Functions
ir_tarval *	computed_value (const ir_node *n)
	If the expression referenced can be evaluated statically computed_value returns a tarval representing the result, tarval_bad if operation will throw an exception/produce undefined behaviour, otherwise tarval_unknown. More...

ir_node *	optimize_in_place (ir_node *n)
	Applies all optimizations to n that are expressible as a pattern in Firm, i.e., they need not a walk of the graph. More...

int	ir_is_negated_value (const ir_node a, const ir_node b)
	checks whether 1 value is the negated other value More...

ir_relation	ir_get_possible_cmp_relations (const ir_node left, const ir_node right)
	(conservatively) approximates all possible relations when comparing the value `left` and `right` More...

void	ir_allow_imprecise_float_transforms (int enable)
	enable/disable imprecise floatingpoint optimizations. More...

int	ir_imprecise_float_transforms_allowed (void)
	return 1 if imprecise float transformations are allowed. More...

Detailed Description

Function Documentation

◆ computed_value()

ir_tarval* computed_value ( const ir_node * n )

If the expression referenced can be evaluated statically computed_value returns a tarval representing the result, tarval_bad if operation will throw an exception/produce undefined behaviour, otherwise tarval_unknown.

◆ ir_allow_imprecise_float_transforms()

void ir_allow_imprecise_float_transforms ( int enable )

enable/disable imprecise floatingpoint optimizations.

These include rules like (a - x) + x = a, or a-0=a. They are wrong in several corner cases but may still be fine for some applications.

◆ ir_get_possible_cmp_relations()

ir_relation ir_get_possible_cmp_relations	(	const ir_node *	left,
		const ir_node *	right
	)

(conservatively) approximates all possible relations when comparing the value left and right

◆ ir_imprecise_float_transforms_allowed()

int ir_imprecise_float_transforms_allowed ( void )

return 1 if imprecise float transformations are allowed.

◆ ir_is_negated_value()

int ir_is_negated_value	(	const ir_node *	a,
		const ir_node *	b
	)

checks whether 1 value is the negated other value

◆ optimize_in_place()

ir_node* optimize_in_place ( ir_node * n )

Applies all optimizations to n that are expressible as a pattern in Firm, i.e., they need not a walk of the graph.

Returns a better node for n. Does not free n – other nodes could reference n.

An equivalent optimization is applied in the constructors defined in ircons.ch. There n is freed if a better node could be found.

Functions