Low cpu usage in 4*4 case for Fermi Hubbard Model

[hz-xiaxz]

Thanks for creating this convinient package!
I've tested the example script on my machine for the $3 \times 4$ case and it runs pretty well. (the doED is a simple function that wraps up the whole script) Also, the multi-core parallelling works well, 59% of my 96 cpus are used to run this work.

However, when I simply change the script from $3 \times 4$ to $4 \times 4$ case, the running process uses only 1 cpu, though the computation time seems reasonably aligning with scaling.

I wonder if this is an issue or a feature. And I want to know why you choose to demonstrate $3 \times 4$ case in the sample script? Thank you!

[waltergu]

This package does not implement manual multi-core parallelization.

The computation of the eigen energies of the 4×4 Fermi-Hubbard model follows the usual two-step procedure:

1. Generate the sparse matrix representation of the Hamiltonian,
2. Compute the smallest eigenvalues of this sparse matrix.

Step 1, as currently implemented, does not support multi-core parallelism. Step 2 may leverage multi-core capabilities depending on the external eigensolver used, which is currently KrylovKit.eigsolve (v0.2), and previously Arpack.eigs(v0.1).

The demonstation of a 3x4 case is just for convenience because it is neither too large nor too small for my computer.

[hz-xiaxz]

thanks!