Towards a Tensor Network ecosystem for large-scale simulations

Sergio Sánchez Ramírez (Barcelona Supercomputing Center)

The realization of Quantum Advantage (aka Quantum Supremacy, Beyond Classical) experiments these last years has put into question what kind of quantum systems can be efficiently simulated. Tensor Networks have been a successful contender for disproving quantum advantage claims and pushing the quantum vs classical frontier.

Multiple software libraries have been developed to ease the design of TN simulations, yet trying to scale simulations in supercomputers still requires the user to deal with internals to maximize performance. Internal hacking suggests incorrect abstractions of the TN data structures and algorithms.

In this talk, I will discuss BSC-Quantic’s effort on correcting design choices on TN software, writing composable software and leveraging TN community cooperation by means of modular packages. Some of the packages that will be introduced are:

  • Tenet.jl, a composable Tensor Network library that allows for tunable executions. Its design has been carefully crafted to provide composability, flexibility and performance.
  • EinExprs.jl, a contraction path search library that offers state-of-art heuristics, visualization utilities and optimizers.