We need to implement a snapshot caching mechanism (maybe using asynchronous programming?) in order to be able to run MPS/DMRG simulations on large systems. Storing 10-100 MPS of length $\mathcal{O}(100)$ is too memory heavy.
The tricky part will be adjusting the RBasis and HamiltonianCache extension functions for this caching mechanism, i.e. performing contractions of two MPS or MPO applications and loading/writing MPS to disk simultaneously.
Of course this mechanism should eventually be snapshot type indepedent.
We need to implement a snapshot caching mechanism (maybe using asynchronous programming?) in order to be able to run MPS/DMRG simulations on large systems. Storing 10-100 MPS of length $\mathcal{O}(100)$ is too memory heavy.
The tricky part will be adjusting the
RBasisandHamiltonianCacheextension functions for this caching mechanism, i.e. performing contractions of two MPS or MPO applications and loading/writing MPS to disk simultaneously.Of course this mechanism should eventually be snapshot type indepedent.