genn
genn copied to clipboard
genn-team
general index into GeNN variables
When we created GeNN, there was a conscious decision to have variables with meaningful names, e.g. V, m, h, n, etc instead of v[0], v[1], v[2], ...
However, at times it would be really nice to just iterate over all variables, e.g. write them all to disk, or initialise them in a loop etc.
Should we just automatically generate a C array of pointers into the variables to help with this?
E.g.
SET_VARS({{"V", "scalar"},{"m", "scalar"},{"h","scalar"},{"NA","scalar"},{"NB","scalar"}})
model.addNeuronPopulation<model>("POP1", ...);
would currently create
scalar *VPOP1, *mPOP1, ...
The suggestion would be to also create
void *_vPOP1[];
where _vPOP1[0]= VPOP1, _vPOP1[1]= mPOP1 etc.
@neworderofjamie , does this make any sense at all?
Hmmm, it does make sense, but I'm not really sure if I think it's a good idea. Firstly, void pointers are a somewhat frowned upon/easy to shoot yourself in the foot with thing to present to the users. Secondly, I can't really think of a lot of use cases where models have enough state variables that you would want to access externally to make it worthwhile (and in those cases you can make void arrays of their state variables yourself unless you also have a profusion of different model types).
Personally I think the solution to this, like the issue with having large numbers of populations you want to do the same thing to (remember this horrorshow), is to use the 'shared library model' approach where you can get variable pointers by name - as this seems a generally useful thing and copy-and-pasted in several places now, I was thinking of making this 'official' and including a user-friendly version with the next GeNN
we can continue discussion offline - I am not entirely sure I understand how the 'shared library model approach' relates to this ... but I don't think I am aware of the details of this approach at all ;-)