Skip to main content
Version: 1.0

TensorMapFloatOOBFill

struct TensorMapFloatOOBFill

Out-of-bounds fill mode for floating-point TMA operations.

Specifies how out-of-bounds memory accesses are handled for floating-point data types during TMA operations.

Implemented traits

AnyType, Copyable, ImplicitlyCopyable, ImplicitlyDestructible, Movable, RegisterPassable, TrivialRegisterPassable

comptime members

NAN_REQUEST_ZERO_FMA

comptime NAN_REQUEST_ZERO_FMA = TensorMapFloatOOBFill(Int32(1))

Fill out-of-bounds values with NaN, request zero for FMA operations.

NONE

comptime NONE = TensorMapFloatOOBFill(Int32(0))

No special out-of-bounds handling.