vector.atomic.cmpxchg

← vector dialect

Atomic compare-exchange at per-lane signed element offsets. Each lane compares the bits at origin + offsets[lane] with expected[lane], writes replacement[lane] on success, and returns the old memory value. Comparison is bitwise for every accepted element type.

Operation contract

Property	Value
Semantic phase	—
Interfaces	CachePolicy, MemoryAccess
Memory effects	readwrite view

Signature

Kind	Name	Type	Cardinality	Description
Operand	expected	vector	required	Fixed-width byte-addressable expected value for each lane.
Operand	replacement	vector	required	Matching replacement value written for each successful lane.
Operand	view	view	required	Typed destination view.
Operand	offsets	vector	required	Per-lane signed element offsets from the logical origin.
Operand	indices	index	variadic	Dynamic logical origin indices.
Result	old	vector	required	Old memory values read by the atomic operations.
Attribute	success_ordering	enum AtomicOrdering	required	Memory ordering used when a lane compare-exchange succeeds.
Attribute	failure_ordering	enum AtomicOrdering	required	Memory ordering used when a lane compare-exchange fails.
Attribute	scope	enum AtomicScope	required	Required atomic synchronization scope.
Attribute	cache_scope	enum CacheScope	optional	Optional cache/coherency scope required by target lowering.
Attribute	cache_temporal	enum CacheTemporal	optional	Optional temporal cache policy required by target lowering.
Attribute	static_indices	i64_array	required	Static logical origin indices with INT64_MIN sentinels for dynamics.

Verification constraints

• HasBytePatternElement(expected)
• HasIndexOrNonI1IntegerElement(offsets)
• SameElementType(expected, replacement, view, old)
• SameShape(offsets, expected, replacement, old)
• SameType(expected, replacement, old)

Examples

%old = vector.atomic.cmpxchg %expected, %replacement, %view[%row, %col][%offsets] {success_ordering = acq_rel, failure_ordering = acquire, scope = workgroup} : vector<4xi32>, view<[%m]x[%n]xi32, %layout>, vector<4xindex> -> vector<4xi32>