// -----

func.func @pack_CPU(%arg0: tensor<100xf32>, %arg1: tensor<100xf32>) -> tensor<2x100xf32> attributes {tac.device = "CPU", tac.interface_name = "func_2"} {
  // CHECK: tac.cost = 1.000000e+02
  %0 = "tfl.pack"(%arg0, %arg1) {axis = 0 : i32, tac.device = "CPU", values_count = 2 : i32} : (tensor<100xf32>, tensor<100xf32>) -> tensor<2x100xf32>
  func.return %0 : tensor<2x100xf32>
}

func.func @func_0_GPU(%arg0: tensor<256x32x32x3xf32>, %arg1: tensor<256x32x32x3xf32>) -> tensor<256x32x32x3xf32> attributes {tac.device = "GPU", tac.interface_name = "func_0"} {
  %0 = "tfl.add"(%arg0, %arg1) {fused_activation_function = "RELU", tac.device = "GPU"} : (tensor<256x32x32x3xf32>, tensor<256x32x32x3xf32>) -> tensor<256x32x32x3xf32>
  func.return %0 : tensor<256x32x32x3xf32>
}

// -----

// CHECK: tac.cost = 1.000000e+03

			b := 3
			return func(y int) int { // ERROR "can inline main.func27.1" "can inline main.main.func27.func34"
				c := 5
				return func(z int) int { // ERROR "can inline main.func27.1.1" "can inline main.main.func27.func34.1" "can inline main.func27.main.func27.1.2" "can inline main.main.func27.main.main.func27.func34.func36"
					return a*x + b*y + c*z
				}(10) // ERROR "inlining call to main.func27.1.1"

// This pass is used to fold tfl.const ops to each subgraph (func::FuncOp):
// See the example below:
//
// In main:
// %0 = tfl.const...
// %1 = tfl.const...
// %2 = call func_1(..., %0,...)
// %3 = call func_2(..., %0, ..., %1...)
// ...
//
// Then those consts will be copied into each function and replace their usage.
// func_1:
//   %0 = tfl.const...
// func_2:
//   %0 = tfl.const...
//   %1 = tfl.const...

// Returns true iff the function's symbol is public.
bool IsPublicFuncOp(func::FuncOp func_op) {
  return SymbolTable::getSymbolVisibility(&*func_op) ==
         SymbolTable::Visibility::Public;
}

}  // namespace

func::FuncOp FindMainFuncOp(ModuleOp module_op) {
  if (const auto main_func_op = module_op.lookupSymbol<func::FuncOp>(
          kImportModelDefaultGraphFuncName);

void ConvertNMSPaddedFunc::RewriteFunc() {
  func_->setAttr(kTFImplements,
                 StringAttr::get(func_.getContext(), kTfNMSPadded));
  Value boxes = func_.getArgument(0);
  Value scores = func_.getArgument(1);
  Value max_output_size = func_.getArgument(2);
  Value iou_threshold = func_.getArgument(3);
  Value score_threshold = func_.getArgument(4);
  auto output_type0 = func_.getFunctionType().getResult(0);

LogicalResult ConvertDenseImageWarpFunc::RewriteFunc() {
  func_.eraseBody();
  func_.addEntryBlock();
  func_->setAttr(kTFImplements,
                 StringAttr::get(func_.getContext(), kImageWarping));

  OpBuilder builder(func_.getBody());
  auto op = builder.create<CustomOp>(func_.getLoc(),
                                     func_.getFunctionType().getResults(),
                                     func_.getArguments(), kImageWarping,

	"deferreturn": abi.FuncIDWrapper,
}

// Get the function ID for the named function in the named file.
// The function should be package-qualified.
func GetFuncID(name string, isWrapper bool) abi.FuncID {
	if isWrapper {
		return abi.FuncIDWrapper
	}
	if strings.HasPrefix(name, "runtime.") {
		if id, ok := funcIDs[name[len("runtime."):]]; ok {
			return id
		}
	}

	// the rest of the test, they will block.
	go func() { // func1
		select {}
	}()
	go func() { // func2
		var c chan int
		c <- 0
	}()
	go func() { // func3
		var c chan int
		<-c
	}()
	done1 := make(chan bool)
	go func() { // func4
		<-done1
	}()
	done2 := make(chan bool)
	go func() { // func5
		done2 <- true
	}()
	c1 := make(chan int)
	c2 := make(chan int)

class ConvertMaxUnpoolingFunc {
 public:
  explicit ConvertMaxUnpoolingFunc(func::FuncOp func, mlir::TF::FuncAttr attr)
      : func_(func), attr_(attr) {}

  LogicalResult RewriteFunc();

  LogicalResult VerifySignature();

 private:
  LogicalResult CreateCustomOptions(std::string& custom_option_buffer);

  func::FuncOp func_;
  mlir::TF::FuncAttr attr_;
};