The core Qobj class in QuTiP specifically uses a custom scipy-derived sparse-matrix format for data storage, which allows simulation and optimisations on large open quantum systems but causes significant memory and computational overhead on smaller-dimensioned systems, and the 32-bit index size can prevent even extremely sparse systems of high numbers of qubits from being representable. This project will decouple the low-level data manipulation procedures from the algebraic manipulations performed within the rest of QuTiP, encapsulating the low-level data manipulation into a higher-level interface, which will allow multiple storage formats to be used at the appropriate situations transparently. Overall, QuTiP will be able to spread its impressive performance to all problem-size domains without compromising on future maintainability or extensibility as demands for numerical quantum simulations increase.

Organization

Student

Jake Lishman

Mentors

  • Nathan Shammah
  • Eric
  • Alex Pitchford
close

2020