Mojo module
reducescatter
Multi-GPU reducescatter implementation for distributed tensor reduction across GPUs.
comptime values
elementwise_epilogue_type
comptime elementwise_epilogue_type = def[dtype: DType, width: Int, *, alignment: Int, ?, .element_types`0x3: KGENParamList[CoordLike]](Coord[element_types], SIMD[dtype, width]) capturing -> None
pdl_level
comptime pdl_level = PDLLevel(0)
Structs
-
ReduceScatterConfig: Configuration for axis-aware reduce-scatter partitioning.
Functions
-
reducescatter: Per-device reducescatter operation with axis-aware scatter.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!