std::string flops_str;
    std::string mac_str;
    if (ops_count < 10000) {
      flops_str = absl::StrFormat("%ld ", ops_count);
      mac_str = absl::StrFormat("%ld ", ops_count / 2);
    } else if (ops_count < billion) {
      flops_str =
          absl::StrFormat("%.3f M ", static_cast<double>(ops_count) / million);
      mac_str = absl::StrFormat("%.3f M ",

}  // namespace
}  // namespace mlir

int main(int argc, char** argv) {
  llvm::InitLLVM y(argc, argv);
  llvm::cl::ParseCommandLineOptions(argc, argv);
  auto file_or_err = llvm::MemoryBuffer::getFileOrSTDIN(inputFileName.c_str());
  if (std::error_code error = file_or_err.getError()) {
    llvm::errs() << argv[0] << ": could not open input file '" << inputFileName
                 << "': " << error.message() << "\n";
    return 1;
  }

  std::string mat_mul_method =
      absl::StrFormat(kMatMulModuleStr, test_case.mat_mul_method);
  TF_ASSERT_OK_AND_ASSIGN(
      XlaCompiler::CompilationResult result,
      CompileMlirModule(
          mat_mul_method.c_str(),
          ConfigProto::Experimental::MLIR_BRIDGE_ROLLOUT_UNSPECIFIED));
}

INSTANTIATE_TEST_SUITE_P(
    BatchMatMulTest, BatchMatMulTest,
    ::testing::ValuesIn<MatMulTestCase>({
        {"BatchMatMul"},

				req.AddCookie(c)
			}

			got := req.CookiesNamed(tt.filter)

			if !reflect.DeepEqual(got, tt.want) {
				asStr := func(v any) string {
					blob, _ := json.MarshalIndent(v, "", "  ")
					return string(blob)
				}
				t.Fatalf("Result mismatch\n\tGot: %s\n\tWant: %s", asStr(got), asStr(tt.want))
			}
		})
	}
}

const (
	fileaContents = "This is a test file."

  return FlatBufferModel::BuildFromFile(model_path.c_str());
}

std::unique_ptr<FlatBufferModel> ReadSharedWeightsTestModel() {
  auto model_path = tensorflow::io::JoinPath(*g_test_model_dir,
                                             internal::kModelWithSharedWeights);
  return FlatBufferModel::BuildFromFile(model_path.c_str());
}

std::unique_ptr<FlatBufferModel> ReadGatherTestModel() {

	// ABIs.
	ABIInternal

	ABICount
)

// ParseABI converts from a string representation in 'abistr' to the
// corresponding ABI value. Second return value is TRUE if the
// abi string is recognized, FALSE otherwise.
func ParseABI(abistr string) (ABI, bool) {
	switch abistr {
	default:
		return ABI0, false
	case "ABI0":
		return ABI0, true
	case "ABIInternal":
		return ABIInternal, true

}  // namespace mlir

int main(int argc, char** argv) {
  llvm::InitLLVM y(argc, argv);
  llvm::cl::ParseCommandLineOptions(argc, argv);
  auto file_or_err = llvm::MemoryBuffer::getFileOrSTDIN(inputFileName.c_str());
  if (std::error_code error = file_or_err.getError()) {
    llvm::errs() << argv[0] << ": could not open input file '" << inputFileName
                 << "': " << error.message() << "\n";
    return 1;
  }

          return absl::OkStatus();
        });

    // Register the corresponding fake op kernel.
    const char* node_name = opdef.name().c_str();
    const char* op_name = opdef.name().c_str();
    const char* device_name = "CPU";
    static auto fake_compute_func = [](void* kernel, TF_OpKernelContext* ctx) {
    };

    TF_KernelBuilder* builder =

	PDBStr                            = "PodDisruptionBudget"
	ReplicaSetStr                     = "ReplicaSet"
	RoleStr                           = "Role"
	RoleBindingStr                    = "RoleBinding"
	SAStr                             = "ServiceAccount"
	ServiceStr                        = "Service"
	SecretStr                         = "Secret"
	StatefulSetStr                    = "StatefulSet"

    node_name = std::string(attr.getValue());
  }
  absl::StatusOr<std::unique_ptr<tensorflow::NodeDef>> node_def =
      tensorflow::ConvertTFDialectOpToNodeDef(
          inst, node_name.c_str(), /*ignore_unregistered_attrs=*/true);
  RETURN_FAILURE_IF_ERROR(node_def.status());

  const FallbackState& fallback_state = GetDefaultFallbackState();