Does "ensemble.compute_ensemble" differ from manually DFT calculations?

[biglinn]

Hi, @mesonepigreco:
The second tutorial gives a manual submission in the front part. This allows user to calculate energies and forces from generated configurations.
However, it also gives a automatical submission and calculates by the function ensemble.compute_ensemble.

I don't know if two ways are the same.

Moreover, there is a method relax = sscha.Relax.SSCHA to automatize all the procedures. User can perform NVT simulation with relax.relax and perform NPT simulation with relax.vc_relax.
This Relax module seemly also call compute_ensemble to perform calculations. How does it control NVT or NPT ensembles?

[mesonepigreco]

The NVT/NPT are the same relaxation at each population. The difference is that vc_relax uses the information on the stress tensor at the end of the minimization to change the primitive cell to reach the isotropic stress at the given target pressure, while relax just starts a new population with the same cell as the one used before.

The ensemble calculation is the same for both NVT and NPT, that is why it does not change.
The manual submission of the ensemble could be useful if you are using a calculator not implemented (e.g., VASP on a cluster). In this case you cannot use vc_relax or relax to do everything automatically, but you need to compute the ensemble and call one single step of vc_relax by choosing a starting pop >= than the maximum population (1 iteration) and telling the code to use the already computed ensemble (do not recompute it) using the restart_from_ens=True option.

relax.max_pop = 1
relax.vc_relax(target_press = 0, static_bulk_modulus = 50, start_pop = 1, restart_from_ens = True)

[biglinn]

So, vc_relax(restart_from_ens = True) skips calling compute_ensemble which perform calculations.

And relax directly starts a new population, does it mean that not using relax or vc_realx actually perform NVT as the same as using relax?