DNS.py 45.2 KB
Newer Older
Cristian Lalescu's avatar
Cristian Lalescu committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
#######################################################################
#                                                                     #
#  Copyright 2015 Max Planck Institute                                #
#                 for Dynamics and Self-Organization                  #
#                                                                     #
#  This file is part of bfps.                                         #
#                                                                     #
#  bfps is free software: you can redistribute it and/or modify       #
#  it under the terms of the GNU General Public License as published  #
#  by the Free Software Foundation, either version 3 of the License,  #
#  or (at your option) any later version.                             #
#                                                                     #
#  bfps is distributed in the hope that it will be useful,            #
#  but WITHOUT ANY WARRANTY; without even the implied warranty of     #
#  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the      #
#  GNU General Public License for more details.                       #
#                                                                     #
#  You should have received a copy of the GNU General Public License  #
#  along with bfps.  If not, see <http://www.gnu.org/licenses/>       #
#                                                                     #
# Contact: Cristian.Lalescu@ds.mpg.de                                 #
#                                                                     #
#######################################################################



import os
import sys
import shutil
import subprocess
import argparse
import h5py
import math
import numpy as np
import warnings

import bfps
from ._code import _code
39
from bfps import tools
Cristian Lalescu's avatar
Cristian Lalescu committed
40
41
42
43
44
45
46
47

