Enum BalancingScheme 

pub enum BalancingScheme {
    BestWorst,
    Tensor,
    Tensors,
    AlternatingTensors,
    IntermediateTensors {
        height_limit: Option<usize>,
    },
    AlternatingIntermediateTensors {
        height_limit: Option<usize>,
    },
    AlternatingTreeTensors {
        height_limit: usize,
    },
}

The scheme used for greedy balancing of partitions.

Variants

BestWorst

Moves a tensor from the slowest subtree to the fastest subtree each time.

Tensor

Identifies the tensor in the slowest subtree and passes it to the subtree with largest memory reduction.

Tensors

Identifies the tensor in the slowest subtree and passes it to the subtree with largest memory reduction. Then identifies the tensor with the largest memory reduction when passed to the fastest subtree. Both slowest and fastest subtrees are updated.

AlternatingTensors

Identifies the tensor in the slowest subtree and passes it to the subtree with largest memory reduction for odd iterations or the tensor with the largest memory reduction when passed to the fastest subtree for even iterations.

IntermediateTensors

Identifies the intermediate tensor in the slowest subtree and passes it to the subtree with largest memory reduction. Then identifies the intermediate tensor with the largest memory reduction when passed to the fastest subtree. Both slowest and fastest subtrees are updated.

Fields

height_limit: Option<usize>

The height up the contraction tree we look when passing intermediate tensors between partitions. A value of Some(1) allows intermediate tensors that are a product of at most 1 contraction process. Using the value of Some(0) is then equivalent to the Tensors method. Setting it to None imposes no height limit.

AlternatingIntermediateTensors

Identifies the intermediate tensor in the slowest subtree and passes it to the subtree with largest memory reduction for odd iterations. Identifies the intermediate tensor with the largest memory reduction when passed to the fastest subtree for odd iterations.

Fields

height_limit: Option<usize>

The height up the contraction tree we look when passing intermediate tensors between partitions. A value of Some(1) allows intermediate tensors that are a product of at most 1 contraction process. Using the value of Some(0) is then equivalent to the Tensors method. Setting it to None imposes no height limit.

AlternatingTreeTensors

Identifies the intermediate tensor in the slowest subtree and passes it to the subtree with largest memory reduction for odd iterations. Identifies the intermediate tensor with the largest memory reduction when passed to the fastest subtree for odd iterations.

Fields

height_limit: usize

The height up the contraction tree we look when passing intermediate tensors between partitions. A value of 1 allows intermediate tensors that are a product of at most 1 contraction process. Using the value of 0 is then equivalent to the Tensors method.

Trait Implementations

impl Clone for BalancingScheme

fn clone(&self) -> BalancingScheme

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for BalancingScheme

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Copy for BalancingScheme