require(name.isNotEmpty()) { "name is empty" }
  for (i in name.indices) {
    val c = name[i]
    require(c in '\u0021'..'\u007e') {
      "Unexpected char 0x${c.charCode()} at $i in header name: $name"
    }
  }
}

internal fun headersCheckValue(
  value: String,
  name: String,
) {
  for (i in value.indices) {
    val c = value[i]
    require(c == '\t' || c in '\u0020'..'\u007e') {

import org.opensearch.action.DocWriteResponse.Result;
import org.opensearch.action.admin.indices.create.CreateIndexResponse;
import org.opensearch.action.admin.indices.exists.indices.IndicesExistsResponse;
import org.opensearch.action.admin.indices.mapping.get.GetMappingsResponse;
import org.opensearch.action.admin.indices.refresh.RefreshResponse;
import org.opensearch.action.bulk.BulkItemResponse;

  if (!ranked_ty) return UnrankedTensorType::get(element_ty);
  int64_t rank = ranked_ty.getRank();

  DenseIntElementsAttr indices;
  if (!matchPattern(reduction_indices, m_Constant(&indices))) {
    // Output type is unranked if reduction indices are not constant and reduced
    // dimensions are not kept.
    if (!keep_dims.getValue()) return UnrankedTensorType::get(element_ty);

    return success();
  }
};

// Lowers DynamicStitch op with constant indices and with static input and
// output shapes using Reshape, UnPack and Pack op.
//
//   %indices0 = "tf.Const"() {value = dense<4> : tensor<i32>}
//   %indices1 = "tf.Const"() {value = dense<[[3, 2], [1, 0]]> :
//   tensor<2x2xi32>} %0 = "tf.DynamicStitch"(%indices0, %indices1, %arg0,
//   %arg1)

  SmallVector<int64_t> indices(index_map_size);
  for (int64_t i = 0; i < index_map_size; i++) {
    indices[i] = dimension_numbers.start_index_map()[i];
  }

  // Fill elements from start_indices with start_index_map
  Value scattered_start_indices = builder.create<TF::TensorScatterUpdateOp>(
      loc, empty_start_indices,
      /*indices=*/
      builder.create<TF::ReshapeOp>(

    // CHECK: "tf.AssignVariableOp"([[VAR]], [[TENSOR]])
    "tf.ResourceScatterAdd"(%resource, %indices, %updates) : (tensor<*x!tf_type.resource<tensor<*xi32>>>, tensor<2x?xi32>, tensor<?x?x?xi32>) -> ()

    tf_device.return
  }) : () -> ()
  func.return
}

// Tests that composite tf.ResourceScatterAdd with 1d Indices operation is decomposed.

// CHECK-LABEL: @decompose_resource_scatter_add_op_1d_indices

  SmallVector<TransposeOp, 2> transpose_ops;

  // Constant operation that defines permutation indices for result transposes.
  ConstOp permutation_op;

  // All operation results must be used by transpose operations with the same
  // permutation indices.
  for (OpResult result : op->getResults()) {
    for (Operation* user : result.getUsers()) {

  let description = [{
    The op compiles and executes a GPU cluster function.

    func_name is the name of the function to be executed on GPU.
    resource_indices are the indices of inputs that are resources.
    used_output_indices are the indices of outputs that have users.

    Example:
      %results = gpurt.compile_and_execute {func_name = "xla_func_0", resource_indices = [1] ...}
  }];

  let arguments = (ins

  // CHECK: [[ch3:%.*]], [[results:%.*]]:2 = tfrt_fallback_async.get_resource [[ready_ch]] {device = "/device:CPU:0", indices = [0, 1]}
  // CHECK: tfrt_fallback_async.executeop key({{.*}}) cost({{.*}}) device("/device:CPU:0") "tf.MatMul"([[results]]#0, [[results]]#1)
  %a, %b = "tf._TfrtGetResource"() {device = "/device:CPU:0", indices = [0, 1], shared_name = ["", ""], container = ["", ""]} : () -> (tensor<3x1xf32>, tensor<1x3xf32>)

//
// values, indices = tfl.topkv2(%inputs, K)
// %1 = tfl.slice(values, 0, k)
// %2 = tfl.slice(indices,0, k)
// .... (values and indices only used for %1 and %2)
//
// %1 or %2 can be absent. If values and indices are only used here,
// this pattern can be replaced with (conceptually)
//
// %values, %indices = tfl.topkv2(%inputs, k)
// replace all use of %1 with values