class DNS(_code):
    """This class is meant to stitch together the C++ code into a final source file,
    compile it, and handle all job launching.
    """
    def __init__(
            self,
            work_dir = './',
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
            simname = 'test'):
        _code.__init__(
                self,
                work_dir = work_dir,
                simname = simname)
        self.host_info = {'type'        : 'cluster',
                          'environment' : None,
                          'deltanprocs' : 1,
                          'queue'       : '',
                          'mail_address': '',
                          'mail_events' : None}
        self.generate_default_parameters()
        return None
    def set_precision(
            self,
            fluid_dtype):
Cristian Lalescu's avatar
Cristian Lalescu committed
64
65
66
67
68
69
70
71
72
73
        if fluid_dtype in [np.float32, np.float64]:
            self.fluid_dtype = fluid_dtype
        elif fluid_dtype in ['single', 'double']:
            if fluid_dtype == 'single':
                self.fluid_dtype = np.dtype(np.float32)
            elif fluid_dtype == 'double':
                self.fluid_dtype = np.dtype(np.float64)
        self.rtype = self.fluid_dtype
        if self.rtype == np.float32:
            self.ctype = np.dtype(np.complex64)
74
            self.C_field_dtype = 'float'
75
            self.fluid_precision = 'single'
Cristian Lalescu's avatar
Cristian Lalescu committed
76
77
        elif self.rtype == np.float64:
            self.ctype = np.dtype(np.complex128)
78
            self.C_field_dtype = 'double'
79
80
            self.fluid_precision = 'double'
        return None
81
82
    def write_src(
            self):
Cristian Lalescu's avatar
Cristian Lalescu committed
83
84
85
86
87
        self.version_message = (
                '/***********************************************************************\n' +
                '* this code automatically generated by bfps\n' +
                '* version {0}\n'.format(bfps.__version__) +
                '***********************************************************************/\n\n\n')
88
89
90
91
92
        self.include_list = [
                '"base.hpp"',
                '"scope_timer.hpp"',
                '"fftw_interface.hpp"',
                '"full_code/main_code.hpp"',
93
                '<cmath>',
94
95
96
97
98
99
100
101
102
                '<iostream>',
                '<hdf5.h>',
                '<string>',
                '<cstring>',
                '<fftw3-mpi.h>',
                '<omp.h>',
                '<cfenv>',
                '<cstdlib>',
                '"full_code/{0}.hpp"\n'.format(self.dns_type)]
Cristian Lalescu's avatar
Cristian Lalescu committed
103
        self.main = """
104
105
106
107
108
109
110
111
            int main(int argc, char *argv[])
            {{
                bool fpe = (
                    (getenv("BFPS_FPE_OFF") == nullptr) ||
                    (getenv("BFPS_FPE_OFF") != std::string("TRUE")));
                return main_code< {0} >(argc, argv, fpe);
            }}
            """.format(self.dns_type + '<{0}>'.format(self.C_field_dtype))
112
113
114
115
116
117
        self.includes = '\n'.join(
                ['#include ' + hh
                 for hh in self.include_list])
        with open(self.name + '.cpp', 'w') as outfile:
            outfile.write(self.version_message + '\n\n')
            outfile.write(self.includes + '\n\n')
118
119
120
121
122
123
124
            outfile.write(
                    self.cread_pars(
                       template_class = '{0}<rnumber>::'.format(self.dns_type),
                        template_prefix = 'template <typename rnumber> ',
                        simname_variable = 'this->simname.c_str()',
                        prepend_this = True) +
                    '\n\n')
125
126
127
128
129
            for rnumber in ['float', 'double']:
                outfile.write(self.cread_pars(
                    template_class = '{0}<{1}>::'.format(self.dns_type, rnumber),
                    template_prefix = 'template '.format(rnumber),
                    just_declaration = True) + '\n\n')
130
            if self.dns_type in ['NSVEparticles', 'NSVE_no_output', 'NSVEparticles_no_output', 'NSVEparticlesP2P']:
131
132
133
                outfile.write('template <typename rnumber> int NSVE<rnumber>::read_parameters(){return EXIT_SUCCESS;}\n')
                outfile.write('template int NSVE<float>::read_parameters();\n')
