Script.atomic.global_scatter_add

Script.atomic.global_scatter_add(dst, *, dim, indices, values, sem='relaxed', scope='gpu', output=None)

Scatter-add into a global tile along dim.

Same contract as shared_scatter_add() but the destination is a GlobalTensor and the default scope is 'gpu'.

Notes

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

• Hardware: Requires compute capability 7.0+ (sm_70).

• PTX: atom.{sem}.{scope}.global.add.s32 (or red.* when the output is unused).

Parameters:

• dst (GlobalTensor)

• dim (int)

• indices (RegisterTensor)

• values (RegisterTensor)

• sem (str)

• scope (str)

• output (RegisterTensor | None)

Return type:

RegisterTensor | None