def make_movie( field="uu1", datadir="data/", proc=-1, extension="xz", format="native", tmin=0.0, tmax=1.0e38, amin=0.0, amax=1.0, transform="", oldfile=False, norm=None, save=None, figsize=(16, 4), fps=12, ): """ read 2D slice files and assemble an animation in a mpg movie. Quickly written from the example at http://matplotlib.sourceforge.net/faq/howto_faq.html Options: field --- which variable to slice datadir--- path to data directory proc --- an integer giving the processor to read a slice from extension --- which plane of xy,xz,yz,Xz. for 2D this should be overwritten. format --- endian. one of little, big, or native (default) tmin --- start time tmax --- end time amin --- minimum value for image scaling amax --- maximum value for image scaling transform --- insert arbitrary numerical code to modify the slice norm --- scales calar data save --- directory to save file figsize --- tuple containing the size of the figure fps --- Frames per seconds for the video """ import os from pencil.io import npfile from pencil import read import numpy as np import pylab as plt from matplotlib import colors # Global configuration: # lines plt.rcParams["lines.linewidth"] = 2 plt.rcParams["lines.color"] = "k" # font plt.rcParams["font.size"] = 30 plt.rcParams["font.family"] = "serif" # legend plt.rcParams["legend.fontsize"] = 20 plt.rcParams["legend.fancybox"] = False plt.rcParams["legend.numpoints"] = 2 plt.rcParams["legend.shadow"] = False plt.rcParams["legend.frameon"] = False # latex plt.rc("text", usetex=True) plt.rcParams["text.latex.preamble"] = [r"\usepackage{amsmath}"] datadir = os.path.expanduser(datadir) if proc < 0: filename = os.path.join(datadir, "slice_" + field + "." + extension) else: filename = os.path.join( datadir, "proc" + str(proc) + "/slice_" + field + "." + extension) # global dim # param = read.param(datadir) dim = read.dim(datadir, proc) if dim.precision == "D": precision = "d" else: precision = "f" grid = read.grid(datadir=datadir, trim=True) # set up slice plane if extension == "xy" or extension == "Xy": hsize = dim.nx vsize = dim.ny xlabel = "x" ylabel = "y" x = grid.x y = grid.y if extension == "xz": hsize = dim.nx vsize = dim.nz xlabel = "x" ylabel = "z" x = grid.x y = grid.z if extension == "yz": hsize = dim.ny vsize = dim.nz xlabel = "y" ylabel = "z" x = grid.y y = grid.z plane = np.zeros((vsize, hsize), dtype=precision) infile = npfile(filename, endian=format) files = [] fig = plt.figure(figsize=figsize) fig.subplots_adjust(left=0.12, bottom=0.1, right=0.98, top=0.96, wspace=0.23, hspace=0.2) ax = fig.add_subplot(111) ifirst = True islice = 0 while 1: try: raw_data = infile.fort_read(precision) except ValueError: break except TypeError: break if oldfile: t = raw_data[-1] plane = raw_data[:-1].reshape(vsize, hsize) else: slice_z2pos = raw_data[-1] t = raw_data[-2] plane = raw_data[:-2].reshape(vsize, hsize) if transform: exec("plane = plane" + transform) if t > tmin and t < tmax: ax.cla() title = "t = %11.3e" % t ax.set_title(title) ax.set_xlabel(xlabel) ax.set_ylabel(ylabel) ax.imshow( plane, origin="lower", vmin=amin, vmax=amax, norm=norm, cmap="hot", extent=[x[0], x[-1], y[0], y[-1]], aspect=1, ) fname = "_tmp%03d.png" % islice print("Saving frame", fname) fig.savefig(fname) files.append(fname) if ifirst: print( "----islice----------t---------min-------max-------delta") print("%10i %10.3e %10.3e %10.3e %10.3e" % (islice, t, plane.min(), plane.max(), plane.max() - plane.min())) ifirst = False islice += 1 if t > tmax: break print("Making movie animation.mpg - this make take a while") os.system( f"mencoder 'mf://_tmp*.png' -mf type=png:fps={fps} -ovc lavc -lavcopts vcodec=wmv2 -oac copy -o animation.mpg" ) if save: os.system(f"mv _tmp*.png {save}") print(f"Moving files to {save}") else: os.system("rm _tmp*.png") print("Removing all files") infile.close()
def calc( self, aver=[], datatopdir=".", lskip_zeros=False, proc=0, rank=0, rmfzeros=1, rmbzeros=1, iy=None, l_correction=False, t_correction=0.0, dim=None, timereducer=None, trargs=[], tindex=(0, None), imask=None, ): """object returns time dependent meridional tensors from Averages object aver.z. u, acoef and bcoef and aver.t For long DNS runs the 'zaverages.dat' file can be very large so MPI may be required and the data is loaded by processor as default. lskip_zeros=True identifies the resetting of the testfield and rmbzeros and rmfzeros number to exclude before and following By default none are removed. iy is the index array that is computed in this MPI process, which may be a subset of the array on this processor l_correction=True permits the pencil coefficients computed prior to the Pencil Code correction implemented after time=t_correction to be rescaled accordingly to match the new formulation. trargs contain optional arguments for the time treatments: mean, smoothing, etc. tindex is set to limit the range of the iterations loaded from Averages in zaverages.dat The index imask, excluding the resets, can be specified to ensure all processes use the same mask """ import numpy as np import os from pencil import read os.chdir(datatopdir) # return to working directory grid = read.grid(proc=proc, trim=True, quiet=True) # if iy None or scalar create numpy array try: iy.size > 0 except: print("exception") if iy == None: print("exception None") iy = np.arange(grid.y.size) else: print("exception int") iy = np.array(iy) if rank == 0: print("iy size is {0}".format(iy.shape)) r, theta = np.meshgrid(grid.x, grid.y[iy], indexing="ij") del (grid, theta) # conserve memory print("rank {0} calculating tensors for proc {1}".format(rank, proc)) # string containers for zaverages.z keys uformat = "u{0}mxy" alpformat = "alp{0}{1}xy" etaformat = "eta{0}{1}{2}xy" # imask calculated once for MPI/processor consistency if rank == 0: print("Removing zeros") old_size = aver.t.shape # if imask is not provided either exclude the zeros or use the full time series try: imask.size > 0 print("imask shape is {}".format(imask.shape)) except: if lskip_zeros: index = alpformat.format(1, 1) izero = np.array( np.where( aver.z.__getattribute__(index) [:, int(aver.z.__getattribute__(index).shape[-2] / 2), int(aver.z.__getattribute__(index).shape[-1] / 2), ] == 0))[0] rmfrange = np.arange(0, rmfzeros - 1) rmbrange = np.arange(0, rmbzeros - 1) rmpoints = np.array([], dtype=int) for zero in izero: rmpoints = np.append(rmpoints, rmfrange + zero) rmpoints = np.append(rmpoints, zero - rmbrange) if izero.size > 0: imask = np.delete(np.where(aver.t), rmpoints) if rank == 0: print("Removed {0} zeros from {1} resets".format( len(rmpoints), len(izero))) print( "Resets occured at save points {0}".format(izero)) else: imask = np.where(aver.t)[0] del (rmpoints, rmbrange, rmfrange) else: imask = np.arange(aver.t.size) if rank == 0: print("Skipped zero removals.") # update the time of the snapshots included self.t = aver.t[imask] # Correction to Pencil Code error may be required on old data if l_correction: if dim == None: dim = read.dim(quiet=True) itcorr = np.where(aver.t[imask] < t_correction)[0] index = alpformat.format(1, 3) aver.z.__getattribute__( index)[itcorr] *= -dim.nprocz / (dim.nprocz - 2.0) for j in range(0, 3): index = alpformat.format(3, j + 1) aver.z.__getattribute__( index)[itcorr] *= -dim.nprocz / (dim.nprocz - 2.0) index = etaformat.format(1, 1, 1) aver.z.__getattribute__( index)[itcorr] *= -dim.nprocz / (dim.nprocz - 2.0) for j in range(0, 3): index = etaformat.format(j + 1, 2, 1) aver.z.__getattribute__( index)[itcorr] *= -dim.nprocz / (dim.nprocz - 2.0) index = etaformat.format(1, 1, 2) aver.z.__getattribute__( index)[itcorr] *= -dim.nprocz / (dim.nprocz - 2.0) for j in range(0, 3): index = etaformat.format(j + 1, 2, 2) aver.z.__getattribute__( index)[itcorr] *= -dim.nprocz / (dim.nprocz - 2.0) # set up place holders for the Pencil Code tensor coefficients index = alpformat.format(1, 1) u = np.zeros( [3, len(imask), aver.z.__getattribute__(index).shape[-2], iy.size]) alp = np.zeros([ 3, 3, len(imask), aver.z.__getattribute__(index).shape[-2], iy.size ]) eta = np.zeros([ 3, 3, 3, len(imask), aver.z.__getattribute__(index).shape[-2], iy.size ]) if rank == 0: print(u.shape, aver.z.__getattribute__(index)[imask, :, :].shape) # store the individual components in the z-averages as tensors for i, coord in zip(range(0, 3), ("x", "y", "z")): try: index = uformat.format(coord) if iy.size > 1: tmp = aver.z.__getattribute__(index)[:, :, iy] u[i, :, :, :] = tmp[imask] else: u[i, :, :, 0] = aver.z.__getattribute__(index)[imask, :, iy] except KeyError: pass for i in range(0, 3): for j in range(0, 3): index = alpformat.format(i + 1, j + 1) if iy.size > 1: tmp = aver.z.__getattribute__(index)[:, :, iy] alp[j, i, :, :, :] = tmp[imask] else: alp[j, i, :, :, 0] = aver.z.__getattribute__(index)[imask, :, iy] for i in range(0, 3): for j in range(0, 3): index1 = etaformat.format(i + 1, j + 1, 1) index2 = etaformat.format(i + 1, j + 1, 2) # Sign difference with Schrinner + r correction if iy.size > 1: tmp = aver.z.__getattribute__(index1)[:, :, iy] # eta[0,j,i,:,:,:] = -tmp[imask] # JOERN, no sign correction eta[0, j, i, :, :, :] = tmp[imask] tmp = aver.z.__getattribute__(index2)[:, :, iy] # eta[1,j,i,:,:,:] = -tmp[imask]*r # JOERN, no sign correction eta[1, j, i, :, :, :] = tmp[imask] * r del tmp else: # eta[0,j,i,:,:,0] = -aver.z.__getattribute__(index1)[imask,:,iy] # JOERN, no sign correction # eta[1,j,i,:,:,0] = -aver.z.__getattribute__(index2)[imask,:,iy]*r[:,0] # JOERN, no sign correction eta[0, j, i, :, :, 0] = aver.z.__getattribute__(index1)[imask, :, iy] eta[1, j, i, :, :, 0] = (aver.z.__getattribute__(index2)[imask, :, iy] * r[:, 0]) # apply the specified averaging or smoothing: 'None' returns unprocessed arrays if callable(timereducer): u = timereducer(u, trargs) alp = timereducer(alp, trargs) eta = timereducer(eta, trargs) if rank == 0: print("Old time dimension has length: {0}".format(old_size)) print("New time dimension has length: {0}".format(alp.shape[-3])) # Create output tensors datatype = alp.dtype datashape = [alp.shape[-3], alp.shape[-2], alp.shape[-1], 1] setattr(self, "utensor", np.zeros([3] + datashape, dtype=datatype)) setattr(self, "alpha", np.zeros([3, 3] + datashape, dtype=datatype)) setattr(self, "beta", np.zeros([3, 3] + datashape, dtype=datatype)) setattr(self, "gamma", np.zeros([3] + datashape, dtype=datatype)) setattr(self, "delta", np.zeros([3] + datashape, dtype=datatype)) setattr(self, "kappa", np.zeros([3, 3, 3] + datashape, dtype=datatype)) setattr(self, "acoef", np.zeros([3, 3] + datashape, dtype=datatype)) setattr(self, "bcoef", np.zeros([3, 3, 3] + datashape, dtype=datatype)) """ All tensors need to be reordered nz,ny,nx,nt for efficient writing to disk """ # Calculating a and b matrices self.acoef[:, :, :, :, :, 0] = np.copy(alp) self.acoef = np.swapaxes(self.acoef, -4, -1) self.acoef = np.swapaxes(self.acoef, -3, -2) self.bcoef[:, :, :, :, :, :, 0] = np.copy(eta) self.bcoef = np.swapaxes(self.bcoef, -4, -1) self.bcoef = np.swapaxes(self.bcoef, -3, -2) irr, ith, iph = 0, 1, 2 # u-tensor print("Calculating utensor on rank {}".format(rank)) # utensor[:,:,:,:,0] = u[:,:,:,:] - np.mean(u[:,:,:,:],axis=1,keepdims=True) self.utensor[:, :, :, :, 0] = u[:, :, :, :] self.utensor = np.swapaxes(self.utensor, -4, -1) self.utensor = np.swapaxes(self.utensor, -3, -2) # Alpha tensor print("Calculating alpha on rank {}".format(rank)) self.alpha[irr, irr, :, :, :, 0] = (alp[irr, irr, :, :, :] - eta[ith, ith, irr, :, :, :] / r) self.alpha[irr, ith, :, :, :, 0] = 0.5 * ( alp[ith, irr, :, :, :] + eta[ith, irr, irr, :, :, :] / r + alp[irr, ith, :, :, :] - eta[ith, ith, ith, :, :, :] / r) self.alpha[irr, iph, :, :, :, 0] = 0.5 * (alp[iph, irr, :, :, :] + alp[irr, iph, :, :, :] - eta[ith, ith, iph, :, :, :] / r) self.alpha[ith, irr, :, :, :, 0] = self.alpha[irr, ith, :, :, :, 0] self.alpha[ith, ith, :, :, :, 0] = (alp[ith, ith, :, :, :] + eta[ith, irr, ith, :, :, :] / r) self.alpha[ith, iph, :, :, :, 0] = 0.5 * (alp[iph, ith, :, :, :] + alp[ith, iph, :, :, :] + eta[ith, irr, iph, :, :, :] / r) self.alpha[iph, irr, :, :, :, 0] = self.alpha[irr, iph, :, :, :, 0] self.alpha[iph, ith, :, :, :, 0] = self.alpha[ith, iph, :, :, :, 0] self.alpha[iph, iph, :, :, :, 0] = alp[iph, iph, :, :, :] self.alpha = np.swapaxes(self.alpha, -4, -1) self.alpha = np.swapaxes(self.alpha, -3, -2) # Gamma vector print("Calculating gamma on rank {}".format(rank)) self.gamma[irr, :, :, :, 0] = -0.5 * (alp[iph, ith, :, :, :] - alp[ith, iph, :, :, :] - eta[ith, irr, iph, :, :, :] / r) self.gamma[ith, :, :, :, 0] = -0.5 * (alp[irr, iph, :, :, :] - alp[iph, irr, :, :, :] - eta[ith, ith, iph, :, :, :] / r) self.gamma[iph, :, :, :, 0] = -0.5 * ( alp[ith, irr, :, :, :] - alp[irr, ith, :, :, :] + eta[ith, irr, irr, :, :, :] / r + eta[ith, ith, ith, :, :, :] / r) self.gamma = np.swapaxes(self.gamma, -4, -1) self.gamma = np.swapaxes(self.gamma, -3, -2) # Beta tensor print("Calculating beta on rank {}".format(rank)) self.beta[irr, irr, :, :, :, 0] = -0.5 * eta[ith, iph, irr, :, :, :] self.beta[irr, ith, :, :, :, 0] = 0.25 * (eta[irr, iph, irr, :, :, :] - eta[ith, iph, ith, :, :, :]) self.beta[irr, iph, :, :, :, 0] = 0.25 * (eta[ith, irr, irr, :, :, :] - eta[ith, iph, iph, :, :, :] - eta[irr, ith, irr, :, :, :]) self.beta[ith, irr, :, :, :, 0] = self.beta[irr, ith, :, :, :, 0] self.beta[ith, ith, :, :, :, 0] = 0.5 * eta[irr, iph, ith, :, :, :] self.beta[ith, iph, :, :, :, 0] = 0.25 * (eta[ith, irr, ith, :, :, :] + eta[irr, iph, iph, :, :, :] - eta[irr, ith, ith, :, :, :]) self.beta[iph, irr, :, :, :, 0] = self.beta[irr, iph, :, :, :, 0] self.beta[iph, ith, :, :, :, 0] = self.beta[ith, iph, :, :, :, 0] self.beta[iph, iph, :, :, :, 0] = 0.5 * (eta[ith, irr, iph, :, :, :] - eta[irr, ith, iph, :, :, :]) # Sign convention to match with meanfield_e_tensor # self.beta = -self.beta #JOERN, not needed self.beta = np.swapaxes(self.beta, -4, -1) self.beta = np.swapaxes(self.beta, -3, -2) # Delta vector print("Calculating delta on rank {}".format(rank)) self.delta[irr, :, :, :, 0] = 0.25 * (eta[irr, ith, ith, :, :, :] - eta[ith, irr, ith, :, :, :] + eta[irr, iph, iph, :, :, :]) self.delta[ith, :, :, :, 0] = 0.25 * (eta[ith, irr, irr, :, :, :] - eta[irr, ith, irr, :, :, :] + eta[ith, iph, iph, :, :, :]) self.delta[iph, :, :, :, 0] = -0.25 * (eta[irr, iph, irr, :, :, :] + eta[ith, iph, ith, :, :, :]) # Sign convention to match with meanfield_e_tensor # self.delta = -self.delta #JOERN, not needed self.delta = np.swapaxes(self.delta, -4, -1) self.delta = np.swapaxes(self.delta, -3, -2) # Kappa tensor print("Calculating kappa on rank {}".format(rank)) for i in range(0, 3): self.kappa[irr, irr, i, :, :, :, 0] = -eta[irr, irr, i, :, :, :] self.kappa[ith, irr, i, :, :, :, 0] = -0.5 * (eta[ith, irr, i, :, :, :] + eta[irr, ith, i, :, :, :]) self.kappa[iph, irr, i, :, :, :, 0] = -0.5 * eta[irr, iph, i, :, :, :] self.kappa[irr, ith, i, :, :, :, 0] = self.kappa[ith, irr, i, :, :, :, 0] self.kappa[ith, ith, i, :, :, :, 0] = -eta[ith, ith, i, :, :, :] self.kappa[iph, ith, i, :, :, :, 0] = -0.5 * eta[ith, iph, i, :, :, :] self.kappa[irr, iph, i, :, :, :, 0] = self.kappa[iph, irr, i, :, :, :, 0] self.kappa[ith, iph, i, :, :, :, 0] = self.kappa[iph, ith, i, :, :, :, 0] self.kappa[iph, iph, i, :, :, :, 0] = 1e-91 # Sign convention to match with meanfield_e_tensor # self.kappa = -self.kappa #JOERN, not needed self.kappa = np.swapaxes(self.kappa, -4, -1) self.kappa = np.swapaxes(self.kappa, -3, -2) setattr(self, "imask", imask)
def zav2h5( folder='.', dataset='', filename='emftensors.h5', timereducer='mean', hdf5dir='data/', l_correction=True, t_correction=8972., rmfzeros=4, rmbzeros=2, dgroup='emftensor', ): """ If large dataset MPI may be required. Loads Averages object and applies tensors calculation and reforms for efficient writing to hdf5 for mean field module simulations. MPI call needs to be improved to avoid MemoryError for large files with read.aver(plane_list=['z']) timereducers needs to be expanded to include various smoothing options """ import numpy as np from pencil import read from pencil.read import aver from pencil.export import create_h5, fvars, create_aver_sph # from pencil.export import create_h5.fvars as fvars # from pencil.export import create_aver_sph from pencil.calc import tensors_sph import h5py import copy timereducers = { 'mean': lambda x, args: np.mean(x, axis=-3, keepdims=True), #np.std(x,axis=-3)), 'mean_last': lambda x, args: np.mean(np.take( x, np.arange(-int(args[0]), 0, 1), axis=-3), axis=-3, keepdims=True), 'none': lambda x, args: x } if not timereducer in timereducers: raise ValueError( 'timereducer "{}" undefined in timereducers'.format(timereducer) + ' options: {}'.format(timereducers.keys())) if len(dataset) == 0: dataset = timereducer with open('zaver.in', 'r') as f: zavers = f.read().splitlines() """ Find out if the calculation is parallel and distribute the arrays according to y-index and ipz=0 processor layout """ try: from mpi4py import MPI comm = MPI.COMM_WORLD rank = comm.Get_rank() # rank of processor on which this script runs size = comm.Get_size() # number of ~ ~ ~ ~ l_mpi = True l_mpi = l_mpi and (size != 1) except ImportError: l_mpi = False rank = 0 size = 1 comm = None dim = read.dim() nx, nny = dim.nx, dim.ny ayindex = np.arange(nny) if l_mpi: y_chunks = np.array_split(ayindex, size, axis=0) yindex = y_chunks[rank] ny = yindex.size else: yindex = ayindex # vector 0 ... nygrid-1 ny = nny ncpus = dim.nprocx * dim.nprocy aprocs = np.arange(ncpus) # vector 0 ... nprocx*nprocy-1 if np.mod(ncpus, size) > 0: print('number of processes must divide {} cpus'.format(ncpus)) quit() if l_mpi: if size > aprocs.size: nsz = size / aprocs.size for ii in range(1, nsz): tmproc = np.append(aprocs, aprocs) aprocs = np.sort(tmproc) proc_chunks = np.array_split(aprocs, size, axis=0) proc = proc_chunks[rank] else: proc = aprocs """Set up hdf5 file and create datasets in which to save the tensors """ lskip_zeros = rmfzeros + rmbzeros > 0 if rank == 0: # if root processor grid = read.grid(trim=True, quiet=True) # read grid zav = read.aver(proc=0, plane_list=['z' ]) # read zaverages of root proc of PC run tensor = tensors_sph( # decompose into individual effect tensors zav, proc=proc[0], rank=0, lskip_zeros=lskip_zeros, iy=[ int(ny / 2 / dim.nprocy), ], timereducer=timereducers[timereducer], #trargs=trargs, rmfzeros=rmfzeros, rmbzeros=rmbzeros, l_correction=l_correction, t_correction=t_correction, dim=dim, #tindex=tindex ) if 'mean' in dataset: nt = 1 else: nt = tensor.t.size create_aver_sph(hdf5dir + filename, dataset, fvars, (1, ny, nx, nt), (0, grid.y, grid.x, tensor.t), hdf5dir=hdf5dir, dgroup=dgroup) if l_mpi: imask = comm.bcast(tensor.imask, root=0) else: imask = tensor.imask import os if os.path.exists(os.path.join(folder, 'averages/z.h5')): zav = aver(plane_list=['z']) # read all averages tensor_buf = tensors_sph( # calculate tensors aver=zav, rank=rank, lskip_zeros=lskip_zeros, timereducer=timereducers[timereducer], #trargs=trargs, rmfzeros=rmfzeros, rmbzeros=rmbzeros, l_correction=l_correction, t_correction=t_correction, dim=dim, #tindex=tindex, imask=imask) else: yndx_tmp = np.array_split(yindex, dim.nprocy) # list of vectors ipy*ny/nprocy ... (ipy+1)*ny/nprocy - 1, ipy=0,nprocy-1 for ipy in range(dim.nprocy): # over all y processors of the PC run for ipx in range( dim.nprocx): # over all x processors of the PC run iproc = dim.nprocx * ipy + ipx # proc rank of the PC run (0 ... nprocx*nprocy-1) yndx = yndx_tmp[ipy] - ipy * int(dim.nygrid / dim.nprocy) zav = aver(proc=iproc, plane_list=['z']) # read averages from proc iproc print('calculating tensors on proc {0} rank {1}'.format( iproc, rank)) """ if iproc==1: # as there is corrupted data on proc 1 with open('zaver.in', 'r') as f: zavers = f.read().splitlines() for zaver in zavers: zav.z.__setattr__(zaver,np.insert( zav.z.__getattribute__(zaver),3766, 0.5*(zav.z.__getattribute__(zaver)[3766]+ zav.z.__getattribute__(zaver)[3767]),axis=0)) zav.t=np.insert(zav.t,3766,0.5*(zav.t[3766]+zav.t[3767]),axis=0) """ tensor_buf = tensors_sph( # calculate tensors aver=zav, proc=iproc, rank=rank, lskip_zeros=lskip_zeros, iy=yndx, timereducer=timereducers[timereducer], #trargs=trargs, rmfzeros=rmfzeros, rmbzeros=rmbzeros, l_correction=l_correction, t_correction=t_correction, dim=dim, #tindex=tindex, imask=imask) if ipx == 0: tensor = copy.deepcopy(tensor_buf) else: for field, comp in fvars: setattr( tensor, field, np.concatenate( (tensor.__getattribute__(field), tensor_buf.__getattribute__(field)), axis=len(comp) + 2)) if l_mpi: comm.barrier() ds = h5py.File(hdf5dir + filename, 'a', driver='mpio', comm=comm) else: ds = h5py.File(hdf5dir + filename, 'a') # open HDF5 file for field, comp in fvars: print('writing {0} from rank {1} for proc {2}'.format( field, rank, iproc)) dsname = '{0}/{1}/{2}'.format(dgroup, field, dataset) if len(comp) == 1: ds[dsname][:, :, yndx_tmp[ipy], :] = tensor.__getattribute__(field) elif len(comp) == 2: ds[dsname][:, :, :, yndx_tmp[ipy], :] = tensor.__getattribute__(field) else: ds[dsname][:, :, :, :, yndx_tmp[ipy], :] = tensor.__getattribute__(field) ds.close()
def find_fixed( self, datadir="data", var_file="VAR0", trace_field="bb", ti=-1, tf=-1, tracer_file_name=None, ): """ Find the fixed points to a snapshot or existing tracer file. call signature:: find_fixed(datadir='data', var_file='VAR0', trace_field='bb', ti=-1, tf=-1, tracer_file_name=None): Keyword arguments: *datadir*: Data directory. *var_file*: Varfile to be read. *trace_field*: Vector field used for the streamline tracing. *ti*: Initial VAR file index for tracer time sequences. Overrides 'var_file'. *tf*: Final VAR file index for tracer time sequences. Overrides 'var_file'. *tracer_file_name* Name of the tracer file to be read. If 'None' compute the tracers. """ import numpy as np import multiprocessing as mp from pencil import read from pencil import math from pencil.diag.tracers import Tracers from pencil.calc.streamlines import Stream from pencil.math.interpolation import vec_int if self.params.int_q == "curly_A": self.curly_A = [] if self.params.int_q == "ee": self.ee = [] # Multi core setup. if not (np.isscalar(self.params.n_proc)) or (self.params.n_proc % 1 != 0): print("Error: invalid processor number") return -1 queue = mp.Queue() # Make sure to read the var files with the correct magic. magic = [] if trace_field == "bb": magic.append("bb") if trace_field == "jj": magic.append("jj") if trace_field == "vort": magic.append("vort") if self.params.int_q == "ee": magic.append("bb") magic.append("jj") dim = read.dim(datadir=datadir) # Check if user wants a tracer time series. if (ti % 1 == 0) and (tf % 1 == 0) and (ti >= 0) and (tf >= ti): series = True var_file = "VAR{0}".format(ti) n_times = tf - ti + 1 else: series = False n_times = 1 self.t = np.zeros(n_times) # Read the initial field. var = read.var( var_file=var_file, datadir=datadir, magic=magic, quiet=True, trimall=True ) self.t[0] = var.t grid = read.grid(datadir=datadir, quiet=True, trim=True) field = getattr(var, trace_field) param2 = read.param(datadir=datadir, quiet=True) if self.params.int_q == "ee": ee = var.jj * param2.eta - math.cross(var.uu, var.bb) self.params.datadir = datadir self.params.var_file = var_file self.params.trace_field = trace_field # Get the simulation parameters. self.params.dx = var.dx self.params.dy = var.dy self.params.dz = var.dz self.params.Ox = var.x[0] self.params.Oy = var.y[0] self.params.Oz = var.z[0] self.params.Lx = grid.Lx self.params.Ly = grid.Ly self.params.Lz = grid.Lz self.params.nx = dim.nx self.params.ny = dim.ny self.params.nz = dim.nz tracers = Tracers() tracers.params = self.params # Create the mapping for all times. if not tracer_file_name: tracers.find_tracers( var_file=var_file, datadir=datadir, trace_field=trace_field, ti=ti, tf=tf, ) else: tracers.read(datadir=datadir, file_name=tracer_file_name) self.tracers = tracers # Set some default values. self.t = np.zeros((tf - ti + 1) * series + (1 - series)) self.fixed_index = np.zeros((tf - ti + 1) * series + (1 - series)) self.poincare = np.zeros( [ int(self.params.trace_sub * dim.nx), int(self.params.trace_sub * dim.ny), n_times, ] ) ix0 = range(0, int(self.params.nx * self.params.trace_sub) - 1) iy0 = range(0, int(self.params.ny * self.params.trace_sub) - 1) self.fixed_points = [] self.fixed_sign = [] self.fixed_tracers = [] # Start the parallelized fixed point finding. for tidx in range(n_times): if tidx > 0: var = read.var( var_file="VAR{0}".format(tidx + ti), datadir=datadir, magic=magic, quiet=True, trimall=True, ) field = getattr(var, trace_field) self.t[tidx] = var.t proc = [] sub_data = [] fixed = [] fixed_sign = [] fixed_tracers = [] for i_proc in range(self.params.n_proc): proc.append( mp.Process( target=self.__sub_fixed, args=(queue, ix0, iy0, field, self.tracers, tidx, var, i_proc), ) ) for i_proc in range(self.params.n_proc): proc[i_proc].start() for i_proc in range(self.params.n_proc): sub_data.append(queue.get()) for i_proc in range(self.params.n_proc): proc[i_proc].join() for i_proc in range(self.params.n_proc): # Extract the data from the single cores. Mind the order. sub_proc = sub_data[i_proc][0] fixed.extend(sub_data[i_proc][1]) fixed_tracers.extend(sub_data[i_proc][2]) fixed_sign.extend(sub_data[i_proc][3]) self.fixed_index[tidx] += sub_data[i_proc][4] self.poincare[sub_proc :: self.params.n_proc, :, tidx] = sub_data[ i_proc ][5] for i_proc in range(self.params.n_proc): proc[i_proc].terminate() # Discard fixed points which lie too close to each other. fixed, fixed_tracers, fixed_sign = self.__discard_close_fixed_points( np.array(fixed), np.array(fixed_sign), np.array(fixed_tracers), var ) if self.fixed_points is None: self.fixed_points = [] self.fixed_sign = [] self.fixed_tracers = [] self.fixed_points.append(np.array(fixed)) self.fixed_sign.append(np.array(fixed_sign)) self.fixed_tracers.append(fixed_tracers) # Compute the traced quantities along the fixed point streamlines. if (self.params.int_q == "curly_A") or (self.params.int_q == "ee"): for t_idx in range(0, n_times): if self.params.int_q == "curly_A": self.curly_A.append([]) if self.params.int_q == "ee": self.ee.append([]) for fixed in self.fixed_points[t_idx]: # Trace the stream line. xx = np.array([fixed[0], fixed[1], self.params.Oz]) # time = np.linspace(0, self.params.Lz/np.max(abs(field[2])), 10) field_strength_z0 = vec_int( xx, field, [var.dx, var.dy, var.dz], [var.x[0], var.y[0], var.z[0]], [len(var.x), len(var.y), len(var.z)], interpolation=self.params.interpolation, ) field_strength_z0 = np.sqrt(np.sum(field_strength_z0 ** 2)) time = np.linspace(0, 4 * self.params.Lz / field_strength_z0, 500) stream = Stream(field, self.params, xx=xx, time=time) # Do the field line integration. if self.params.int_q == "curly_A": curly_A = 0 for l in range(stream.iterations - 1): aaInt = vec_int( (stream.tracers[l + 1] + stream.tracers[l]) / 2, var.aa, [var.dx, var.dy, var.dz], [var.x[0], var.y[0], var.z[0]], [len(var.x), len(var.y), len(var.z)], interpolation=self.params.interpolation, ) curly_A += np.dot( aaInt, (stream.tracers[l + 1] - stream.tracers[l]) ) self.curly_A[-1].append(curly_A) if self.params.int_q == "ee": ee_p = 0 for l in range(stream.iterations - 1): eeInt = vec_int( (stream.tracers[l + 1] + stream.tracers[l]) / 2, ee, [var.dx, var.dy, var.dz], [var.x[0], var.y[0], var.z[0]], [len(var.x), len(var.y), len(var.z)], interpolation=self.params.interpolation, ) ee_p += np.dot( eeInt, (stream.tracers[l + 1] - stream.tracers[l]) ) self.ee[-1].append(ee_p) if self.params.int_q == "curly_A": self.curly_A[-1] = np.array(self.curly_A[-1]) if self.params.int_q == "ee": self.ee[-1] = np.array(self.ee[-1]) return 0
def update(self, hard=False, quiet=True): """Update simulation object: if not read in: - read param.nml - read grid and ghost grid Set hard=True to force update. """ from os.path import exists from os.path import join from pencil.read import param, grid, dim REEXPORT = False if hard == True: self.param = False self.grid = False self.ghost_grid = False self.dim = False REEXPORT = True if self.param == False: try: if exists(join(self.datadir, 'param.nml')): print('~ Reading param.nml.. ') param = param(quiet=quiet, datadir=self.datadir) self.param = {} # read params into Simulation object for key in dir(param): if key.startswith('_') or key == 'read': continue if type(getattr(param, key)) in [bool, list, float, int, str]: self.param[key] = getattr(param, key) else: try: # allow for nested param objects self.param[key] = {} for subkey in dir(getattr(param, key)): if subkey.startswith( '_') or subkey == 'read': continue if type( getattr(getattr(param, key), subkey)) in [ bool, list, float, int, str ]: self.param[key][subkey] = getattr( getattr(param, key), subkey) except: # not nested param objects continue REEXPORT = True else: if not quiet: print('? WARNING: for ' + self.path + '\n? Simulation has ' + 'not run yet! Meaning: No param.nml found!') REEXPORT = True except: print('! ERROR: while reading param.nml for ' + self.path) self.param = False REEXPORT = True if self.param != False and (self.grid == False or self.ghost_grid == False): # read grid only if param is not False try: print('~ Reading grid.. ') self.grid = grid(datadir=self.datadir, trim=True, quiet=True) print('~ Reading ghost_grid.. ') self.ghost_grid = grid(datadir=self.datadir, trim=False, quiet=True) print('~ Reading dim.. ') self.dim = dim(datadir=self.datadir) if not quiet: print('# Updating grid and ghost_grid succesfull') REEXPORT = True # adding lx, dx etc to params self.param['Lx'] = self.grid.Lx self.param['Ly'] = self.grid.Ly self.param['Lz'] = self.grid.Lz self.param['lx'] = self.grid.Lx self.param['ly'] = self.grid.Ly self.param['lz'] = self.grid.Lz self.param['dx'] = self.grid.dx self.param['dy'] = self.grid.dy self.param['dz'] = self.grid.dz except: if not quiet: print( '? WARNING: Updating grid and ghost_grid ' + 'was not successfull, since run has not yet started.') if self.started() or (not quiet): print('? WARNING: Couldnt load grid for ' + self.path) self.grid = False self.ghost_grid = False self.dim = False REEXPORT = True elif self.param == False: if not quiet: print('? WARNING: Updating grid and ghost_grid ' + 'was not successfull, since run has not yet started.') self.grid = False self.ghost_grid = False self.dim = False REEXPORT = True if REEXPORT == True: self.export() return self
def kernel_smooth( sim_path, src, dst, magic=["meanuu"], par=[], comm=None, gd=[], grp_overwrite=False, overwrite=False, rank=0, size=1, nghost=3, kernel=1., status="a", chunksize=1000.0, dtype=np.float64, quiet=True, nmin=32, typ='piecewise', mode=list(), ): if comm: overwrite = False if isinstance(par, list): os.chdir(sim_path) par = read.param(quiet=True, conflicts_quiet=True) if isinstance(gd, list): os.chdir(sim_path) gd = read.grid(quiet=True) # get data dimensions nx, ny, nz = ( src["settings"]["nx"][0], src["settings"]["ny"][0], src["settings"]["nz"][0], ) mx, my, mz = ( src["settings"]["mx"][0], src["settings"]["my"][0], src["settings"]["mz"][0], ) # extend gost zones to include up to 1.5 * kernel length) dx = max(src['grid/dx'][()], src['grid/dy'][()], src['grid/dz'][()]) nkernel = np.int(2.5 * kernel / dx) sigma = kernel / dx print('sigma {:.2f}, kernel {:.2f}, dx {:.2f}'.format(sigma, kernel, dx)) # split data into manageable memory chunks dstchunksize = 8 * nx * ny * nz / 1024 * 1024 if dstchunksize > chunksize: nchunks = cpu_optimal( nx, ny, nz, quiet=quiet, mvar=src["settings/mvar"][0], maux=src["settings/maux"][0], MBmin=chunksize, nmin=nmin, size=size, )[1] else: nchunks = [1, 1, 1] print("nchunks {}".format(nchunks)) # for mpi split chunks across processes if size > 1: locindx = np.array_split(np.arange(nx) + nghost, nchunks[0]) locindy = np.array_split(np.arange(ny) + nghost, nchunks[1]) locindz = np.array_split(np.arange(nz) + nghost, nchunks[2]) indx = [ locindx[np.mod( rank + int(rank / nchunks[2]) + int(rank / nchunks[1]), nchunks[0])] ] indy = [locindy[np.mod(rank + int(rank / nchunks[2]), nchunks[1])]] indz = [locindz[np.mod(rank, nchunks[2])]] allchunks = 1 else: locindx = np.array_split(np.arange(nx) + nghost, nchunks[0]) locindy = np.array_split(np.arange(ny) + nghost, nchunks[1]) locindz = np.array_split(np.arange(nz) + nghost, nchunks[2]) indx = np.array_split(np.arange(nx) + nghost, nchunks[0]) indy = np.array_split(np.arange(ny) + nghost, nchunks[1]) indz = np.array_split(np.arange(nz) + nghost, nchunks[2]) allchunks = nchunks[0] * nchunks[1] * nchunks[2] if 1 in nchunks: mode = ["reflect", "reflect", "reflect"] for ich in range(3): if nchunks[ich] == 1: mode[2 - ich] = "wrap" if mode[2 - ich] == "reflect": typ = "piecewise" else: typ = "all" print('mode:', mode, 'typ:', typ) # save time dataset_h5( dst, "time", status=status, data=src["time"][()], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype, ) # ensure derived variables are in a list if isinstance(magic, list): magic = magic else: magic = [magic] # initialise group group = group_h5( dst, "data", status="a", overwrite=grp_overwrite, comm=comm, rank=rank, size=size, ) for key in magic: if is_vector(key): dataset_h5( group, key + str(nkernel), status=status, shape=[3, mz, my, mx], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype, ) print("writing " + key + " shape {}".format([3, mz, my, mx])) else: dataset_h5( group, key + str(nkernel), status=status, shape=[mz, my, mx], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype, ) print("writing " + key + " shape {}".format([mz, my, mx])) for ichunk in range(allchunks): for iz in [indz[np.mod(ichunk, nchunks[2])]]: if nchunks[2] == 1: zextra = nghost else: zextra = nkernel + nghost n1, n2 = iz[0] - zextra, iz[-1] + zextra + 1 lindz = np.arange(n1, n2) n1out = n1 + zextra n2out = n2 - zextra varn1 = zextra varn2 = -zextra if iz[0] == locindz[0][0]: n1out = 0 varn1 = zextra - nghost if iz[-1] == locindz[-1][-1]: n2out = n2 - zextra + nghost varn2 = n2 - n1 - zextra + nghost if n1 < 0: lindz[np.where(lindz < nghost)[0]] += nz if n2 > mz - 1: lindz[np.where(lindz > mz - 1 - nghost)[0]] -= nz print('n1out {},n2out {},varn1 {},varn2 {},zextra {}'.format( n1out, n2out, varn1, varn2, zextra)) for iy in [ indy[np.mod(ichunk + int(ichunk / nchunks[2]), nchunks[1])] ]: if nchunks[1] == 1: yextra = nghost else: yextra = nkernel + nghost m1, m2 = iy[0] - yextra, iy[-1] + yextra + 1 lindy = np.arange(m1, m2) m1out = m1 + yextra m2out = m2 + 1 - yextra varm1 = yextra varm2 = -yextra if iy[0] == locindy[0][0]: m1out = 0 varm1 = yextra - nghost if iy[-1] == locindy[-1][-1]: m2out = m2 - yextra + nghost varm2 = m2 - m1 - yextra + nghost if m1 < 0: lindy[np.where(lindy < 0)[0]] += ny if m2 > my - 1: lindy[np.where(lindy > my - 1)[0]] -= ny print( 'm1out {},m2out {},varm1 {},varm2 {},yextra {}'.format( m1out, m2out, varm1, varm2, yextra)) for iy in [ indy[np.mod(ichunk + int(ichunk / nchunks[2]), nchunks[1])] ]: for ix in [ indx[np.mod( ichunk + int(ichunk / nchunks[2]) + int(ichunk / nchunks[1]), nchunks[0], )] ]: if nchunks[1] == 1: xextra = nghost else: xextra = nkernel + nghost l1, l2 = ix[0] - xextra, ix[-1] + xextra + 1 lindx = np.arange(l1, l2) l1out = l1 + xextra l2out = l2 + 1 - xextra varl1 = xextra varl2 = -xextra if ix[0] == locindx[0][0]: l1out = 0 varl1 = xextra - nghost if ix[-1] == locindx[-1][-1]: l2out = l2 - xextra + nghost varl2 = l2 - l1 - xextra + nghost if l1 < 0: lindx[np.where(lindx < 0)[0]] += nx if l2 > mx - 1: lindx[np.where(lindx > mx - 1)[0]] -= nx print('l1out {},l2out {},varl1 {},varl2 {},xextra {}'. format(l1out, l2out, varl1, varl2, xextra)) if not quiet: print("remeshing " + key + " chunk {}".format([iz, iy, ix])) print('sending ichunk {} with index ranges {}'.format( ichunk, [n1, n2, m1, m2, l1, l2])) var = smoothed_data(src["data"], dst["data"], key, par, gd, lindx, lindy, lindz, nghost, sigma, typ, mode) print( 'ichunk {}, var min {:.1e}, var max {:.1e}'.format( ichunk, var.min(), var.max())) # print('var shape {}'.format(var.shape)) # if not quiet: # print('writing '+key+ # ' shape {} chunk {}'.format( # var.shape, [iz,iy,ix])) print('ichunk: out indices {}'.format( [n1out, n2out, m1out, m2out, l1out, l2out])) if is_vector(key): dst["data"][key + str(nkernel)][:, n1out:n2out, m1out:m2out, l1out:l2out] = dtype( var[:, varn1:varn2, varm1:varm2, varl1:varl2]) else: dst["data"][key + str(nkernel)][n1out:n2out, m1out:m2out, l1out:l2out] = dtype( var[varn1:varn2, varm1:varm2, varl1:varl2])
def rhs_data(sim_path, src, dst, magic=["uxb","etadel2a"], par=[], comm=None, gd=[], grp_overwrite=False, overwrite=False, rank=0, size=1, nghost=3,status="a", chunksize = 1000.0, dtype=np.float64, quiet=True, nmin=32, Reynolds_shock=False, lmix=False ): if comm: overwrite = False if isinstance(par, list): os.chdir(sim_path) par = read.param(quiet=True,conflicts_quiet=True) if isinstance(gd, list): os.chdir(sim_path) gd = read.grid(quiet=True) #get data dimensions nx, ny, nz = src["settings"]["nx"][0],\ src["settings"]["ny"][0],\ src["settings"]["nz"][0] mx, my, mz = src["settings"]["mx"][0],\ src["settings"]["my"][0],\ src["settings"]["mz"][0] #split data into manageable memory chunks dstchunksize = 8*nx*ny*nz/1024*1024 if dstchunksize > chunksize: nchunks = cpu_optimal(nx,ny,nz,quiet=quiet, mvar=src["settings/mvar"][0], maux=src["settings/maux"][0], MBmin=chunksize,nmin=nmin,size=size)[1] else: nchunks = [1,1,1] print("nchunks {}".format(nchunks)) # for mpi split chunks across processes if size > 1: locindx = np.array_split(np.arange(nx)+nghost,nchunks[0]) locindy = np.array_split(np.arange(ny)+nghost,nchunks[1]) locindz = np.array_split(np.arange(nz)+nghost,nchunks[2]) indx = [locindx[np.mod(rank+int(rank/nchunks[2]) +int(rank/nchunks[1]),nchunks[0])]] indy = [locindy[np.mod(rank+int(rank/nchunks[2]),nchunks[1])]] indz = [locindz[np.mod(rank,nchunks[2])]] allchunks = 1 else: locindx = np.array_split(np.arange(nx)+nghost,nchunks[0]) locindy = np.array_split(np.arange(ny)+nghost,nchunks[1]) locindz = np.array_split(np.arange(nz)+nghost,nchunks[2]) indx = np.array_split(np.arange(nx)+nghost,nchunks[0]) indy = np.array_split(np.arange(ny)+nghost,nchunks[1]) indz = np.array_split(np.arange(nz)+nghost,nchunks[2]) allchunks = nchunks[0]*nchunks[1]*nchunks[2] # save time dataset_h5(dst, "time", status=status, data=src["time"][()], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype) # ensure derived variables are in a list if isinstance(magic, list): magic = magic else: magic = [magic] # confirm exists group group_h5(dst, "data", status="a", overwrite=grp_overwrite, comm=comm, rank=rank, size=size) # initialise group group = group_h5(dst, "calc", status="a", overwrite=grp_overwrite, comm=comm, rank=rank, size=size) for key in magic: if is_vector(key): dataset_h5(group, key, status=status, shape=[3,mz,my,mx], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype) print("writing "+key+" shape {}".format([3,mz,my,mx])) else: dataset_h5(group, key, status=status, shape=[mz,my,mx], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype) print("writing "+key+" shape {}".format([mz,my,mx])) for ichunk in range(allchunks): for iz in [indz[np.mod(ichunk,nchunks[2])]]: n1, n2 = iz[ 0]-nghost,\ iz[-1]+nghost+1 n1out = n1+nghost n2out = n2-nghost varn1 = nghost varn2 = -nghost if iz[0] == locindz[0][0]: n1out = 0 varn1 = 0 if iz[-1] == locindz[-1][-1]: n2out = n2 varn2 = n2 for iy in [indy[np.mod(ichunk+ int(ichunk/nchunks[2]),nchunks[1])]]: m1, m2 = iy[ 0]-nghost,\ iy[-1]+nghost+1 m1out = m1+nghost m2out = m2-nghost varm1 = nghost varm2 = -nghost if iy[0] == locindy[0][0]: m1out = 0 varm1 = 0 if iy[-1] == locindy[-1][-1]: m2out = m2 varm2 = m2 for ix in [indx[np.mod(ichunk+int(ichunk/nchunks[2]) +int(ichunk/nchunks[1]),nchunks[0])]]: l1, l2 = ix[ 0]-nghost,\ ix[-1]+nghost+1 l1out = l1+nghost l2out = l2-nghost varl1 = nghost varl2 = -nghost if ix[0] == locindx[0][0]: l1out = 0 varl1 = 0 if ix[-1] == locindx[-1][-1]: l2out = l2 varl2 = l2 if not quiet: print("remeshing "+key+" chunk {}".format( [iz,iy,ix])) var = calc_rhs_data(src, dst, key, par, gd, l1, l2, m1, m2, n1, n2, nghost=nghost, Reynolds_shock=Reynolds_shock, lmix=lmix) if is_vector(key): dst["calc"][key][:,n1out:n2out, m1out:m2out, l1out:l2out] = dtype(var[:, varn1:varn2, varm1:varm2, varl1:varl2]) else: dst["calc"][key][n1out:n2out, m1out:m2out, l1out:l2out] = dtype(var[ varn1:varn2, varm1:varm2, varl1:varl2])
def write_h5_snapshot( snapshot, file_name="VAR0", datadir="data/allprocs", precision="d", nghost=3, persist=None, settings=None, param=None, grid=None, lghosts=False, indx=None, proc=None, ipx=None, ipy=None, ipz=None, procdim=None, unit=None, t=None, x=None, y=None, z=None, state="a", quiet=True, lshear=False, driver=None, comm=None, overwrite=False, rank=0, size=1, ): """ Write a snapshot given as numpy array. We assume by default that a run simulation directory has already been constructed and start completed successfully in h5 format so that files dim, grid and param files are already present. If not the contents of these will need to be supplied as dictionaries along with persist if included. call signature: write_h5_snapshot(snapshot, file_name='VAR0', datadir='data/allprocs', precision='d', nghost=3, persist=None, settings=None, param=None, grid=None, lghosts=False, indx=None, unit=None, t=None, x=None, y=None, z=None, procdim=None, quiet=True, lshear=False, driver=None, comm=None) Keyword arguments: *snapshot*: Numpy array containing the snapshot. Must be of shape [nvar, nz, ny, nx] without boundaries or. Must be of shape [nvar, mz, my, mx] with boundaries for lghosts=True. *file_name*: Name of the snapshot file to be written, e.g. VAR0 or var. *datadir*: Directory where the data is stored. *precision*: Single 'f' or double 'd' precision. *persist*: optional dictionary of persistent variable. *settings*: optional dictionary of persistent variable. *param*: optional Param object. *grid*: optional Pencil Grid object of grid parameters. *nghost*: Number of ghost zones. *lghosts*: If True the snapshot contains the ghost zones. *indx* Index object of index for each variable in f-array *unit*: Optional dictionary of simulation units. *quiet*: Option to print output. *t*: Time of the snapshot. *xyz*: xyz arrays of the domain with ghost zones. This will normally be obtained from Grid object, but facility to redefine an alternative grid value. *lshear*: Flag for the shear. *driver* File driver for hdf5 io for use in serial or MPI parallel. *comm* MPI objects supplied if driver is 'mpio'. *overwrite* flag to replace existing h5 snapshot file. *rank* rank of process with root=0. """ import numpy as np from os.path import join from pencil import read from pencil.io import open_h5, group_h5, dataset_h5 from pencil import is_sim_dir # test if simulation directory if not is_sim_dir(): print("ERROR: Directory needs to be a simulation") sys.stdout.flush() if indx == None: indx = read.index() # if settings == None: settings = {} skeys = [ "l1", "l2", "m1", "m2", "n1", "n2", "nx", "ny", "nz", "mx", "my", "mz", "nprocx", "nprocy", "nprocz", "maux", "mglobal", "mvar", "precision", ] dim = read.dim() for key in skeys: settings[key] = dim.__getattribute__(key) settings["precision"] = precision.encode() settings["nghost"] = nghost settings["version"] = np.int32(0) nprocs = settings["nprocx"] * settings["nprocy"] * settings["nprocz"] gkeys = [ "x", "y", "z", "Lx", "Ly", "Lz", "dx", "dy", "dz", "dx_1", "dy_1", "dz_1", "dx_tilde", "dy_tilde", "dz_tilde", ] if grid == None: grid = read.grid(quiet=True) else: gd_err = False for key in gkeys: if not key in grid.__dict__.keys(): print("ERROR: key " + key + " missing from grid") sys.stdout.flush() gd_err = True if gd_err: print("ERROR: grid incomplete") sys.stdout.flush() ukeys = [ "length", "velocity", "density", "magnetic", "time", "temperature", "flux", "energy", "mass", "system", ] if param == None: param = read.param(quiet=True) param.__setattr__("unit_mass", param.unit_density * param.unit_length**3) param.__setattr__("unit_energy", param.unit_mass * param.unit_velocity**2) param.__setattr__("unit_time", param.unit_length / param.unit_velocity) param.__setattr__("unit_flux", param.unit_mass / param.unit_time**3) param.unit_system = param.unit_system.encode() # check whether the snapshot matches the simulation shape if lghosts: try: snapshot.shape[0] == settings["mvar"] snapshot.shape[1] == settings["mx"] snapshot.shape[2] == settings["my"] snapshot.shape[3] == settings["mz"] except ValueError: print("ERROR: snapshot shape {} ".format(snapshot.shape) + "does not match simulation dimensions with ghosts.") sys.stdout.flush() else: try: snapshot.shape[0] == settings["mvar"] snapshot.shape[1] == settings["nx"] snapshot.shape[2] == settings["ny"] snapshot.shape[3] == settings["nz"] except ValueError: print("ERROR: snapshot shape {} ".format(snapshot.shape) + "does not match simulation dimensions without ghosts.") sys.stdout.flush() # Determine the precision used and ensure snapshot has correct data_type. if precision == "f": data_type = np.float32 snapshot = np.float32(snapshot) elif precision == "d": data_type = np.float64 snapshot = np.float64(snapshot) else: print("ERROR: Precision {0} not understood.".format(precision) + " Must be either 'f' or 'd'") sys.stdout.flush() return -1 # Check that the shape does not conflict with the proc numbers. if ((settings["nx"] % settings["nprocx"] > 0) or (settings["ny"] % settings["nprocy"] > 0) or (settings["nz"] % settings["nprocz"] > 0)): print("ERROR: Shape of the input array is not compatible with the " + "cpu layout. Make sure that nproci devides ni.") sys.stdout.flush() return -1 # Check the shape of the xyz arrays if specified and overwrite grid values. if x != None: if len(x) != settings["mx"]: print("ERROR: x array is incompatible with the shape of snapshot.") sys.stdout.flush() return -1 grid.x = data_type(x) if y != None: if len(y) != settings["my"]: print("ERROR: y array is incompatible with the shape of snapshot.") sys.stdout.flush() return -1 grid.y = data_type(y) if z != None: if len(z) != settings["mz"]: print("ERROR: z array is incompatible with the shape of snapshot.") sys.stdout.flush() return -1 grid.z = data_type(z) # Define a time. if t is None: t = data_type(0.0) # making use of pc_hdf5 functionality: if not proc == None: state = "a" else: state = "w" filename = join(datadir, file_name) print("write_h5_snapshot: filename =", filename) with open_h5( filename, state, driver=driver, comm=comm, overwrite=overwrite, rank=rank, size=size, ) as ds: data_grp = group_h5( ds, "data", status=state, delete=False, overwrite=overwrite, rank=rank, size=size, ) if not procdim: for key in indx.__dict__.keys(): if key in ["uu", "keys", "aa", "KR_Frad", "uun", "gg", "bb"]: continue #create ghost zones if required if not lghosts: tmp_arr = np.zeros([ snapshot.shape[1] + 2 * nghost, snapshot.shape[2] + 2 * nghost, snapshot.shape[3] + 2 * nghost, ]) tmp_arr[dim.n1:dim.n2 + 1, dim.m1:dim.m2 + 1, dim.l1:dim.l2 + 1] = np.array( snapshot[indx.__getattribute__(key) - 1]) dataset_h5( data_grp, key, status=state, data=tmp_arr, dtype=data_type, overwrite=overwrite, rank=rank, comm=comm, size=size, ) else: dataset_h5( data_grp, key, status=state, data=np.array(snapshot[indx.__getattribute__(key) - 1]), dtype=data_type, overwrite=overwrite, rank=rank, comm=comm, size=size, ) else: for key in indx.__dict__.keys(): if key in ["uu", "keys", "aa", "KR_Frad", "uun", "gg", "bb"]: continue dataset_h5( data_grp, key, status=state, shape=(settings["mz"], settings["my"], settings["mx"]), dtype=data_type, rank=rank, comm=comm, size=size, ) # adjust indices to include ghost zones at boundaries l1, m1, n1 = procdim.l1, procdim.m1, procdim.n1 if procdim.ipx == 0: l1 = 0 if procdim.ipy == 0: m1 = 0 if procdim.ipz == 0: n1 = 0 l2, m2, n2 = procdim.l2, procdim.m2, procdim.n2 if procdim.ipx == settings["nprocx"] - 1: l2 = procdim.l2 + settings["nghost"] if procdim.ipy == settings["nprocy"] - 1: m2 = procdim.m2 + settings["nghost"] if procdim.ipz == settings["nprocz"] - 1: n2 = procdim.n2 + settings["nghost"] nx, ny, nz = procdim.nx, procdim.ny, procdim.nz ipx, ipy, ipz = procdim.ipx, procdim.ipy, procdim.ipz for key in indx.__dict__.keys(): if key in ["uu", "keys", "aa", "KR_Frad", "uun", "gg", "bb"]: continue tmp_arr = np.array(snapshot[indx.__getattribute__(key) - 1]) data_grp[key][n1 + ipz * nz:n2 + ipz * nz + 1, m1 + ipy * ny:m2 + ipy * ny + 1, l1 + ipx * nx:l2 + ipx * nx + 1, ] = tmp_arr[n1:n2 + 1, m1:m2 + 1, l1:l2 + 1] dataset_h5( ds, "time", status=state, data=np.array(t), size=size, dtype=data_type, rank=rank, comm=comm, overwrite=overwrite, ) # add settings sets_grp = group_h5( ds, "settings", status=state, delete=False, overwrite=overwrite, rank=rank, size=size, ) for key in settings.keys(): if "precision" in key: dataset_h5( sets_grp, key, status=state, data=(settings[key], ), dtype=None, rank=rank, comm=comm, size=size, overwrite=overwrite, ) else: dataset_h5( sets_grp, key, status=state, data=(settings[key], ), dtype=data_type, rank=rank, comm=comm, size=size, overwrite=overwrite, ) # add grid grid_grp = group_h5( ds, "grid", status=state, delete=False, overwrite=overwrite, rank=rank, size=size, ) for key in gkeys: dataset_h5( grid_grp, key, status=state, data=(grid.__getattribute__(key)), dtype=data_type, rank=rank, comm=comm, size=size, overwrite=overwrite, ) dataset_h5( grid_grp, "Ox", status=state, data=(param.__getattribute__("xyz0")[0], ), dtype=data_type, rank=rank, comm=comm, size=size, overwrite=overwrite, ) dataset_h5( grid_grp, "Oy", status=state, data=(param.__getattribute__("xyz0")[1], ), dtype=data_type, rank=rank, comm=comm, size=size, overwrite=overwrite, ) dataset_h5( grid_grp, "Oz", status=state, data=(param.__getattribute__("xyz0")[2], ), dtype=data_type, rank=rank, comm=comm, size=size, overwrite=overwrite, ) # add physical units unit_grp = group_h5( ds, "unit", status=state, delete=False, overwrite=overwrite, rank=rank, size=size, ) for key in ukeys: if "system" in key: dataset_h5( unit_grp, key, status=state, data=(param.__getattribute__("unit_" + key), ), rank=rank, comm=comm, size=size, overwrite=overwrite, ) else: dataset_h5( unit_grp, key, status=state, data=param.__getattribute__("unit_" + key), rank=rank, comm=comm, size=size, overwrite=overwrite, ) # add optional persistent data if persist != None: pers_grp = group_h5( ds, "persist", status=state, size=size, delete=False, overwrite=overwrite, rank=rank, ) for key in persist.keys(): if not quiet: print(key, type(persist[key][()])) sys.stdout.flush() arr = np.empty(nprocs, dtype=type(persist[key][()])) arr[:] = persist[key][()] dataset_h5( pers_grp, key, status=state, data=(arr), size=size, dtype=data_type, rank=rank, comm=comm, overwrite=overwrite, )