                outfile.write('template int NSVE<double>::read_parameters();\n\n')
134
135
136
137
            if self.dns_type in ['NSVEparticles_no_output']:
                outfile.write('template <typename rnumber> int NSVEparticles<rnumber>::read_parameters(){return EXIT_SUCCESS;}\n')
                outfile.write('template int NSVEparticles<float>::read_parameters();\n')
                outfile.write('template int NSVEparticles<double>::read_parameters();\n\n')
138
            outfile.write(self.main + '\n')
139
140
141
        return None
    def generate_default_parameters(self):
        # these parameters are relevant for all DNS classes
142
143
144
145
146
147
148
        self.parameters['dealias_type'] = int(1)
        self.parameters['dkx'] = float(1.0)
        self.parameters['dky'] = float(1.0)
        self.parameters['dkz'] = float(1.0)
        self.parameters['niter_todo'] = int(8)
        self.parameters['niter_stat'] = int(1)
        self.parameters['niter_out'] = int(8)
149
        self.parameters['checkpoints_per_file'] = int(1)
150
        self.parameters['dt'] = float(0.01)
151
        self.parameters['nu'] = float(0.1)
152
        self.parameters['fmode'] = int(1)
153
154
155
156
157
158
159
160
        self.parameters['famplitude'] = float(0.5)
        self.parameters['fk0'] = float(2.0)
        self.parameters['fk1'] = float(4.0)
        self.parameters['forcing_type'] = 'linear'
        self.parameters['histogram_bins'] = int(256)
        self.parameters['max_velocity_estimate'] = float(1)
        self.parameters['max_vorticity_estimate'] = float(1)
        # parameters specific to particle version
161
162
163
164
165
166
        self.NSVEp_extra_parameters = {}
        self.NSVEp_extra_parameters['niter_part'] = int(1)
        self.NSVEp_extra_parameters['nparticles'] = int(10)
        self.NSVEp_extra_parameters['tracers0_integration_steps'] = int(4)
        self.NSVEp_extra_parameters['tracers0_neighbours'] = int(1)
        self.NSVEp_extra_parameters['tracers0_smoothness'] = int(1)
Cristian Lalescu's avatar
Cristian Lalescu committed
167
        return None
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
    def get_kspace(self):
        kspace = {}
        if self.parameters['dealias_type'] == 1:
            kMx = self.parameters['dkx']*(self.parameters['nx']//2 - 1)
            kMy = self.parameters['dky']*(self.parameters['ny']//2 - 1)
            kMz = self.parameters['dkz']*(self.parameters['nz']//2 - 1)
        else:
            kMx = self.parameters['dkx']*(self.parameters['nx']//3 - 1)
            kMy = self.parameters['dky']*(self.parameters['ny']//3 - 1)
            kMz = self.parameters['dkz']*(self.parameters['nz']//3 - 1)
        kspace['kM'] = max(kMx, kMy, kMz)
        kspace['dk'] = min(self.parameters['dkx'],
                           self.parameters['dky'],
                           self.parameters['dkz'])
        nshells = int(kspace['kM'] / kspace['dk']) + 2
        kspace['nshell'] = np.zeros(nshells, dtype = np.int64)
        kspace['kshell'] = np.zeros(nshells, dtype = np.float64)
        kspace['kx'] = np.arange( 0,
                                  self.parameters['nx']//2 + 1).astype(np.float64)*self.parameters['dkx']
        kspace['ky'] = np.arange(-self.parameters['ny']//2 + 1,
                                  self.parameters['ny']//2 + 1).astype(np.float64)*self.parameters['dky']
        kspace['ky'] = np.roll(kspace['ky'], self.parameters['ny']//2+1)
        kspace['kz'] = np.arange(-self.parameters['nz']//2 + 1,
                                  self.parameters['nz']//2 + 1).astype(np.float64)*self.parameters['dkz']
        kspace['kz'] = np.roll(kspace['kz'], self.parameters['nz']//2+1)
        return kspace
    def get_data_file_name(self):
        return os.path.join(self.work_dir, self.simname + '.h5')
    def get_data_file(self):
        return h5py.File(self.get_data_file_name(), 'r')
    def get_particle_file_name(self):
        return os.path.join(self.work_dir, self.simname + '_particles.h5')
    def get_particle_file(self):
        return h5py.File(self.get_particle_file_name(), 'r')
    def get_postprocess_file_name(self):
        return os.path.join(self.work_dir, self.simname + '_postprocess.h5')
    def get_postprocess_file(self):
        return h5py.File(self.get_postprocess_file_name(), 'r')
    def compute_statistics(self, iter0 = 0, iter1 = None):
        """Run basic postprocessing on raw data.
        The energy spectrum :math:`E(t, k)` and the enstrophy spectrum
        :math:`\\frac{1}{2}\omega^2(t, k)` are computed from the

