class NamedParallelDevice {
 public:
  NamedParallelDevice(const std::string& name,
                      std::unique_ptr<ParallelDevice> parallel_device)
      : device_name_(name), parallel_device_(std::move(parallel_device)) {}
  const std::string& name() const { return device_name_; }
  const ParallelDevice& device() const { return *parallel_device_; }

 private:
  std::string device_name_;
  std::unique_ptr<ParallelDevice> parallel_device_;
};

    shape_ = std::vector<int64_t>(dim_sizes.begin(), dim_sizes.end());
  }
  *shape = &*shape_;
  return absl::OkStatus();
}

Status ParallelTensor::SummarizeValue(std::string& summary) {
  summary = "{";
  std::vector<std::string> summarized_devices = device_.SummarizeDeviceNames();
  for (int component_index = 0; component_index < tensors_.size();
       ++component_index) {

  // Process output names.
  std::vector<string> output_names_vec;
  if (output_names) {
    output_names_vec.reserve(noutputs);
    for (int i = 0; i < noutputs; ++i) {
      output_names_vec.push_back(string(output_names[i]));
    }
  }

  // Process control output names.
  std::vector<string> control_output_names_vec;
  if (control_output_names) {

  ExpectScalarEq<float>(second_components[1].get(), 9.);

  // Verify that the mirrors are placed on the component devices.
  std::string first_device = TFE_TensorHandleBackingDeviceName(
      second_components[0].get(), status.get());
  ASSERT_EQ(second_underlying_devices[0], first_device);
  std::string second_device = TFE_TensorHandleBackingDeviceName(
      second_components[1].get(), status.get());

    ExpectScalarEq<float>(components[0].get(), 20.);
    ExpectScalarEq<float>(components[1].get(), 20.);

    std::string first_device =
        TFE_TensorHandleBackingDeviceName(components[0].get(), status.get());
    ASSERT_EQ(underlying_devices[0], first_device);
    std::string second_device =
        TFE_TensorHandleBackingDeviceName(components[1].get(), status.get());
    ASSERT_EQ(underlying_devices[1], second_device);

  std::string translated_src = TranslateName(src);
  std::string translated_target = TranslateName(target);
  ops_->copy_file(filesystem_.get(), translated_src.c_str(),
                  translated_target.c_str(), plugin_status.get());
  return StatusFromTF_Status(plugin_status.get());
}

std::string ModularFileSystem::TranslateName(const std::string& name) const {

TEST(RamFileBlockCacheTest, ValidateAndUpdateFileSignature) {
  int calls = 0;
  auto fetcher = [&calls](const string& filename, size_t offset, size_t n,
                          char* buffer, TF_Status* status) -> int64_t {
    calls++;
    memset(buffer, 'x', n);
    TF_SetStatus(status, TF_OK, "");
    return n;
  };
  string filename = "file";
  tf_gcs_filesystem::RamFileBlockCache cache(16, 32, 0, fetcher);
  std::vector<char> out;

#include "tensorflow/core/platform/test_benchmark.h"
#include "tensorflow/core/protobuf/cluster.pb.h"
#include "tensorflow/core/protobuf/config.pb.h"

using tensorflow::string;

namespace tensorflow {
namespace {

static bool HasSubstr(absl::string_view base, absl::string_view substr) {
  bool ok = absl::StrContains(base, substr);
  EXPECT_TRUE(ok) << base << ", expected substring " << substr;
  return ok;
}

      RemoveFile_Locked(std::string(it->first.first));
    }
  }
}

void RamFileBlockCache::Flush() {
  absl::MutexLock lock(&mu_);
  block_map_.clear();
  lru_list_.clear();
  lra_list_.clear();
  cache_size_ = 0;
}

void RamFileBlockCache::RemoveFile(const std::string& filename) {
  absl::MutexLock lock(&mu_);
  RemoveFile_Locked(filename);
}

#include "tensorflow/core/platform/status.h"
#include "tensorflow/core/platform/test.h"

using tensorflow::Status;
using tensorflow::string;
using tensorflow::TF_StatusPtr;

namespace tensorflow {
namespace {

// The tests are parameterized on:
// - a string representing the tracing implementation: "mlir" or "graphdef".
// - a boolean that when true enables TFRT as the execution engine.
class UnifiedCAPI