Hi @DavidKing2020. Did you double-check that you get the same number of spikes as with the rand()<rate*dt method? The code for all code objects except the input_neurons_spike_thresholder_codeobject should be exactly identical to the previous code, so I don’t have any explication why this one should take more time than before…
Hi @mstimberg , my memory actually slipped. The timing of s_input2e0_post_codeobject is actually the same with the new and original method as shown the following timing profile with the “rand()<rate*dt” method:.
input_neurons_spike_thresholder_codeobject 15.14 ms 42.87 %
s_input2e0_post_codeobject 10.61 ms 30.03 %
s_input2e0_pre_plastic_codeobject 1.68 ms 4.76 %
exci_neurons0_stateupdater_codeobject 1.42 ms 4.02 %
The Possion spike timing is indeed improved by ~x5. Thanks so much for all the help!
1 Like