Translating network models to parallel hardware in NEURON (Hines and Carnevale 2008)

 Download zip file   Auto-launch 
Help downloading and running models
Shows how to move a working network model written in NEURON from a serial processor to a parallel machine in such a way that the final result will produce numerically identical results on either serial or parallel hardware.
1 . Hines ML, Carnevale NT (2008) Translating network models to parallel hardware in NEURON. J Neurosci Methods 169:425-55 [PubMed]
Model Information (Click on a link to find other models with that property)
Model Type: Realistic Network;
Brain Region(s)/Organism:
Cell Type(s):
Gap Junctions:
Simulation Environment: NEURON;
Model Concept(s): Simplified Models; Methods;
Implementer(s): Carnevale, Ted [Ted.Carnevale at]; Hines, Michael [Michael.Hines at];
{load_file("nrngui.hoc")}  // load the GUI and standard run libraries

objref pc
pc = new ParallelContext()

// Step 1: Define the cell classes


// Steps 2 and 3 are to create the cells and connect the cells

NCELL = 20  // total number of cells in the ring network
  // identical to total number of cells on all machines

objref cells, nclist  // cells will be a List that holds 
  // all instances of network cells that exist on this host
  // nclist will hold all NetCon instances that exist on this host
  // and connect network spike sources to targets on this host (nclist)

proc mkring() {
  mkcells($1)  // create the cells
  connectcells()  // connect them together

// creates the cells and appends them to a List called cells
// argument is the number of cells to be created
proc mkcells() {local i  localobj cell, nc, nil
  cells = new List()
  // each host gets every nhost'th cell,
  // starting from the id of the host
  // and continuing until all cells have been dealt out
  for (; i < $1; i += pc.nhost) {
    cell = new B_BallStick()
    pc.set_gid2node(i,  // associate gid i with this host
    nc = cell.connect2target(nil)  // attach spike detector to cell
    pc.cell(i, nc)  // associate gid i with spike detector

// connects the cells
// appends the NetCons to a List called nclist
proc connectcells() {local i, targid  localobj src, target, syn, nc
  nclist = new List()
  for i=0, NCELL - 1 {  // iterating over source gids
    targid = (i+1)%NCELL
    if (!pc.gid_exists(targid)) { continue }
    target = pc.gid2cell(targid)
    syn = target.synlist.object(0)  // the first object in synlist
        // is an ExpSyn with e = 0, therefore an excitatory synapse
    nc = pc.gid_connect(i, syn)
    nc.delay = 1
    nc.weight = 0.01

mkring(NCELL)  // go ahead and create the net!

// Instrumentation, i.e. stimulation and recording

// stim will be an artificial spiking cell that generates a "spike" event
// that is delivered to the first cell in the net by ncstim
// in order to initiate network spiking.
// We won't bother including this "external stimulus source" or its NetCon
// in the network's lists of cells or NetCons.
objref stim, ncstim
proc mkstim() {
  // exit if the first cell in the net does not exist on this host
  if (!pc.gid_exists(0)) { return }
  stim = new NetStim()
  stim.number = 1
  stim.start = 0
  ncstim = new NetCon(stim, pc.gid2cell(0).synlist.object(0))
  ncstim.delay = 0
  ncstim.weight = 0.01


objref tvec, idvec  // will be Vectors that record all spike times (tvec)
        // and the corresponding id numbers of the cells that spiked (idvec)
proc spikerecord() {local i  localobj nc, nil
  tvec = new Vector()
  idvec = new Vector()
  for i=0, cells.count-1 {
    nc = cells.object(i).connect2target(nil)
    nc.record(tvec, idvec, nc.srcgid)
    // the Vector will continue to record spike times
    // even after the NetCon has been destroyed


// Simulation control

tstop = 100

// Report simulation results

proc spikeout() { local i, rank
  pc.barrier()  // wait for all hosts to get to this point
  if ( printf("\ntime\t cell\n")  // print header once
  for rank=0, pc.nhost-1 {  // host 0 first, then 1, 2, etc.
    if ( {
      for i=0, tvec.size-1 {
        printf("%g\t %d\n", tvec.x[i], idvec.x[i])
    pc.barrier()  // wait for all hosts to get to this point



Loading data, please wait...