Denis Alevi

PR adding more Windows support: https://github.com/brian-team/brian2/pull/1135

Definitely take into account synapse sorting during synapse generation. See my last comment in #32

The test in #190 uses 30GB, too. It only has 10 neurons and 1 Synapses object but multiple run calls.

Where possible, make the Brian2 tests `standalone-compatible` instead of `cpp_standalon`, `standalone-only`. I think for some this should be trivial and would avoid us having to update the `cuda-standalone` only tests...

Same happens for [test_spikegenerator_change_spikes](https://github.com/brian-team/brian2/blob/master/brian2/tests/test_spikegenerator.py#L165-L187), even if I add a different spikemonitor for each run. The test only passes if I also remove the previous spikemonitor from the network. before running...

Alright, I think I got it. There are three hardware limits for the execution of blocks on SMs: 1. A maximal number of threads, for the A100 (and I think...

Uhh. Looks like I'm not being notified of new issues 🙈. Changed the notification settings, sorry for the late response. As @mstimberg mentioned, I haven't worked on Windows support yet...

Fixed in 2c7e4da46a47e9722c4f743b4702ba19eac96abc, will be merged into master with PR #217 .

Copied relevant part for `source` mode optimization from [my comment](https://github.com/brian-team/brian2cuda/issues/12#issuecomment-280185166) in #12: **No serializing needed for `no_or_const_delay_mode` when presynaptic variables are modified** In `no_or_const_delay_mode` we don't need to serialize when...

Came across an [example using STDP with neuronal trace](https://github.com/brian-team/brian2/blob/320a5f6b36e2aa5dfe7e6b22bd46c3652ad9cc99/brian2/tests/test_synapses.py#L1793-L1819), which results in `source` mode pathways. So this can be relevant. @moritzaugustin