Cell splitting in neural networks extends strong scaling (Hines et al. 2008)

 Download zip file   Auto-launch 
Help downloading and running models
Neuron tree topology equations can be split into two subtrees and solved on different processors with no change in accuracy, stability, or computational effort; communication costs involve only sending and receiving two double precision values by each subtree at each time step. Application of the cell splitting method to two published network models exhibits good runtime scaling on twice as many processors as could be effectively used with whole-cell balancing.
1 . Hines ML, Eichner H, Schürmann F (2008) Neuron splitting in compute-bound parallel network simulations enables runtime scaling with twice as many processors. J Comput Neurosci 25:203-10 [PubMed]
Citations  Citation Browser
Model Information (Click on a link to find other models with that property)
Model Type: Realistic Network;
Brain Region(s)/Organism: Generic;
Cell Type(s):
Gap Junctions:
Simulation Environment: NEURON;
Model Concept(s): Methods;
Implementer(s): Hines, Michael [Michael.Hines at Yale.edu];
durand.hoc *
groucho_gapbld.hoc *
groucho_gapbld_mix.hoc *
serial_or_par_wrapper.hoc *
synaptic_compmap_construct.hoc *
synaptic_map_construct.hoc *
objref compmap, allow, x

obfunc synaptic_compmap_construct () { local nrow, ncol  localobj f, s, tmpmap
Parameter, Description:
$1 thisno, maybe this double will be replaced in NEURON?
$2 num_postsynaptic_cells,  another double
// returned compmap(i,j), Matrix object=compartment #on postsyn cell j for ith presyn input
$3 num_presyninputs_perpostsyn_cell, a double 
$4 num_allowcomp, another double
$o5 allow, a Vector object of allowed postsyn compartments
$6 display, another double

c Construct a map of compartments at connections of one presynaptic
c cell to type to a postsynaptic cell type.
c compmap (i,j) = compartment number on postsynaptic cell j of its
c  i'th presynaptic input.
c display is an integer flag.  If display = 1, print compmap

        INTEGER thisno,
     &   num_postsynaptic_cells,
     &   num_presyninputs_perpostsyn_cell,
     &   compmap (num_presyninputs_perpostsyn_cell, 
     &                  num_postsynaptic_cells),
     &   num_allowcomp, allow(num_allowcomp)
c num_allowcomp = number of different allowed compartments
c allow = list of allowed compartments
        INTEGER i,j,k,l,m,n,o,p
        INTEGER display

        double precision seed, x(1)
//	print "arrived"
//	objref seed
	seed = new Vector()

        num_postsynaptic_cells = $2
        num_presyninputs_perpostsyn_cell = $3
	num_allowcomp = $4
	objref allow
	allow = $o5
	display = $6

	objref compmap
	compmap = new Matrix(num_presyninputs_perpostsyn_cell+1, num_postsynaptic_cells+1)
  if (!use_p2c_net_connections) {
//            map = 0
            k = 1
// print "num_postsynaptic_cells, num_presyninputs_perpostsyn_cell = ",num_postsynaptic_cells, num_presyninputs_perpostsyn_cell
// print "matrix size = ",compmap.nrow(),compmap.ncol()

        for ii = 1, num_postsynaptic_cells {
        for jj = 1, num_presyninputs_perpostsyn_cell {
            x = durand (seed, k, x)
// c This defines a compartment     
           LL = int ( x.x[0] * (num_allowcomp) ) + 1
//	 print "jj,ii: ",jj,ii, " LL=",LL
        if (LL > num_allowcomp) {
		print " unnexpected boundary issue in synaptic_compmap_construct()"
		LL = num_allowcomp
// print allow.x(L)
           compmap.x[jj][ii] = allow.x[LL]


	thisno = $1
// c Possibly print out map when done.
       if ((display == 1) && (thisno == 0)) {
        for i = 1, num_postsynaptic_cells {
         printf("%6d %6d %6d\n", compmap.x(1,i), compmap.x(2,i), \

	// read from file created by port2colossus
	s = new String()
	sprint(s.s, "../../p2c/compmap/%s.dat", $s7)
//printf("%s %d %d\n", s.s, nrow, ncol)
	f = new File()
	tmpmap = new Matrix(ncol, nrow) // need to transpose
	tmpmap.scanf(f, ncol, nrow)
	tmpmap = tmpmap.transpose
	tmpmap.bcopy(0,0,nrow, ncol, 1, 1, compmap)
       return compmap