        .. math::

            \sum_{k \\leq \\|\\mathbf{k}\\| \\leq k+dk}\\hat{u_i} \\hat{u_j}^*, \\hskip .5cm
            \sum_{k \\leq \\|\\mathbf{k}\\| \\leq k+dk}\\hat{\omega_i} \\hat{\\omega_j}^*

        tensors, and the enstrophy spectrum is also used to
        compute the dissipation :math:`\\varepsilon(t)`.
        These basic quantities are stored in a newly created HDF5 file,
        ``simname_postprocess.h5``.
        """
        if len(list(self.statistics.keys())) > 0:
            return None
        self.read_parameters()
        with self.get_data_file() as data_file:
            if 'moments' not in data_file['statistics'].keys():
                return None
            iter0 = min((data_file['statistics/moments/velocity'].shape[0] *
                         self.parameters['niter_stat']-1),
                        iter0)
            if type(iter1) == type(None):
                iter1 = data_file['iteration'].value
            else:
                iter1 = min(data_file['iteration'].value, iter1)
            ii0 = iter0 // self.parameters['niter_stat']
            ii1 = iter1 // self.parameters['niter_stat']
            self.statistics['kshell'] = data_file['kspace/kshell'].value
            self.statistics['kM'] = data_file['kspace/kM'].value
            self.statistics['dk'] = data_file['kspace/dk'].value
            computation_needed = True
            pp_file = h5py.File(self.get_postprocess_file_name(), 'a')
            if 'ii0' in pp_file.keys():
                computation_needed =  not (ii0 == pp_file['ii0'].value and
                                           ii1 == pp_file['ii1'].value)
                if computation_needed:
                    for k in pp_file.keys():
                        del pp_file[k]
            if computation_needed:
                pp_file['iter0'] = iter0
                pp_file['iter1'] = iter1
                pp_file['ii0'] = ii0
                pp_file['ii1'] = ii1
                pp_file['t'] = (self.parameters['dt']*
                                self.parameters['niter_stat']*
                                (np.arange(ii0, ii1+1).astype(np.float)))
                pp_file['energy(t, k)'] = (
                    data_file['statistics/spectra/velocity_velocity'][ii0:ii1+1, :, 0, 0] +
                    data_file['statistics/spectra/velocity_velocity'][ii0:ii1+1, :, 1, 1] +
                    data_file['statistics/spectra/velocity_velocity'][ii0:ii1+1, :, 2, 2])/2
                pp_file['enstrophy(t, k)'] = (
                    data_file['statistics/spectra/vorticity_vorticity'][ii0:ii1+1, :, 0, 0] +
                    data_file['statistics/spectra/vorticity_vorticity'][ii0:ii1+1, :, 1, 1] +
                    data_file['statistics/spectra/vorticity_vorticity'][ii0:ii1+1, :, 2, 2])/2
                pp_file['vel_max(t)'] = data_file['statistics/moments/velocity']  [ii0:ii1+1, 9, 3]
                pp_file['renergy(t)'] = data_file['statistics/moments/velocity'][ii0:ii1+1, 2, 3]/2
            for k in ['t',
                      'energy(t, k)',
                      'enstrophy(t, k)',
                      'vel_max(t)',
                      'renergy(t)']:
                if k in pp_file.keys():
                    self.statistics[k] = pp_file[k].value
            self.compute_time_averages()
        return None
    def compute_time_averages(self):
        """Compute easy stats.

        Further computation of statistics based on the contents of
        ``simname_postprocess.h5``.
        Standard quantities are as follows
        (consistent with [Ishihara]_):

        .. math::

            U_{\\textrm{int}}(t) = \\sqrt{\\frac{2E(t)}{3}}, \\hskip .5cm
            L_{\\textrm{int}}(t) = \\frac{\pi}{2U_{int}^2(t)} \\int \\frac{dk}{k} E(t, k), \\hskip .5cm
            T_{\\textrm{int}}(t) =
            \\frac{L_{\\textrm{int}}(t)}{U_{\\textrm{int}}(t)}

            \\eta_K = \\left(\\frac{\\nu^3}{\\varepsilon}\\right)^{1/4}, \\hskip .5cm
            \\tau_K = \\left(\\frac{\\nu}{\\varepsilon}\\right)^{1/2}, \\hskip .5cm
            \\lambda = \\sqrt{\\frac{15 \\nu U_{\\textrm{int}}^2}{\\varepsilon}}

            Re = \\frac{U_{\\textrm{int}} L_{\\textrm{int}}}{\\nu}, \\hskip
            .5cm
            R_{\\lambda} = \\frac{U_{\\textrm{int}} \\lambda}{\\nu}

