triton.experimental.gluon.language.nvidia.ampere.async_copy.async_load

triton.experimental.gluon.language.nvidia.ampere.async_copy.async_load(smem, pointer, mask=None, cache_modifier=None, eviction_policy=None, volatile=False, cache_policy=None, _semantic=None)

Asynchronously load elements from global memory to shared memory.

Parameters:

  • smem (shared_memory_descriptor) – Destination shared memory descriptor.

  • pointer (tensor) – Source pointer tensor.

  • mask (tensor, optional) – Mask tensor for predicated loads. Defaults to None.

  • cache_modifier (str) – Cache modifier specifier. Defaults to None.

  • eviction_policy (str) – Eviction policy specifier. Defaults to None.

  • volatile (bool) – Whether the load is volatile. Defaults to False.

  • cache_policy (CachePolicy) – Cache policy. Cannot be combined with cache_modifier or eviction_policy. Defaults to None.