Skip to content

Tucker and tuckerproduct Performance Optimization #9

@njericha

Description

@njericha

We should be able to multiple matrix factors with the core of a Tucker product in an optimal order based on the size of the matrix factors. For example,

G ×₆ M[6] ×₅ M[5] ×₄ M[4] ×₃ M[3] ×₂ M[2] ×₁ M[1]

can be faster than

 G ×₁ M[1] ×₂ M[2] ×₃ M[3] ×₄ M[4] ×₅ M[5] ×₆ M[6]

if the matrices go up in size.

But _fulltuckerproduct uses less memory so it may not be worth it if we expect this will cause more garbage collection.

See file tensorproducts.jl.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions