sizeof2t Class Reference

sizeof(T) expression. More...

#include <irep2_expr.h>

Inheritance diagram for sizeof2t:

Public Member Functions
	sizeof2t (const type2tc &type, const expr2tc &val, const type2tc &sizeof_t)
	sizeof2t (const sizeof2t &ref)=default
expr2tc	do_simplify () const override
	expr-specific simplification methods.
Public Member Functions inherited from expr2t
virtual	~expr2t ()=default
expr2tc	clone () const
expr2tc	with_type (const type2tc &new_type) const
	Build a fresh expression of the same kind and field values as this one, but with a different type.
bool	operator== (const expr2t &ref) const
bool	operator< (const expr2t &ref) const
bool	operator!= (const expr2t &ref) const
std::string	pretty (unsigned int indent=0) const
	Produce textual representation of this expr.
DUMP_METHOD void	dump () const
	Write textual representation of this object to stdout.
size_t	crc () const
	Calculate a hash/digest of the current expr.
bool	cmp (const expr2t &ref) const
	Structural comparison.
int	lt (const expr2t &ref) const
	Trinary structural ordering.
list_of_memberst	tostring (unsigned int indent) const
	Convert fields of subclasses to a string representation.
const expr2tc *	get_sub_expr (size_t idx) const
	Fetch a sub-operand.
size_t	get_num_sub_exprs () const
	Count the number of sub-exprs there are.
expr2tc	simplify () const
	Simplify an expression.
expr2tc	simplify (bool suppress_reassoc) const
	Simplify with reassociation suppression.
template<typename T>
void	foreach_operand (T &&t) const
	Indirect, abstract operand iteration.
template<typename T>
void	Foreach_operand (T &&t)
Public Member Functions inherited from irep2t
	irep2t () noexcept
	irep2t (const irep2t &) noexcept
irep2t &	operator= (const irep2t &)=delete
virtual	~irep2t ()=default
void	mark_writer () const noexcept
void	clear_writer () const noexcept

Public Attributes
expr2tc	value
type2tc	sizeof_type
Public Attributes inherited from expr2t
const expr_ids	expr_id
	Instance of expr_ids recording tihs exprs type.
const type2tc	type
	Type of this expr.
std::atomic< size_t >	crc_val
Public Attributes inherited from irep2t
std::atomic< unsigned int >	refcount
std::atomic< std::uintptr_t >	writer_thread

Static Public Attributes
static constexpr auto	fields
static std::string	field_names [esbmct::num_type_fields]

Additional Inherited Members
Public Types inherited from expr2t
enum	expr_ids { end_expr_id }
	Enumeration identifying each sort of expr. More...
typedef esbmct::function_ref< void(const expr2tc &)>	const_op_delegate
typedef esbmct::function_ref< void(expr2tc &)>	op_delegate
typedef expr2tc	container_type
typedef expr2t	base_type
Static Public Member Functions inherited from irep2t
static std::uintptr_t	current_thread_tag () noexcept
Protected Member Functions inherited from expr2t
	expr2t (const type2tc &type, expr_ids id)
	Primary constructor.
	expr2t (const expr2t &ref)
	Copy constructor.
void	foreach_operand_impl_const (const_op_delegate &expr) const
void	foreach_operand_impl (op_delegate &expr)

Detailed Description

sizeof(T) expression.

Replaces the legacy sizeof-type irep attribute side channel (esbmc/esbmc#5337) with two first-class reflected fields:

• value: the byte size, computed eagerly by the frontend so it matches clang's authoritative sizeof (constants via Expr::EvaluateAsInt, VLAs via c_sizeof with a namespace). Keeping clang's value — rather than recomputing it from sizeof_type via type_byte_size — is what makes flexible-array-members, bitfields and VLAs agree with the source.
• sizeof_type: the measured type T, read by allocation typing (get_alloc_type), the scanf overflow check and the pretty-printer. do_simplify() unwraps to value; the SMT backend lowers the node to value directly so it is sound even under –no-simplify. Both fields ride in fields, so the node migrates / copies / hashes with no side channel.

Constructor & Destructor Documentation

sizeof2t() [1/2]

sizeof2t::sizeof2t ( const type2tc & type, const expr2tc & val, const type2tc & sizeof_t )

inline

sizeof2t() [2/2]

sizeof2t::sizeof2t ( const sizeof2t & ref )

default

Member Function Documentation

do_simplify()

expr2tc sizeof2t::do_simplify ( ) const

overridevirtual

expr-specific simplification methods.

By default, an expression can't be simplified, and this method returns a nil expression to show that. However if simplification is possible, the subclass overrides this and if it can simplify its operands, returns a new simplified expression. It should attempt to modify itself (it's const).

If simplification failed the first time around, the simplify method will simplify this expressions individual operands, and will then call an expr with the simplified operands to see if it's now become simplifiable. This call occurs whether or not any operands were actually simplified, see below.

The 'second' parameter can be used to avoid invoking expensive attempts to simplify an expression more than once - on the first call to do_simplify this parameter will be false, then on the second it's be true, allowing method implementation to save the expensive stuff until all of its operands have certainly been simplified.

Currently simplification does some things that it shouldn't: pointer arithmetic for example. I'm not sure where this can be relocated to though.

Returns

expr2tc A nil expression if no simplifcation could occur, or a new simplified object if it can.

Reimplemented from expr2t.

Member Data Documentation

field_names

std::string sizeof2t::field_names

static

Initial value:

=
  {"value", "sizeof_type", "", "", ""}

fields

auto sizeof2t::fields

staticconstexpr

Initial value:

=
    std::make_tuple(&expr2t::type, &sizeof2t::value, &sizeof2t::sizeof_type)

sizeof_type

type2tc sizeof2t::sizeof_type

value

expr2tc sizeof2t::value

---

The documentation for this class was generated from the following files:

• src/irep2/irep2_expr.h
• src/irep2/irep2_expr.cpp
• src/util/expr/expr_simplifier.cpp