapedType b) {
  if (!a.hasRank()) {
    return b;
  }
  if (!b.hasRank()) {
    return a;
  }
  int64_t rank = a.getRank();
  SmallVector<int64_t, 4> dims;
  dims.resize(rank);
  for (int i = 0, e = rank; i != e; i++) {
    int64_t dim0 = a.getDimSize(i);
    int64_t dim1 = b.getDimSize(i);

// packed_value = bitwise_or(packed_low, packed_high)
Value PackOperand(OpBuilder &builder, Location loc, Value value, int pack_dim) {
  ShapedType value_type = mlir::cast<ShapedType>(value.getType());
  const int rank = value_type.getRank();

  SmallVector<int64_t> packed_shape(value_type.getShape().begin(),
                                    value_type.getShape().end());
  RankedTensorType shape_type =

      CPred<"$_op.getOperand(" # n #
      ").getType().cast<ShapedType>().getRank() == " # m>]>>;

// Returns true if the n-th result has unknown rank or has rank m.
class TF_ResultHasRank<int n, int m> :
  PredOpTrait<"result " # n # " is " # m # "-D",
    Or<[TF_ResultIsUnrankedPred<n>,
      CPred<"$_op.getResult(" # n #
      ").getType().cast<ShapedType>().getRank() == " # m>]>>;