def difference(delta):
x = MC.runMC(MCcycles2, delta, idum, alpha)
return x.accepted * 1.0 / MCcycles2 - .5 #We want 50% accepted moves
#Loop over alpha values
for alpha in values:
#Determination of optimal delta value (for each alpha), i.e.
#finding the zero-point of the difference function by the bisection method
minimum = delta_min
maximum = delta_max
while maximum - minimum > tolerance:
if difference(minimum) * difference((minimum + maximum) / 2) < 0:
maximum = (minimum + maximum) / 2
else:
minimum = (minimum + maximum) / 2
delta = (minimum + maximum) / 2
#Run MC calculation (store results in x)
x = MC.runMC(MCcycles / nprocs, delta, idum, alpha)
idum = x.idum
#All nodes send their stats to the master
if myid != 0:
pypar.send((x.sum, x.squaresum, x.N, x.accepted), destination=0)
#The master receives, calculates and writes to file
if myid == 0:
for i in xrange(1, nprocs):
isum, isquaresum, iN, iaccepted = pypar.receive(source=i)
x.sum += isum
x.squaresum += isquaresum
x.N += iN
x.accepted += iaccepted
def difference(delta):
x = MC.runMC(MCcycles2,delta,idum,alpha)
return x.accepted*1.0/MCcycles2 - .5 #We want 50% accepted moves
#Loop over alpha values
for alpha in values:
#Determination of optimal delta value (for each alpha), i.e.
#finding the zero-point of the difference function by the bisection method
minimum = delta_min
maximum = delta_max
while maximum - minimum > tolerance:
if difference(minimum)*difference((minimum+maximum)/2) < 0:
maximum = (minimum + maximum) / 2
else:
minimum = (minimum + maximum) / 2
delta = (minimum + maximum) / 2
#Run MC calculation (store results in x)
x = MC.runMC(MCcycles/nprocs,delta,idum,alpha)
idum = x.idum
#All nodes send their stats to the master
if myid != 0:
pypar.send((x.sum,x.squaresum,x.N,x.accepted),destination=0)
#The master receives, calculates and writes to file
if myid == 0:
for i in xrange(1,nprocs):
isum,isquaresum,iN,iaccepted = pypar.receive(source=i)
x.sum += isum
x.squaresum += isquaresum
x.N += iN
x.accepted += iaccepted
#If it isn't the termination signal, do the job and return the result
if alpha != 0.0:
#Determination of optimal delta value (for each alpha), i.e.
#finding the zero-point of the difference function by the bisection method
minimum = delta_min
maximum = delta_max
while maximum - minimum > tolerance:
if difference(minimum)*difference((minimum+maximum)/2) < 0:
maximum = (minimum + maximum) / 2
else:
minimum = (minimum + maximum) / 2
delta = (minimum + maximum) / 2
#Run MC calculation (store results in x)
x = MC.runMC(MCcycles,delta,idum,alpha)
idum = x.idum
#Calculate statistics and store them in a result object
thisresult = result()
thisresult.alpha = alpha
thisresult.id = myid
thisresult.E = x.sum / x.N
E2 = x.squaresum / x.N
thisresult.sigma = sqrt(E2 - thisresult.E**2)
thisresult.acceptance = x.accepted*1.0/MCcycles
thisresult.error = thisresult.sigma / sqrt(x.N)
#Send to master
pypar.send(thisresult,destination=0)
#Loop over alpha values
for alpha in values:
#Determination of optimal delta value (for each alpha), i.e.
#finding the zero-point of the difference function by the bisection method
minimum = delta_min
maximum = delta_max
while maximum - minimum > tolerance:
if difference(minimum)*difference((minimum+maximum)/2) < 0:
maximum = (minimum + maximum) / 2
else:
minimum = (minimum + maximum) / 2
delta = (minimum + maximum) / 2
#Run MC calculation (store results in x)
x = MC.runMC(MCcycles,delta,idum,alpha)
idum = x.idum
#Calculate statistics
E = x.sum / x.N
E2 = x.squaresum / x.N
sigma = sqrt(E2 - E**2)
acceptance = x.accepted*1.0/MCcycles
error = sigma / sqrt(x.N)
#Print results to screen
#print 'alpha = %f, <E> = %f, sigma = %f, error = %f, acceptance = %f' %(alpha,E,sigma,error,acceptance)
#Print results to file
outfile.write('%f %f %f %f %f\n' %(alpha,E,sigma,error,acceptance))
