Script.cluster.map_shared_addr

Script.cluster.map_shared_addr

Script.cluster.map_shared_addr(addr, target_rank)[source]

Map shared memory address(es) to the corresponding address(es) in another CTA’s shared memory.

This instruction uses the PTX mapa.shared::cluster instruction to translate shared memory addresses from the current CTA’s address space to another CTA’s address space within the same cluster.

Parameters:

  • addr (RegisterTensor) – A register tensor of dtype uint32 containing shared memory address(es) to map.

  • target_rank (Expr | int) – The rank of the target CTA in the cluster.

Returns:

A register tensor with the same shape and dtype as addr, containing the mapped addresses.

Return type:

RegisterTensor

Notes

  • Thread group: Can be executed by any sized thread group.

  • Hardware: Requires compute capability 9.0+ (sm_90).

  • PTX: mapa.shared::cluster

Contents