kernel.async.gather.mask

← kernel dialect

Predicated form of kernel.async.gather. False predicates perform no source or destination access for the current invocation but still produce a completed token, preserving a uniform async group shape for tails and guarded tiles.

Operation contract

Property	Value
Semantic phase	—
Target contracts	kernel.async
Interfaces	CachePolicy
Memory effects	read source, write dest

Signature

Kind	Name	Type	Cardinality	Description
Operand	source	view	required	Per-invocation global-like source fragment.
Operand	dest	view	required	Subgroup-uniform workgroup destination tile with a leading subgroup-lane axis.
Operand	predicate	i1	required	Scalar predicate controlling this invocation's gather.
Result	token	any	required	Opaque async-copy token for the predicated gather.
Attribute	cache_scope	enum CacheScope	required	Required cache/coherency scope for the transfer.
Attribute	cache_temporal	enum CacheTemporal	required	Required temporal cache hint for the transfer.

Examples

%copy = kernel.async.gather.mask %src_lane to %lds_tile, %in_bounds {cache_scope = cu, cache_temporal = regular} : view<4xi8> to view<[%wave]x4xi8>, i1 -> kernel.async.token