// to produce the same result. It first slices the first `$index` rows. Then
  // expands the dimension of the `$item`, followed by another slice of the
  // remaining rows starting from `$index` + 1. Lastly it concatenates the
  // three parts together.
  // On a high level, it's doing something like:
  // def : Pat<(TF_TensorListSetItemOp $input, $index, $item),
  //      (Concat
  //        concat_dim = 0,

// each config will result in a final dest rule set (*.foo.com, and *.com).
//
// The following is the merge logic:
// 1. Unique subsets (based on subset name) are concatenated to the original rule's list of subsets
// 2. If the original rule did not have any top level traffic policy, traffic policies from the new rule will be
// used.

  [
    Pure,
    TFL_SameFirstOperandAndFirstResultElementType,
    SameOperandsAndResultsScale,
    QuantizableResult
  ]> {
  let summary = "Concatenation operator";

  let description = [{
    Concatenates tensors along one dimension
  }];

  let arguments = (
    ins TFL_VariadicTensorOf<
      [F32, I64, I32, I16, I8, QI8, QUI8, UI8, UI32, I1]>:$values,
    I32Attr:$axis,

}

def TF_ConcatOp : TF_Op<"Concat", [Pure]> {
  let summary = "Concatenates tensors along one dimension.";

  let arguments = (ins
    Arg<TF_Int32Tensor, [{0-D.  The dimension along which to concatenate.  Must be in the
range [0, rank(values)).}]>:$concat_dim,
    Arg<Variadic<TF_Tensor>, [{The `N` Tensors to concatenate. Their ranks and types must match,

    ): FirLiteralExpression? {
        val concatenated = buildString {
            for (arg in stringConcatenationCall.arguments) {
                val evaluated = evaluate(arg) ?: return null
                append(evaluated.value.toString())
            }
        }

        return ConstantValueKind.String.toLiteralExpression(stringConcatenationCall.source, concatenated)
    }

    private fun evaluateFunctionCall(

   * position + 8. The array must have at least 8 bytes from offset (inclusive).
   *
   * @param input the input bytes
   * @param offset the offset into the array at which to start
   * @return a long of a concatenated 8 bytes
   */
  static long load64(byte[] input, int offset) {
    // We don't want this in production code as this is the most critical part of the loop.
    assert input.length >= offset + 8;

   * position + 8. The array must have at least 8 bytes from offset (inclusive).
   *
   * @param input the input bytes
   * @param offset the offset into the array at which to start
   * @return a long of a concatenated 8 bytes
   */
  static long load64(byte[] input, int offset) {
    // We don't want this in production code as this is the most critical part of the loop.
    assert input.length >= offset + 8;

new files. **Do not add RELNOTE=yes comments in CLs.** Instead, add a file to
the CL (or ask the author to do so).

At the end of the development cycle, the files will be merged by being
concatenated in sorted order by pathname. Files in the directory matching the
glob "*stdlib/*minor" are treated specially. They should be in subdirectories
corresponding to standard library package paths, and headings for those package

  %15 = mhlo.add %13, %4 : tensor<32x32xi64>
  %16 = mhlo.select %14, %15, %13 : tensor<32x32xi1>, tensor<32x32xi64>
  %17 = mhlo.reshape %16 : (tensor<32x32xi64>) -> tensor<32x32x1xi64>
  %18 = "mhlo.concatenate"(%12, %17) <{dimension = 2 : i64}> : (tensor<32x32x1xi64>, tensor<32x32x1xi64>) -> tensor<32x32x2xi64>

		in.PrintText(&buf)
	}
	return buf.String()
}

// Merge combines the markdown documents (files ending in ".md") in the tree rooted
// at fs into a single document.
// The blocks of the documents are concatenated in lexicographic order by filename.
// Heading with no content are removed.
// The link keys must be unique, and are combined into a single map.
//