We are currently using 32bit integers for neuron and synapse IDs. Given the current minimal memory demand for a synapse and the maximal available GPU memory (12GB currently?), we should not be able to have more synapses than a 32bit integer can index. But this might change in the future with new GPUs etc. We need to make sure that this is reported correctly to the user. Currently I am just printing an error message when the number of synapses exceeds the unsigned integer range (see here). That message should reference for example this issue or another issue for adding 64bit integer support. And probably should also exit the simulation, currently it just prints a message.
We are currently using
32bitintegers for neuron and synapse IDs. Given the current minimal memory demand for a synapse and the maximal available GPU memory (12GBcurrently?), we should not be able to have more synapses than a32bitinteger can index. But this might change in the future with new GPUs etc. We need to make sure that this is reported correctly to the user. Currently I am just printing an error message when the number of synapses exceeds the unsigned integer range (see here). That message should reference for example this issue or another issue for adding64bitinteger support. And probably should also exit the simulation, currently it just prints a message.