Struct IdentityComm 

pub struct IdentityComm;

A single-process “communicator” that passes tensors through unchanged.

Useful for testing parallel code on a single device.

This reduces nothing, and that is correct. With world_size == 1 there is no peer to combine with, so the identity is the reduction. It is not a working multi-device collective and must not be read as one — for that, use DeviceGroup, which holds every shard and folds them in-graph.

Trait Implementations

impl Communicator for IdentityComm

fn info(&self) -> CommInfo

Communicator info (rank, world size).

fn all_reduce(&self, tensor: &Tensor, _op: ReduceOp) -> Result<Tensor>

All-reduce: compute element-wise reduction across all ranks, distributing the result to every rank.

fn all_gather(&self, tensor: &Tensor, _dim: usize) -> Result<Tensor>

All-gather: concatenate tensors from all ranks along dim.

fn reduce_scatter( &self, tensor: &Tensor, _op: ReduceOp, _dim: usize, ) -> Result<Tensor>

Reduce-scatter: reduce across ranks then scatter equal chunks to each rank.

fn broadcast(&self, tensor: &Tensor, _root: usize) -> Result<Tensor>

Broadcast tensor from root to all ranks.

fn barrier(&self) -> Result<()>

Barrier: block until all ranks reach this point.