(target_op_names.empty() && !op->isRegistered())) {
      custom_ops.push_back(op);
    }
  });

  for (auto *op : custom_ops) {
    builder.setInsertionPoint(op);
    Location loc = op->getLoc();
    auto custom_op = builder.create<CustomTfOp>(loc, op->getResultTypes(),
                                                op->getOperands());
    Region region;
    Block *new_block = new Block;

    return failure();
  }

  results.clear();
  builder.setInsertionPointAfter(op);
  for (const auto& result_value : result_values) {
    results.push_back(builder.create<TF::ConstOp>(op->getLoc(), result_value));
  }
  return success();
}

bool IsOperationFoldable(Operation* op) {
  if (isa<TF::ConstOp>(op)) return true;

  if (op->getDialect()->getNamespace() != "tf" || !TF::CanBeFolded(op)) {

System.err.println 'this is stderr'

def slf4jLogger = org.slf4j.LoggerFactory.getLogger('some-logger')
slf4jLogger.warn('A warning using SLF4j')

def jclLogger = org.apache.commons.logging.LogFactory.getLog('some-logger')
jclLogger.warn('A warning using JCL')

def log4jLogger = org.apache.log4j.Logger.getLogger('some-logger')
log4jLogger.warn('A warning using Log4j')

  std::string device = "/job:localhost/replica:0/task:0/device:CPU:0";
  module->walk(
      [&](mlir::tf_device::ClusterOp descendant) { cluster = descendant; });
  mlir::OpBuilder builder(&context);
  auto loc = cluster->getLoc();

  // Wrap the cluster op into a Launch op
  auto launch_op = tensorflow::WrapOpInLaunch(&builder, loc, cluster, device);

  EXPECT_TRUE(llvm::isa<mlir::tf_device::LaunchOp>(launch_op));
  launch_op->erase();

    // This occurs when doing qi16 convolution, as bias is passed as a
    // non-quantized int64
    if (allTypesFp || allTypesQuantizedOrInt) return failure();

    Location loc = op->getLoc();
    SmallVector<Value> newOperands;
    newOperands.reserve(op->getNumOperands());
    for (auto operand : op->getOperands()) {
      if (QuantizedType::getQuantizedElementType(operand.getType())) {

                    "one use for all functions. Functions in the "
                    "recursive cycle are: ";
      llvm::interleaveComma(*scci, err, [&](CallGraphNode *node) {
        err << node->getCallableRegion()->getLoc();
      });
      return err;
    }
  }
  return success();
}

#define GEN_PASS_DEF_GUARANTEEALLFUNCSONEUSEPASS
#include "tensorflow/compiler/mlir/tensorflow/transforms/tf_passes.h.inc"

         static_cast<float>(max.getValueAsDouble())});
    ElementsAttr axis_stats;
    IntegerAttr axis;

    quantfork::StatisticsOp stats_op = rewriter.create<quantfork::StatisticsOp>(
        op->getLoc(), op.getInput(), layer_stats, axis_stats, axis);
    op.getOutput().replaceAllUsesWith(stats_op.getResult());
    return success();
  }
};

static PassRegistration<ConvertCustomAggregationOpToQuantStatsPass> pass;

  const auto composite_resource_users =
      GetResourceOpsUsingCompositeArgsInReplicate(replicate_op);
  for (auto resource_user : composite_resource_users) {
    WrapOpInLaunch(builder, resource_user->getLoc(), resource_user,
                   tensorflow::GetDeviceAliasForLogicalCore(0));
  }
}

void TPUColocateCompositeResourceOps::runOnOperation() {

  if (call_op_has_platform_index_arg) {
    arg_mapper.map(func_op.getArgument(0),
                   builder.create<mhlo::ConstantOp>(
                       func_op.getLoc(), builder.getI16IntegerAttr(0)));
  }
  for (auto [func_arg, operand] : llvm::zip_equal(
           func_op.getArguments().take_back(call_op.getNumOperands()),
           call_op.getOperands())) {

  if (result_tuple_ty.size() != 1) return;
  auto result_ty = mlir::dyn_cast<ShapedType>(result_tuple_ty.getType(0));

  func.eraseBody();
  func.addEntryBlock();
  ImplicitLocOpBuilder builder(func.getLoc(), func.getBody());
  llvm::SmallVector<int32_t> result_shape_i32;
  auto result_shape = result_ty.getShape();
  for (auto element : result_shape) {
    result_shape_i32.push_back(static_cast<int32_t>(element));
  }