Using Multiple nodes for TEBD.

Hi!

Is there any option to use multiple nodes for TEBD calculation (in Julia version)? I am not using the in-built TEBD function. I have an MPS & I multiply it with Trotter gates to do the time evolution.Now,gates are applied on even & odd sites alternatively.So, gates in each layer commute among themselves. I guess it is then possible to push the portions of a large MPS into different cluster nodes & apply the gates in a single layer in parallel instead of applying the gates in sequence from an array.

Unfortunately, I can’t find out a simple way to do that. This is not multi threading (which is already being done by Intel MKL/BLAS). I need to access multiple nodes . I looked at https://github.com/ITensor/ITensorParallel.jl & the tutorials. It seems like there is an implementation there but I couldn’t understand how to tune it for my purpose.

Thank you in advance for your kind support. I really appreciate the awesome work being done by the ITensor maintainers to provide us this awesome library.

Unfortunately we don’t have a public version available for that.

Right now, ITensorParallel.jl is focused on parallelizing over sums of Hamiltonians, while what you are asking for is real space parallelization (more akin to [1301.3494] Real-Space Parallel Density Matrix Renormalization Group, though with TEBD it is even simpler). It is something that we are definitely hoping to do but haven’t had time to implement a public version yet. One of our postdocs Benedikt Kloss who has been contributing to ITensorParallel.jl has a private version of real space parallelized TEBD which maybe he can share but I don’t know what the status of that is.