Script.tcgen05.copy

Script.tcgen05.copy

Script.tcgen05.copy(src, dst)[source]

Copy data from shared memory to tensor memory.

Asynchronously copies a 2D shared tensor into a 2D tensor memory tensor. Use tcgen05.commit to signal completion via an mbarrier.

Parameters:

  • src (SharedTensor) – The source shared tensor. Must be 2D.

  • dst (TMemoryTensor) – The destination tensor memory tensor. Must be 2D.

Return type:

None

Notes

  • Thread group: Must be executed by a warp-aligned thread group.

  • Hardware: Requires compute capability 10.0+ (sm_100).

  • PTX: tcgen05.cp

Contents