nest-gpu
nest-gpu copied to clipboard
nest
NEST GPU
Currently connections are first created in the RAM, then copied to GPU memory just before starting the simulation. The building phase could be much faster if connections were created in...
To be able to register NESTML compiled models at run time there is the need to dynamically load models to the C++ / CUDA level.
Implement possibility to choose between at least two possibilities for the number of bytes used for each connection (excluding structures for non-static connections): either 12 bytes or 16 bytes per...
The ext_neuron class is used both to represent the images of NEST (CPU) neurons, through a software interface, and the images remote source neurons communicating spikes to their images through...
Image nodes are just spike buffers that accumulate spikes from remote source nodes, to be delivered to local target nodes. Avoid calling any update functions other than the spike-buffer update...
Improve performance in the creation of remote connection structures by removing unnecessary steps in case source nodes in a Connect command are defined by a sequence, i.e. have contiguous indexes.