MTLN is ~2x slower than holland

[lmdiazangulo]

[Alberto-o]

If the cause of the slowness is in the calls to ngspice, there are things that can be changed to have less calls:

Simulation with sources

Sources are defined as independent sources with a constant DC value. Each timestep there is a call to updateCircuitSources, which changes the value of the source using the ngsice command alter , passing a value interpolated from two arrays of time and value read from the excitation file

However, a PWL source can be defined if ngspice is compiled with xspice (which it is, by default). A model for the source is generated, and a A device source is defined from that model

* Define the model referencing your file
.model input_file_mod filesource(file="data.txt" timescale=1 amplscale=[1.0])

* Instantiate the source (n1, n2 are nodes)
A1 [%v(n1) %v(n2)] input_file_mod

In preprocess, where the terminations are defined, if there is a source it could defined this way. Then, the call to updateCircuitSources could be eliminated

Updating the current of the nodes

Instead of referring to each node current independently I think there is a way to refer to them all in a single call using them as a vector (see manual section 13.5.3). WIP