Corrected calcBunchofWn

Author: pavel-esir

.gitignore

@@ -14,3 +14,6 @@ py_hh_cpu.cpp
 !n_cycle.npy
 !m_cycle.npy
 !h_cycle.npy
+*.png
+*.out
+*.txt

hh_main.cpp

@@ -105,7 +105,10 @@ void freeMemory(){
 }
 
 void set_calc_params(unsigned int devIdx, unsigned int Tsim, unsigned int cutoff_ns_tm, unsigned int Nneur, unsigned int Ncon, unsigned int recInt, float h){
-    cudaSetDevice(devIdx);
+//     int deviceCount = 0;
+//     cudaGetDeviceCount(&deviceCount);
+//     printf("%u, device count %i\n", devIdx, deviceCount);
+    cudaSetDevice(devIdx % 3);
     if (firstRun){
         firstRun = false;
     } else {

scripts/calcBunchofWn.py

@@ -9,6 +9,8 @@
 import py_hh as phh
 import numpy as np
 from numpy import random
+import matplotlib
+matplotlib.use('Agg')
 import matplotlib.pylab as pl
 pl.ioff()
 from distribute_delays import getDelays
@@ -18,21 +20,39 @@
 rank = comm.Get_rank()
 size = comm.Get_size()
 
-w_n_range = np.linspace(1.3, 1.5, size, endpoint=False)
+#N = 60
+w_n_range = np.linspace(1.6, 2.0, size)
+
+#N = 70
+#w_n_range = np.linspace(1.6, 1.8, size)
+
+# N = 80
+#w_n_range = np.linspace(1.3, 1.5, size)
+
+# N = 90
+#w_n_range = np.linspace(1.2, 1.4, size)
+
+#w_n_range = np.linspace(1.1, 1.39, size)
+
+# Neur = 120
+#w_n_range = np.linspace(1.14, 1.16, size)
+
+# Nneur = 130
+#w_n_range = np.linspace(1.00, 1.12, size)
 
 random.seed(0)
 psn_seed = 0
 
-SimTime = 10000.
+SimTime = 1000000.
 h = 0.02
 Tcutoff = np.iinfo(np.int32).max
 Tsim = int(SimTime/h)
 recInt = np.iinfo(np.int32).max
 
-w_ps = np.arange(1.88, 1.951, 0.01)
+w_ps = np.arange(1.87, 1.951, 0.01)
 nw = len(w_ps)
 
-N = 100
+N = 50
 Nneur = N*nw
 
 tau_psc = 0.2   # ms
@@ -42,7 +62,7 @@
 I0 = 5.27
 
 rate = np.zeros(Nneur, dtype=fltSize) + 185.0    # Poisson noise rate, Hz (shouldn't  be 0)
-pcon = 0.2
+pcon = 0.25
 Ncon = int(N*N*pcon)
 pre = np.zeros(Ncon*nw, dtype='uint32')
 post = np.zeros(Ncon*nw, dtype='uint32')
@@ -68,8 +88,8 @@
 
 Vrec = np.zeros((int((Tsim  + recInt - 1)/recInt), Nneur), dtype=fltSize)
 
-#v0, n0, m0, h0 = 32.906693, 0.574676, 0.913177, 0.223994
-v0, n0, m0, h0 = -60.8457, 0.3763, 0.0833, 0.4636
+v0, n0, m0, h0 = 32.906693, 0.574676, 0.913177, 0.223994
+#v0, n0, m0, h0 = -60.8457, 0.3763, 0.0833, 0.4636
 Vm = np.zeros(Nneur, dtype=fltSize) + v0
 ns = np.zeros(Nneur, dtype=fltSize) + n0
 ms = np.zeros(Nneur, dtype=fltSize) + m0
@@ -85,7 +105,7 @@
 incTimes = np.zeros((NnumSp, Nneur), dtype='uint32')
 incSpWeights = np.zeros((NnumSp, Nneur), dtype=fltSize) + wInc*np.e/tau_psc
 #%%
-phh.setCalcParams(rank % 3, Tsim, Tcutoff, Nneur, Ncon*nw, recInt, h)
+phh.setCalcParams(rank, Tsim, Tcutoff, Nneur, Ncon*nw, recInt, h)
 
 phh.setIncomSpikes(incTimes, nums, incSpWeights, NnumSp)
 phh.setNeurVars(Vm, Vrec, ns, ms, hs)
@@ -134,4 +154,4 @@
 #    a2.set_xlim((0, SimTime/1000))
 #pl.show()
 
-pl.savefig('N_{}_rate_{}_w_n_{}_Ie_{:.2f}_pcon_{:.2f}_delayRND.png'.format(N, rate[0], w_n, I0, pcon))
+pl.savefig('N_{}_rate_{:.2f}_w_n_{:.3f}_Ie_{:.2f}_pcon_{:.2f}_delayRND.png'.format(N, rate[0], w_n, I0, pcon))