BIG PIPE TIMESLICE SERIES
BIG PIPE TIMESLICE WINDOWS

If I were you, I'd try SwitchToThread() to just relinquish processor time slice and return as soon as possible.Īs for your particular problem, it is too bad the synchronization has to be inter-process. I am actually surprised to hear there's a difference between Sleep(0) and Sleep(1)īecause as a minimum, the thread will have to wait for at least 1 processor time slice (not 100% guaranteed, but quite probable).

BIG PIPE TIMESLICE WINDOWS

setbVal ( 0, Qb ) print 'size of matrix storing pressure head history is' print shape ( Hs ) #watch out if this matrix is huge-make Mi bigger or Nt smaller so you don't have such large data sets.A call to Sleep() makes the OS suspend the calling thread and any processor time remaining from its slice is lost and Windows will immediately schedule another thread. setbVal ( j, bvt ) bv = bvt #record boundary value #specify an boundary inflow at node 0 n1.

BIG PIPE TIMESLICE SERIES

nodeTypes )) for j in range ( len ( aa )): #find the junction1s and set boundary value time series to.

ones ( M + 1 ) aa = arange ( 0, len ( n1. load ( open ( 'summary_results_saved_at2015_11_15_16_40_13.pkl', 'rb' )) Af = ICdata Qf = ICdata for i in range ( 0, Np ): #print 'Af = %f, A0= %f, H0/D = %f'%(Ai, A0, p0.HofA(A0,False)/Ds) if startfromsaved : #subsequent runs use old data A0 = Af Q0 = Qf else : Ai = (( Ds ** 2 ) / 4. M ) #now set the initial conditions if startfromsaved : #initialize with data from a previous simulation ICdata = pickle. config with new info ( fi, fc ) = rewritePipes ( fn, oldinp, Ns, Ls, Mrs, Ds, jt, bt, bv, r, h0s, q0s, T, M, a, elevs ) #create the network you're actually going to run n1 = PyNetwork ( fi, fc, 1 ) dt = n1. inp file with network connectivity oldinp = './indata/Alameda62.0.0.inp' #write new files. config files that will be used for actual runtime parameters fn = './indata/Alameda_m4' #old. zeros (( M / Mi * Nt, sum ( Ns ))) #new prefix for. zeros (( M / Mi * Nt, sum ( Ns ))) Us = np. H_Np], # where H0 is the pressure data as a function of x in pipe 0 at time slice i, etc # same indexing scheme holds for Us Hs = np. clock () startfromsaved = 0 #set = 1 to start from saved data #create matrices to store pressure and velocity data # i^th row of Hs is the i^th timeslice, where we take Mi time steps of length dt in between storing each time slice # the i^th time slice consists of the, H1.

config file to (A,Q) =(0,0) in all pipes q0s = * Np Mrs = * Np #Manning roughness coefficient =0.007 for all pipes Nstar = 1 #measuring point for each pipe #elevation data for all nodes elevs = nodeTypes Nn = len ( jt ) #boundary types: need to specify three parameters for each junction1: #-r: reflection (1 means reflect all waves, -1 means reflect none, 0 means apply one of the cases (0-2) in bt ) #-bt: boundary type (0 means specify A, 1 means specify Q, 2 means orifice outflow) #-bv: boundary value (a float that specifies value, if constant) # if nonconstant, subsequently use setBval(i,Y) for time series Y (length M+1) #set boundary type (this will only matter for nodes where r =0) bt = * Nn #all nodes orifice outflow bt = 1 #node 0 we'll specify inflow if bt = 0 : sbtype = 'A' elif bt = 1 : sbtype = 'Q' #set boundary values to 0(this will only matter for nodes where r =0) bv = * Nn #set reflection r = * Nn r = 0 # this node we'll specify inflow r = 1 #these nodes will reflect everything r = 1 r = 1 bt = 1 #set boundary type of node 0 to specify Q(t) h0s = * Np #initialize. 8 )) * 4 #set time steps based on dx to assure CFL condition M = ( M + Mi - M % Mi ) #round it up to be an even multiple of Mi jt = n0. Ds dx = / Ns for i in range ( Np )] M = int ( T * a / ( max ( dx ) *. Nedges a = 100 #pressure wave speed (this determines slot width) Ns = # set dx = 1 meter by assigning 1 grid cell per meter of length Ds = n0. Ls T = 20 #set simulation time per run Mi = 10 #number of time steps in between writes Nt = 1 #number of simulations to do Ttot = Nt * T # total simulation time Np = n0.

Big pipe timeslice