Mojo struct
TensorMapL2Promotion
struct TensorMapL2Promotion
L2 cache promotion hint for TMA tensor map descriptors.
Specifies how much data to promote into the L2 cache during TMA operations. Promoting data to L2 can improve performance when the same data will be accessed multiple times.
Implemented traits
AnyType,
Copyable,
ImplicitlyCopyable,
ImplicitlyDestructible,
Movable,
RegisterPassable,
TrivialRegisterPassable
comptime members
L2_128B
comptime L2_128B = TensorMapL2Promotion(SIMD(2))
Promote 128 bytes to L2 cache.
L2_256B
comptime L2_256B = TensorMapL2Promotion(SIMD(3))
Promote 256 bytes to L2 cache.
L2_64B
comptime L2_64B = TensorMapL2Promotion(SIMD(1))
Promote 64 bytes to L2 cache.
NONE
comptime NONE = TensorMapL2Promotion(SIMD(0))
No L2 promotion.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!