"Adding" solutions is slow (casadi.horzcat)

[rtimms]

"Adding" solutions can take up a large portion of the overall solution time. The time is spend calling casadi.horzcat.

E.g. in the below example casadi.horzcat takes about 9% of the overall time.

import pybamm
pybamm.set_logging_level("INFO")
options = {
    "sei": "solvent-diffusion limited",
    "sei film resistance": "distributed",
    "sei porosity change": "true",
}
model = pybamm.lithium_ion.DFN(options=options)
params = pybamm.ParameterValues(chemistry=pybamm.parameter_sets.Chen2020)
params.update(
    {
        "Reference temperature [K]": 278.15,
        "Ambient temperature [K]": 278.15,
        "Initial temperature [K]": 278.15,
        "Upper voltage cut-off [V]": 4.21,
    }
)
exp = pybamm.Experiment(
    ["Hold at 4.2V until C/100", "Rest for 4 hours (3 minute period)"]
    + [
        "Discharge at 1C until 2.5V",
        "Charge at 0.3C until 4.2V (3 minute period)",
        "Hold at 4.2V until C/100 (3 minute period)",
    ]
    * 50
)
sim = pybamm.Simulation(model=model, parameter_values=params, experiment=exp)
sim.solve()

[valentinsulzer]

Should be fixed, together with #1276 , by reformatting Solution to create a processed variable from its sub-solutions, without first creating solution.y

[valentinsulzer]

This was a great find @rtimms - new results look promising!

On develop, this experiment took 7:28
On new branch, this experiment took 5:35