Support (/ Fix) code generation with Artificial cells for NEURON simulator

[pramodk]

Overview of the issue

When testing the ringtest model with NMODL as transpiler for NEURON, I see that the spikes are not generated.

Expected result/behavior

Spike results between nmodl and nocmodl transpiler should be the same

Setup

• NEURON Version: master
• NMODL Version: master
• Installation method: typical CMake

cmake .. -DNRN_ENABLE_INTERVIEWS=OFF -DNRN_ENABLE_MPI=OFF -DNRN_ENABLE_RX3D=OFF -DNRN_ENABLE_CORENEURON=ON -DCMAKE_INSTALL_PREFIX=$(pwd)/install

Minimal working example

Model Repisotiry

git clone https://github.com/neuronsimulator/ringtest.git
cd ringtest

I applied below patch so that it's easy to test nocmodl and nmodl:

$ git diff
diff --git a/ring.py b/ring.py
index f2071ff..6301f10 100644
--- a/ring.py
+++ b/ring.py
@@ -102,8 +102,11 @@ class Ring(object):

         if not settings.pc.gid_exists(self.gidstart):
             return
-
-        self.stim = h.NetStim()
+        import os
+        if os.getenv("NETSTIM_NMODL") == '1':
+            self.stim = h.NetStimNMODL()
+        else:
+            self.stim = h.NetStim()

In order to test like the above, I created a copy of netstim.mod:

cp nrn/src/nrnoc/netstim.mod mod/netstim_nmodl.mod
sed -i '' 's/NetStim/NetStimNMODL/g' mod/netstim_nmodl.mod

Build mod files:

<nrn-install>/bin/nrnivmodl -nmodl <nrn-install>/bin/nmodl mod/

Check the spikes between nocmodl and nmodl:

$ ./arm64/special -python ringtest.py
NEURON -- VERSION 9.0a-299-g314d42222+ pramodk/docs-enh (314d42222+) 2024-07-16
...
0.000982046s wrote 688 spikes

vs

$ NETSTIM_NMODL=1 ./arm64/special -python ringtest.py
...
0.000211954s wrote 0 spikes

i.e. 688 spikes ve 0

Preliminary Investigation

Yesterday I was quickly looking through the generated code and I noticed the following. Maybe separate issues for each of these might be better (?):

• [x] RANDOM construct is not handled - e.g. stream allocation in nrn_alloc() is missing
• [ ] VERBATIM block contents is not printed (not the main cause)
• [x] ARTIFICIAL_CELL should have a call to artcell_net_send() instead of net_send()
• [x] mechname_reg() missing call to add_nrn_has_net_event(mech_type)

(I might be missing a few other aspects, which I will add here if I see/remember but you can easily see these as stimulus is not working)