Supported data types in GeNN models

[Jaime67]

Hello!

I'm working with GeNN and would like to know the data types available for defining an SNN model. Specifically, I’m referring to this data type definition:

model = GeNNModel("float", "mnist_mb")

In the documentation tutorials and user projects, float is commonly used. I've also tested double. I'm considering using half-precision float, but as far as I know, it's not supported. Are there any other data types (such as half-float or alternatives to float and double) currently supported in GeNN? Are there expectations to support new data types soon?

Thanks!

[neworderofjamie]

The only data types currently supported for setting at this level are "float" and "double" although you are also free to use integer types for individual neuron and synapse model state variables if you want to experiment with e.g. neuromorphic hardware friendly neuron models. We are working on support for fp16 and bfloat16 but I can't promise when that work will be done.