def write_h5_grid( file_name="grid", datadir="data", precision="d", nghost=3, settings=None, param=None, grid=None, unit=None, quiet=True, driver=None, comm=None, overwrite=False, rank=0, ): """ Write the grid information as hdf5. We assume by default that a run simulation directory has already been constructed, but start has not been executed in h5 format so that binary sim files dim, grid and param files are already present in the sim directory, or provided from an old binary sim source directory as inputs. call signature: write_h5_grid(file_name='grid', datadir='data', precision='d', nghost=3, settings=None, param=None, grid=None, unit=None, quiet=True, driver=None, comm=None) Keyword arguments: *file_name*: Prefix of the file name to be written, 'grid'. *datadir*: Directory where 'grid.h5' is stored. *precision*: Single 'f' or double 'd' precision. *nghost*: Number of ghost zones. *settings*: Optional dictionary of persistent variable. *param*: Optional Param object. *grid*: Optional Pencil Grid object of grid parameters. *unit*: Optional dictionary of simulation units. *quiet*: Option to print output. """ from os.path import join import numpy as np from pencil import read from pencil.io import open_h5, group_h5, dataset_h5 from pencil import is_sim_dir # test if simulation directory if not is_sim_dir(): print("ERROR: Directory needs to be a simulation") sys.stdout.flush() # if settings == None: settings = {} skeys = [ "l1", "l2", "m1", "m2", "n1", "n2", "nx", "ny", "nz", "mx", "my", "mz", "nprocx", "nprocy", "nprocz", "maux", "mglobal", "mvar", "precision", ] dim = read.dim() for key in skeys: settings[key] = dim.__getattribute__(key) settings["precision"] = precision.encode() settings["nghost"] = nghost settings["version"] = np.int32(0) gkeys = [ "x", "y", "z", "Lx", "Ly", "Lz", "dx", "dy", "dz", "dx_1", "dy_1", "dz_1", "dx_tilde", "dy_tilde", "dz_tilde", ] if grid == None: grid = read.grid(quiet=True) else: gd_err = False for key in gkeys: if not key in grid.__dict__.keys(): print("ERROR: key " + key + " missing from grid") sys.stdout.flush() gd_err = True if gd_err: print("ERROR: grid incomplete") sys.stdout.flush() ukeys = [ "length", "velocity", "density", "magnetic", "time", "temperature", "flux", "energy", "mass", "system", ] if param == None: param = read.param(quiet=True) param.__setattr__("unit_mass", param.unit_density * param.unit_length**3) param.__setattr__("unit_energy", param.unit_mass * param.unit_velocity**2) param.__setattr__("unit_time", param.unit_length / param.unit_velocity) param.__setattr__("unit_flux", param.unit_mass / param.unit_time**3) param.unit_system = param.unit_system.encode() # open file for writing data filename = join(datadir, file_name + ".h5") with open_h5(filename, "w", driver=driver, comm=comm, overwrite=overwrite, rank=rank) as ds: # add settings sets_grp = group_h5(ds, "settings", status="w") for key in settings.keys(): if "precision" in key: dataset_h5(sets_grp, key, status="w", data=(settings[key], )) else: dataset_h5(sets_grp, key, status="w", data=(settings[key], )) # add grid grid_grp = group_h5(ds, "grid", status="w") for key in gkeys: dataset_h5(grid_grp, key, status="w", data=(grid.__getattribute__(key))) dataset_h5(grid_grp, "Ox", status="w", data=(param.__getattribute__("xyz0")[0], )) dataset_h5(grid_grp, "Oy", status="w", data=(param.__getattribute__("xyz0")[1], )) dataset_h5(grid_grp, "Oz", status="w", data=(param.__getattribute__("xyz0")[2], )) # add physical units unit_grp = group_h5(ds, "unit", status="w") for key in ukeys: if "system" in key: dataset_h5( unit_grp, key, status="w", data=(param.__getattribute__("unit_" + key), ), ) else: dataset_h5( unit_grp, key, status="w", data=param.__getattribute__("unit_" + key), )
def sim2h5( newdir=".", olddir=".", varfile_names=None, todatadir="data/allprocs", fromdatadir="data", precision="d", nghost=3, lpersist=True, x=None, y=None, z=None, lshear=False, snap_by_proc=False, aver_by_proc=False, lremove_old_snapshots=False, lremove_old_slices=False, lread_all_videoslices=False, vlarge=100000000, lremove_old_averages=False, execute=False, quiet=True, l2D=True, lvars=True, lvids=True, laver=True, laver2D=False, lremove_deprecated_vids=False, lsplit_slices=False, ): """ Copy a simulation object written in Fortran binary to hdf5. The default is to copy all snapshots from/to the current simulation directory. Optionally the old files can be removed to call signature: sim2h5(newdir='.', olddir='.', varfile_names=None, todatadir='data/allprocs', fromdatadir='data', precision='d', nghost=3, lpersist=False, x=None, y=None, z=None, lshear=False, snap_by_proc=False, aver_by_proc=False, lremove_old_snapshots=False, lremove_old_slices=False, lread_all_videoslices=True, lremove_old_averages=False, execute=False, quiet=True, l2D=True, lvars=True, lvids=True, laver=True) Keyword arguments: *olddir*: String path to simulation source directory. Path may be relative or absolute. *newdir*: String path to simulation destination directory. Path may be relative or absolute. *varfile_names*: A list of names of the snapshot files to be written, e.g. VAR0 If None all varfiles in olddir+'/data/proc0/' will be converted *todatadir*: Directory to which the data is stored. *fromdatadir*: Directory from which the data is collected. *precision*: Single 'f' or double 'd' precision for new data. *nghost*: Number of ghost zones. TODO: handle switching size of ghost zones. *lpersist*: option to include persistent variables from snapshots. *xyz*: xyz arrays of the domain with ghost zones. This will normally be obtained from Grid object, but facility to redefine an alternative grid value. *lshear*: Flag for the shear. *execute*: optional confirmation required if lremove_old. *lremove_old_snapshots*: If True the old snapshot data will be deleted once the new h5 data has been saved. *lremove_old_slices*: If True the old video slice data will be deleted once the new h5 data has been saved. *lremove_old_averages*: If True the old averages data will be deleted once the new h5 data has been saved. *aver_by_proc* Option to read old binary files by processor and write in parallel *laver2D* If True apply to each plane_list 'y', 'z' and load each variable sequentially *l_mpi*: Applying MPI parallel process *driver*: HDF5 file io driver either None or mpio *comm*: MPI library calls *rank*: Integer ID of processor *size*: Number of MPI processes """ import glob import numpy as np import os from os.path import exists, join import subprocess as sub import sys from .. import read from .. import sim from . import write_h5_grid from pencil.util import is_sim_dir try: from mpi4py import MPI comm = MPI.COMM_WORLD rank = comm.Get_rank() size = comm.Get_size() driver = "mpio" l_mpi = True l_mpi = l_mpi and (size != 1) except ImportError: comm = None driver = None rank = 0 size = 1 l_mpi = False if not l_mpi: comm = None driver = None print("rank {} and size {}".format(rank, size)) sys.stdout.flush() if rank == size - 1: print("l_mpi", l_mpi) sys.stdout.flush() # test if simulation directories if newdir == ".": newdir = os.getcwd() if olddir == ".": olddir = os.getcwd() os.chdir(olddir) if not is_sim_dir(): if rank == 0: print("ERROR: Directory (" + olddir + ") needs to be a simulation") sys.stdout.flush() return -1 if newdir != olddir: if not exists(newdir): cmd = "pc_newrun -s " + newdir if rank == size - 1: process = sub.Popen(cmd.split(), stdout=sub.PIPE) output, error = process.communicate() print(cmd, output, error) # os.system(cmd) if comm: comm.Barrier() os.chdir(newdir) if not is_sim_dir(): if rank == 0: print("ERROR: Directory (" + newdir + ") needs to be a simulation") sys.stdout.flush() return -1 # lremove_old = lremove_old_snapshots or lremove_old_slices or lremove_old_averages if lremove_old: if not execute: os.chdir(olddir) if rank == 0: print("WARNING: Are you sure you wish to remove the Fortran" + " binary files from \n" + os.getcwd() + ".\n" + "Set execute=True to proceed.") sys.stdout.flush() return -1 os.chdir(olddir) if lvars: if varfile_names == None: os.chdir(fromdatadir + "/proc0") lVARd = False varfiled_names = natural_sort(glob.glob("VARd*")) if len(varfiled_names) > 0: varfile_names = natural_sort(glob.glob("VAR*")) for iv in range(len(varfile_names) - 1, -1, -1): if "VARd" in varfile_names[iv]: varfile_names.remove(varfile_names[iv]) lVARd = True else: varfile_names = natural_sort(glob.glob("VAR*")) os.chdir(olddir) else: lVARd = False if isinstance(varfile_names, list): varfile_names = varfile_names else: varfile_names = [varfile_names] varfiled_names = [] tmp_names = [] for varfile_name in varfile_names: if "VARd" in varfile_names: varfiled_names.append(varfile_name) lVARd = True else: tmp_names.append(varfile_name) varfile_names = tmp_names gkeys = [ "x", "y", "z", "Lx", "Ly", "Lz", "dx", "dy", "dz", "dx_1", "dy_1", "dz_1", "dx_tilde", "dy_tilde", "dz_tilde", ] grid = None if rank == size - 1: grid = read.grid(quiet=True) if l_mpi: grid = comm.bcast(grid, root=size - 1) if not quiet: print(rank, grid) sys.stdout.flush() for key in gkeys: if not key in grid.__dict__.keys(): if rank == 0: print("ERROR: key " + key + " missing from grid") sys.stdout.flush() return -1 # obtain the settings from the old simulation settings = {} skeys = [ "l1", "l2", "m1", "m2", "n1", "n2", "nx", "ny", "nz", "mx", "my", "mz", "nprocx", "nprocy", "nprocz", "maux", "mglobal", "mvar", "precision", ] if rank == 0: olddim = read.dim() for key in skeys: settings[key] = np.array(olddim.__getattribute__(key)) olddim = None settings["nghost"] = np.array(nghost) settings["precision"] = precision.encode() if l_mpi: settings = comm.bcast(settings, root=0) if snap_by_proc: nprocs = settings["nprocx"] * settings["nprocy"] * settings["nprocz"] if np.mod(nprocs, size) != 0: print("WARNING: efficiency requires cpus to divide ncpus") sys.stdout.flush() if not quiet: print(rank, grid) sys.stdout.flush() # obtain physical units from old simulation ukeys = [ "length", "velocity", "density", "magnetic", "time", "temperature", "flux", "energy", "mass", "system", ] param = read.param(quiet=True) param.__setattr__("unit_mass", param.unit_density * param.unit_length**3) param.__setattr__("unit_energy", param.unit_mass * param.unit_velocity**2) param.__setattr__("unit_time", param.unit_length / param.unit_velocity) param.__setattr__("unit_flux", param.unit_mass / param.unit_time**3) param.unit_system = param.unit_system.encode() # index list for variables in f-array if not quiet: print(rank, param) sys.stdout.flush() indx = None if rank == 0: indx = read.index() if l_mpi: indx = comm.bcast(indx, root=0) # check consistency between Fortran binary and h5 data os.chdir(newdir) dim = None if is_sim_dir(): if rank == size - 1: if exists(join(newdir, "data", "dim.dat")): try: dim = read.dim() except ValueError: pass if l_mpi: dim = comm.bcast(dim, root=size - 1) if dim: if not quiet: print(rank, dim) sys.stdout.flush() try: dim.mvar == settings["mvar"] dim.mx == settings["mx"] dim.my == settings["my"] dim.mz == settings["mz"] except ValueError: if rank == size - 1: print("ERROR: new simulation dimensions do not match.") sys.stdout.flush() return -1 dim = None os.chdir(olddir) if rank == size - 1: print("precision is ", precision) sys.stdout.flush() if laver2D: aver2h5( newdir, olddir, todatadir="data/averages", fromdatadir="data", l2D=False, precision=precision, quiet=quiet, laver2D=laver2D, lremove_old_averages=False, aver_by_proc=aver_by_proc, l_mpi=l_mpi, driver=driver, comm=comm, rank=rank, size=size, ) l2D = False # copy snapshots if lvars and len(varfile_names) > 0: var2h5( newdir, olddir, varfile_names, todatadir, fromdatadir, snap_by_proc, precision, lpersist, quiet, nghost, settings, param, grid, x, y, z, lshear, lremove_old_snapshots, indx, l_mpi=l_mpi, driver=driver, comm=comm, rank=rank, size=size, ) # copy downsampled snapshots if present if lvars and lVARd: var2h5( newdir, olddir, varfiled_names, todatadir, fromdatadir, False, precision, lpersist, quiet, nghost, settings, param, grid, x, y, z, lshear, lremove_old_snapshots, indx, trimall=True, l_mpi=l_mpi, driver=driver, comm=comm, rank=rank, size=size, ) if lvars: var2h5( newdir, olddir, [ "var.dat", ], todatadir, fromdatadir, snap_by_proc, precision, lpersist, quiet, nghost, settings, param, grid, x, y, z, lshear, lremove_old_snapshots, indx, l_mpi=l_mpi, driver=driver, comm=comm, rank=rank, size=size, ) # copy old video slices to new h5 sim if lvids: if lremove_deprecated_vids: for ext in [ "bb.", "uu.", "ux.", "uy.", "uz.", "bx.", "by.", "bz." ]: cmd = "rm -f " + join(olddir, fromdatadir, "proc*", "slice_" + ext + "*") if rank == 0: process = sub.Popen(cmd.split(), stdout=sub.PIPE) output, error = process.communicate() print(cmd, output, error) cmd = "rm -f " + join(fromdatadir, "slice_" + ext + "*") if rank == 0: process = sub.Popen(cmd.split(), stdout=sub.PIPE) output, error = process.communicate() print(cmd, output, error) if comm: comm.Barrier() cmd = "src/read_all_videofiles.x" if rank == size - 1 and lread_all_videoslices: process = sub.Popen(cmd.split(), stdout=sub.PIPE) output, error = process.communicate() print(cmd, output, error) if comm: comm.Barrier() slices2h5( newdir, olddir, grid, todatadir="data/slices", fromdatadir=fromdatadir, precision=precision, quiet=quiet, vlarge=vlarge, lsplit_slices=lsplit_slices, lremove_old_slices=lremove_old_slices, l_mpi=l_mpi, driver=driver, comm=comm, rank=rank, size=size, ) # copy old averages data to new h5 sim if laver: aver2h5( newdir, olddir, todatadir="data/averages", fromdatadir=fromdatadir, l2D=l2D, precision=precision, quiet=quiet, aver_by_proc=False, lremove_old_averages=lremove_old_averages, l_mpi=l_mpi, driver=driver, comm=comm, rank=rank, size=size, ) # check some critical sim files are present for new sim without start # construct grid.h5 sim information if requied for new h5 sim os.chdir(newdir) if l_mpi: comm.Barrier() if rank == 0: write_h5_grid( file_name="grid", datadir="data", precision=precision, nghost=nghost, settings=settings, param=param, grid=grid, unit=None, quiet=quiet, ) source_file = join(olddir, fromdatadir, "proc0/varN.list") target_file = join(newdir, todatadir, "varN.list") if exists(source_file): cmd = "cp " + source_file + " " + target_file process = sub.Popen(cmd.split(), stdout=sub.PIPE) output, error = process.communicate() print(cmd, output, error) items = [ "def_var.pro", "index.pro", "jobid.dat", "param.nml", "particle_index.pro", "pc_constants.pro", "pointmass_index.pro", "pt_positions.dat", "sn_series.dat", "svnid.dat", "time_series.dat", "tsnap.dat", "tspec.dat", "tvid.dat", "t2davg.dat", "var.general", "variables.pro", "varname.dat", ] for item in items: source_file = join(olddir, fromdatadir, item) target_file = join(newdir, fromdatadir, item) if exists(source_file): if not exists(target_file): cmd = "cp " + source_file + " " + target_file process = sub.Popen(cmd.split(), stdout=sub.PIPE) output, error = process.communicate() print(cmd, output, error) print("Simulation Fortran to h5 completed on rank {}.".format(rank)) sys.stdout.flush()
def find_tracers(self, var_file='VAR0', datadir='data', trace_field='bb', ti=-1, tf=-1): """ Trace streamlines of the vectofield 'field' from z = z0 to z = z1 and integrate quantities 'int_q' along the lines. Creates a 2d mapping as in 'streamlines.f90'. call signature: find_tracers(var_file='VAR0', datadir='data', trace_field='bb', ti=-1, tf=-1) Keyword arguments: *var_file*: Varfile to be read. *datadir*: Directory where the data is stored. *trace_field*: Vector field used for the streamline tracing. *ti*: Initial VAR file index for tracer time sequences. Overrides 'var_file'. *tf*: Final VAR file index for tracer time sequences. Overrides 'var_file'. """ import numpy as np import multiprocessing as mp from pencil import read from pencil import math # Write the tracing parameters. self.params.trace_field = trace_field self.params.datadir = datadir # Multi core setup. if not(np.isscalar(self.params.n_proc)) or (self.params.n_proc%1 != 0): print("error: invalid processor number") return -1 queue = mp.Queue() # Read the data. magic = [] if trace_field == 'bb': magic.append('bb') if trace_field == 'jj': magic.append('jj') if trace_field == 'vort': magic.append('vort') if self.params.int_q == 'ee': magic.append('bb') magic.append('jj') dim = read.dim(datadir=datadir) self.params.var_file = var_file # Check if user wants a tracer time series. if (ti%1 == 0) and (tf%1 == 0) and (ti >= 0) and (tf >= ti): series = True nTimes = tf-ti+1 else: series = False nTimes = 1 # Initialize the arrays. self.x0 = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes]) self.y0 = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes]) self.x1 = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes]) self.y1 = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes]) self.z1 = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes]) self.l = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes]) if self.params.int_q == 'curly_A': self.curly_A = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes]) if self.params.int_q == 'ee': self.ee = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes]) self.mapping = np.zeros([int(self.params.trace_sub*dim.nx), int(self.params.trace_sub*dim.ny), nTimes, 3]) self.t = np.zeros(nTimes) for t_idx in range(ti, tf+1): if series: var_file = 'VAR' + str(t_idx) # Read the data. var = read.var(var_file=var_file, datadir=datadir, magic=magic, quiet=True, trimall=True) grid = read.grid(datadir=datadir, quiet=True, trim=True) param2 = read.param(datadir=datadir, quiet=True) self.t[t_idx] = var.t # Extract the requested vector trace_field. field = getattr(var, trace_field) if self.params.int_q == 'curly_A': self.aa = var.aa if self.params.int_q == 'ee': self.ee = var.jj*param2.eta - math.cross(var.uu, var.bb) # Get the simulation parameters. self.params.dx = var.dx self.params.dy = var.dy self.params.dz = var.dz self.params.Ox = var.x[0] self.params.Oy = var.y[0] self.params.Oz = var.z[0] self.params.Lx = grid.Lx self.params.Ly = grid.Ly self.params.Lz = grid.Lz self.params.nx = dim.nx self.params.ny = dim.ny self.params.nz = dim.nz # Initialize the tracers. for ix in range(int(self.params.trace_sub*dim.nx)): for iy in range(int(self.params.trace_sub*dim.ny)): self.x0[ix, iy, t_idx] = grid.x[0] + grid.dx/self.params.trace_sub*ix self.x1[ix, iy, t_idx] = self.x0[ix, iy, t_idx].copy() self.y0[ix, iy, t_idx] = grid.y[0] + grid.dy/self.params.trace_sub*iy self.y1[ix, iy, t_idx] = self.y0[ix, iy, t_idx].copy() self.z1[ix, iy, t_idx] = grid.z[0] proc = [] sub_data = [] for i_proc in range(self.params.n_proc): proc.append(mp.Process(target=self.__sub_tracers, args=(queue, field, t_idx, i_proc, self.params.n_proc))) for i_proc in range(self.params.n_proc): proc[i_proc].start() for i_proc in range(self.params.n_proc): sub_data.append(queue.get()) for i_proc in range(self.params.n_proc): proc[i_proc].join() for i_proc in range(self.params.n_proc): # Extract the data from the single cores. Mind the order. sub_proc = sub_data[i_proc][0] self.x1[sub_proc::self.params.n_proc, :, t_idx] = sub_data[i_proc][1] self.y1[sub_proc::self.params.n_proc, :, t_idx] = sub_data[i_proc][2] self.z1[sub_proc::self.params.n_proc, :, t_idx] = sub_data[i_proc][3] self.l[sub_proc::self.params.n_proc, :, t_idx] = sub_data[i_proc][4] self.mapping[sub_proc::self.params.n_proc, :, t_idx, :] = sub_data[i_proc][5] if self.params.int_q == 'curly_A': self.curly_A[sub_proc::self.params.n_proc, :, t_idx] = sub_data[i_proc][6] if self.params.int_q == 'ee': self.ee[sub_proc::self.params.n_proc, :, t_idx] = sub_data[i_proc][7] for i_proc in range(self.params.n_proc): proc[i_proc].terminate() return 0
def slices2vtk(field='', extension='', datadir='data', destination='slices', proc=-1): """ Convert slices from PencilCode format to vtk. call signature:: slices2vtk(field='', extension='', datadir='data', destination='slices', proc=-1) Read slice files specified by *variables* and convert them into vtk format for the specified extensions. Write the result in *destination*. NB: You need to have called src/read_videofiles.x before using this script. Keyword arguments: *field*: All allowed fields which can be written as slice files, e.g. b2, uu1, lnrho, ... See the pencil code manual for more (chapter: "List of parameters for `video.in'"). *extension*: List of slice positions. *datadir*: Directory where the data is stored. *destination*: Destination files. *proc*: Processor which should be read. Set to -1 for all processors. """ import sys import numpy as np from pencil import read # Convert single variable string into length 1 list of arrays. if (len(field) > 0): if (len(field[0]) == 1): field = [field] if (len(extension) > 0): if (len(extension[0]) == 1): extension = [extension] # Read the grid dimensions. grid = read.grid(datadir=datadir, proc=proc, trim=True, quiet=True) # Read the dimensions. dim = read.dim(datadir=datadir, proc=proc) # Read the user given parameters for the slice positions. params = read.param(quiet=True) # Read the slice file for all specified variables and extensions. slices = read.slices(field=field, extension=extension, datadir=datadir, proc=proc) # Determine the position of the slices. if params.ix != -1: x0 = grid.x[params.ix] elif params.slice_position == 'm': x0 = grid.x[int(len(grid.x) / 2)] if params.iy != -1: y0 = grid.y[params.iy] elif params.slice_position == 'm': y0 = grid.y[int(len(grid.y) / 2)] if params.iz != -1: z0 = grid.z[params.iz] elif params.slice_position == 'm': z0 = grid.z[int(len(grid.z) / 2)] if params.iz2 != -1: z02 = grid.z[params.iz] elif params.slice_position == 'm': z02 = grid.z[int(len(grid.z) / 2)] for t_idx, t in enumerate(slices.t): for ext in extension: # Open the destination file for writing. fd = open(destination + '_' + ext + '_' + str(t_idx) + '.vtk', 'wb') # Write the header. fd.write('# vtk DataFile Version 2.0\n'.encode('utf-8')) fd.write('slices {0}\n'.format(ext).encode('utf-8')) fd.write('BINARY\n'.encode('utf-8')) fd.write('DATASET STRUCTURED_POINTS\n'.encode('utf-8')) if ext == 'xy': fd.write('DIMENSIONS {0:9} {1:9} {2:9}\n'.format( dim.nx, dim.ny, 1).encode('utf-8')) fd.write('ORIGIN {0:8.12} {1:8.12} {2:8.12}\n'.format( grid.x[0], grid.y[0], z0).encode('utf-8')) fd.write('SPACING {0:8.12} {1:8.12} {2:8.12}\n'.format( grid.dx, grid.dy, 1.).encode('utf-8')) dim_p = dim.nx dim_q = dim.ny if ext == 'xy2': fd.write('DIMENSIONS {0:9} {1:9} {2:9}\n'.format( dim.nx, dim.ny, 1).encode('utf-8')) fd.write('ORIGIN {0:8.12} {1:8.12} {2:8.12}\n'.format( grid.x[0], grid.y[0], z02).encode('utf-8')) fd.write('SPACING {0:8.12} {1:8.12} {2:8.12}\n'.format( grid.dx, grid.dy, 1.).encode('utf-8')) dim_p = dim.nx dim_q = dim.ny if ext == 'xz': fd.write('DIMENSIONS {0:9} {1:9} {2:9}\n'.format( dim.nx, 1, dim.nz).encode('utf-8')) fd.write('ORIGIN {0:8.12} {1:8.12} {2:8.12}\n'.format( grid.x[0], y0, grid.z[0]).encode('utf-8')) fd.write('SPACING {0:8.12} {1:8.12} {2:8.12}\n'.format( grid.dx, 1., grid.dz).encode('utf-8')) dim_p = dim.nx dim_q = dim.nz if ext == 'yz': fd.write('DIMENSIONS {0:9} {1:9} {2:9}\n'.format( 1, dim.ny, dim.nz).encode('utf-8')) fd.write('ORIGIN {0:8.12} {1:8.12} {2:8.12}\n'.format( x0, grid.y[0], grid.z[0]).encode('utf-8')) fd.write('SPACING {0:8.12} {1:8.12} {2:8.12}\n'.format( 1., grid.dy, grid.dz).encode('utf-8')) dim_p = dim.ny dim_q = dim.nz fd.write('POINT_DATA {0:9}\n'.format(dim_p * dim_q).encode('utf-8')) # Write the data. for fi in field: data = getattr(getattr(slices, ext), fi) fd.write( ('SCALARS ' + ext + '_' + fi + ' float\n').encode('utf-8')) fd.write('LOOKUP_TABLE default\n'.encode('utf-8')) if sys.byteorder == 'little': data = data.astype(np.float32).byteswap() else: data = data.astype(np.float32) fd.write(data[t_idx].tobytes()) fd.close()
def var2vtk(var_file='var.dat', datadir='data', proc=-1, variables=None, b_ext=False, magic=[], destination='work', quiet=True, trimall=True, ti=-1, tf=-1): """ Convert data from PencilCode format to vtk. call signature:: var2vtk(var_file='', datadir='data', proc=-1, variables='', b_ext=False, destination='work', quiet=True, trimall=True, ti=-1, tf=-1) Read *var_file* and convert its content into vtk format. Write the result in *destination*. Keyword arguments: *var_file*: The original var_file. *datadir*: Directory where the data is stored. *proc*: Processor which should be read. Set to -1 for all processors. *variables*: List of variables which should be written. If None all. *b_ext*: Add the external magnetic field. *destination*: Destination file. *quiet*: Keep quiet when reading the var files. *trimall*: Trim the data cube to exclude ghost zones. *ti, tf*: Start and end index for animation. Leave negative for no animation. Overwrites variable var_file. """ import numpy as np import sys from pencil import read from pencil import math # Determine of we want an animation. if ti < 0 or tf < 0: animation = False else: animation = True # If no variables specified collect all by default if not variables: variables = [] indx = read.index() for key in indx.__dict__.keys(): if 'keys' not in key: variables.append(key) if 'uu' in variables: magic.append('vort') variables.append('vort') if 'rho' in variables or 'lnrho' in variables: if 'ss' in variables: magic.append('tt') variables.append('tt') magic.append('pp') variables.append('pp') if 'aa' in variables: magic.append('bb') variables.append('bb') magic.append('jj') variables.append('jj') variables.append('ab') variables.append('b_mag') variables.append('j_mag') else: # Convert single variable string into length 1 list of arrays. if (len(variables) > 0): if (len(variables[0]) == 1): variables = [variables] if 'tt' in variables: magic.append('tt') if 'pp' in variables: magic.append('pp') if 'bb' in variables: magic.append('bb') if 'jj' in variables: magic.append('jj') if 'vort' in variables: magic.append('vort') if 'b_mag' in variables and not 'bb' in magic: magic.append('bb') if 'j_mag' in variables and not 'jj' in magic: magic.append('jj') if 'ab' in variables and not 'bb' in magic: magic.append('bb') for t_idx in range(ti, tf + 1): if animation: var_file = 'VAR' + str(t_idx) # Read the PencilCode variables and set the dimensions. var = read.var(var_file=var_file, datadir=datadir, proc=proc, magic=magic, trimall=True, quiet=quiet) grid = read.grid(datadir=datadir, proc=proc, trim=trimall, quiet=True) params = read.param(quiet=True) # Add external magnetic field. if (b_ext == True): B_ext = np.array(params.b_ext) var.bb[0, ...] += B_ext[0] var.bb[1, ...] += B_ext[1] var.bb[2, ...] += B_ext[2] dimx = len(grid.x) dimy = len(grid.y) dimz = len(grid.z) dim = dimx * dimy * dimz dx = (np.max(grid.x) - np.min(grid.x)) / (dimx - 1) dy = (np.max(grid.y) - np.min(grid.y)) / (dimy - 1) dz = (np.max(grid.z) - np.min(grid.z)) / (dimz - 1) # Write the vtk header. if animation: fd = open(destination + str(t_idx) + '.vtk', 'wb') else: fd = open(destination + '.vtk', 'wb') fd.write('# vtk DataFile Version 2.0\n'.encode('utf-8')) fd.write('VAR files\n'.encode('utf-8')) fd.write('BINARY\n'.encode('utf-8')) fd.write('DATASET STRUCTURED_POINTS\n'.encode('utf-8')) fd.write('DIMENSIONS {0:9} {1:9} {2:9}\n'.format(dimx, dimy, dimz).encode('utf-8')) fd.write('ORIGIN {0:8.12} {1:8.12} {2:8.12}\n'.format( grid.x[0], grid.y[0], grid.z[0]).encode('utf-8')) fd.write('SPACING {0:8.12} {1:8.12} {2:8.12}\n'.format( dx, dy, dz).encode('utf-8')) fd.write('POINT_DATA {0:9}\n'.format(dim).encode('utf-8')) # Write the data. for v in variables: print('Writing {0}.'.format(v)) # Prepare the data to the correct format. if v == 'ab': data = math.dot(var.aa, var.bb) elif v == 'b_mag': data = np.sqrt(math.dot2(var.bb)) elif v == 'j_mag': data = np.sqrt(math.dot2(var.jj)) else: data = getattr(var, v) if sys.byteorder == 'little': data = data.astype(np.float32).byteswap() else: data = data.astype(np.float32) # Check if we have vectors or scalars. if data.ndim == 4: data = np.moveaxis(data, 0, 3) fd.write('VECTORS {0} float\n'.format(v).encode('utf-8')) else: fd.write('SCALARS {0} float\n'.format(v).encode('utf-8')) fd.write('LOOKUP_TABLE default\n'.encode('utf-8')) fd.write(data.tobytes()) del (var) fd.close()
def derive_data(sim_path, src, dst, magic=['pp', 'tt'], par=[], comm=None, gd=[], overwrite=False, rank=0, size=1, nghost=3, status='a', chunksize=1000.0, dtype=np.float64, quiet=True, nmin=32): if comm: overwrite = False if isinstance(par, list): os.chdir(sim_path) par = read.param(quiet=True, conflicts_quiet=True) if isinstance(gd, list): os.chdir(sim_path) gd = read.grid(quiet=True) #get data dimensions nx, ny, nz = src['settings']['nx'][0],\ src['settings']['ny'][0],\ src['settings']['nz'][0] mx, my, mz = src['settings']['mx'][0],\ src['settings']['my'][0],\ src['settings']['mz'][0] #split data into manageable memory chunks dstchunksize = 8 * nx * ny * nz / 1024 * 1024 if dstchunksize > chunksize: nchunks = cpu_optimal(nx, ny, nz, quiet=quiet, mvar=src['settings/mvar'][0], maux=src['settings/maux'][0], MBmin=chunksize, nmin=nmin, size=size)[1] else: nchunks = [1, 1, 1] print('nchunks {}'.format(nchunks)) # for mpi split chunks across processes if size > 1: locindx = np.array_split(np.arange(nx) + nghost, nchunks[0]) locindy = np.array_split(np.arange(ny) + nghost, nchunks[1]) locindz = np.array_split(np.arange(nz) + nghost, nchunks[2]) indx = [ locindx[np.mod( rank + int(rank / nchunks[2]) + int(rank / nchunks[1]), nchunks[0])] ] indy = [locindy[np.mod(rank + int(rank / nchunks[2]), nchunks[1])]] indz = [locindz[np.mod(rank, nchunks[2])]] allchunks = 1 else: locindx = np.array_split(np.arange(nx) + nghost, nchunks[0]) locindy = np.array_split(np.arange(ny) + nghost, nchunks[1]) locindz = np.array_split(np.arange(nz) + nghost, nchunks[2]) indx = np.array_split(np.arange(nx) + nghost, nchunks[0]) indy = np.array_split(np.arange(ny) + nghost, nchunks[1]) indz = np.array_split(np.arange(nz) + nghost, nchunks[2]) allchunks = nchunks[0] * nchunks[1] * nchunks[2] # save time dataset_h5(dst, 'time', status=status, data=src['time'][()], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype) # ensure derived variables are in a list if isinstance(magic, list): magic = magic else: magic = [magic] # initialise group group = group_h5(dst, 'data', status='a', overwrite=overwrite, comm=comm, rank=rank, size=size) for key in magic: if is_vector(key): dataset_h5(group, key, status=status, shape=[3, mz, my, mx], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype) print('writing ' + key + ' shape {}'.format([3, mz, my, mx])) else: dataset_h5(group, key, status=status, shape=[mz, my, mx], comm=comm, size=size, rank=rank, overwrite=overwrite, dtype=dtype) print('writing ' + key + ' shape {}'.format([mz, my, mx])) for ichunk in range(allchunks): for iz in [indz[np.mod(ichunk, nchunks[2])]]: n1, n2 = iz[ 0]-nghost,\ iz[-1]+nghost+1 n1out = n1 + nghost n2out = n2 - nghost varn1 = nghost varn2 = -nghost if iz[0] == locindz[0][0]: n1out = 0 varn1 = 0 if iz[-1] == locindz[-1][-1]: n2out = n2 varn2 = n2 for iy in [ indy[np.mod(ichunk + int(ichunk / nchunks[2]), nchunks[1])] ]: m1, m2 = iy[ 0]-nghost,\ iy[-1]+nghost+1 m1out = m1 + nghost m2out = m2 - nghost varm1 = nghost varm2 = -nghost if iy[0] == locindy[0][0]: m1out = 0 varm1 = 0 if iy[-1] == locindy[-1][-1]: m2out = m2 varm2 = m2 for ix in [ indx[np.mod( ichunk + int(ichunk / nchunks[2]) + int(ichunk / nchunks[1]), nchunks[0])] ]: l1, l2 = ix[ 0]-nghost,\ ix[-1]+nghost+1 l1out = l1 + nghost l2out = l2 - nghost varl1 = nghost varl2 = -nghost if ix[0] == locindx[0][0]: l1out = 0 varl1 = 0 if ix[-1] == locindx[-1][-1]: l2out = l2 varl2 = l2 if not quiet: print('remeshing ' + key + ' chunk {}'.format([iz, iy, ix])) var = calc_derived_data(src['data'], dst['data'], key, par, gd, l1, l2, m1, m2, n1, n2, nghost=nghost) #print('var shape {}'.format(var.shape)) #if not quiet: # print('writing '+key+ # ' shape {} chunk {}'.format( # var.shape, [iz,iy,ix])) if is_vector(key): dst['data'][key][:, n1out:n2out, m1out:m2out, l1out:l2out] = dtype( var[:, varn1:varn2, varm1:varm2, varl1:varl2]) else: dst['data'][key][n1out:n2out, m1out:m2out, l1out:l2out] = dtype( var[varn1:varn2, varm1:varm2, varl1:varl2])