        .. [Ishihara] T. Ishihara et al,
                      *Small-scale statistics in high-resolution direct numerical
                      simulation of turbulence: Reynolds number dependence of
                      one-point velocity gradient statistics*.
                      J. Fluid Mech.,
                      **592**, 335-366, 2007
        """
        for key in ['energy', 'enstrophy']:
            self.statistics[key + '(t)'] = (self.statistics['dk'] *
                                            np.sum(self.statistics[key + '(t, k)'], axis = 1))
        self.statistics['Uint(t)'] = np.sqrt(2*self.statistics['energy(t)'] / 3)
        self.statistics['Lint(t)'] = ((self.statistics['dk']*np.pi /
                                       (2*self.statistics['Uint(t)']**2)) *
                                      np.nansum(self.statistics['energy(t, k)'] /
                                                self.statistics['kshell'][None, :], axis = 1))
        for key in ['energy',
                    'enstrophy',
                    'vel_max',
                    'Uint',
                    'Lint']:
            if key + '(t)' in self.statistics.keys():
                self.statistics[key] = np.average(self.statistics[key + '(t)'], axis = 0)
        for suffix in ['', '(t)']:
            self.statistics['diss'    + suffix] = (self.parameters['nu'] *
                                                   self.statistics['enstrophy' + suffix]*2)
            self.statistics['etaK'    + suffix] = (self.parameters['nu']**3 /
                                                   self.statistics['diss' + suffix])**.25
            self.statistics['tauK'    + suffix] =  (self.parameters['nu'] /
                                                    self.statistics['diss' + suffix])**.5
            self.statistics['Re' + suffix] = (self.statistics['Uint' + suffix] *
                                              self.statistics['Lint' + suffix] /
                                              self.parameters['nu'])
            self.statistics['lambda' + suffix] = (15 * self.parameters['nu'] *
                                                  self.statistics['Uint' + suffix]**2 /
                                                  self.statistics['diss' + suffix])**.5
            self.statistics['Rlambda' + suffix] = (self.statistics['Uint' + suffix] *
                                                   self.statistics['lambda' + suffix] /
                                                   self.parameters['nu'])
            self.statistics['kMeta' + suffix] = (self.statistics['kM'] *
                                                 self.statistics['etaK' + suffix])
            if self.parameters['dealias_type'] == 1:
                self.statistics['kMeta' + suffix] *= 0.8
        self.statistics['Tint'] = self.statistics['Lint'] / self.statistics['Uint']
        self.statistics['Taylor_microscale'] = self.statistics['lambda']
        return None
    def set_plt_style(
            self,
            style = {'dashes' : (None, None)}):
        self.style.update(style)
        return None
    def convert_complex_from_binary(
            self,
            field_name = 'vorticity',
            iteration = 0,
            file_name = None):
        """read the Fourier representation of a vector field.

        Read the binary file containing iteration ``iteration`` of the
        field ``field_name``, and write it in a ``.h5`` file.
        """
        data = np.memmap(
                os.path.join(self.work_dir,
                             self.simname + '_{0}_i{1:0>5x}'.format('c' + field_name, iteration)),
                dtype = self.ctype,
                mode = 'r',
                shape = (self.parameters['ny'],
                         self.parameters['nz'],
                         self.parameters['nx']//2+1,
                         3))
        if type(file_name) == type(None):
            file_name = self.simname + '_{0}_i{1:0>5x}.h5'.format('c' + field_name, iteration)
            file_name = os.path.join(self.work_dir, file_name)
        f = h5py.File(file_name, 'a')
        f[field_name + '/complex/{0}'.format(iteration)] = data
        f.close()
        return None
    def write_par(
            self,
            iter0 = 0,
            particle_ic = None):
        assert (self.parameters['niter_todo'] % self.parameters['niter_stat'] == 0)
        assert (self.parameters['niter_todo'] % self.parameters['niter_out']  == 0)
        assert (self.parameters['niter_out']  % self.parameters['niter_stat'] == 0)
380
        if self.dns_type in ['NSVEparticles_no_output', 'NSVEparticlesP2P', 'NSVEparticles']:
381
382
            assert (self.parameters['niter_todo'] % self.parameters['niter_part'] == 0)
            assert (self.parameters['niter_out']  % self.parameters['niter_part'] == 0)
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
        _code.write_par(self, iter0 = iter0)
        with h5py.File(self.get_data_file_name(), 'r+') as ofile:
            ofile['bfps_info/exec_name'] = self.name
            kspace = self.get_kspace()
            for k in kspace.keys():
                ofile['kspace/' + k] = kspace[k]
            nshells = kspace['nshell'].shape[0]
            kspace = self.get_kspace()
            nshells = kspace['nshell'].shape[0]
            vec_stat_datasets = ['velocity', 'vorticity']
            scal_stat_datasets = []
            for k in vec_stat_datasets:
                time_chunk = 2**20//(8*3*3*nshells)
                time_chunk = max(time_chunk, 1)
                ofile.create_dataset('statistics/spectra/' + k + '_' + k,
                                     (1, nshells, 3, 3),
                                     chunks = (time_chunk, nshells, 3, 3),
                                     maxshape = (None, nshells, 3, 3),
                                     dtype = np.float64)
                time_chunk = 2**20//(8*4*10)
                time_chunk = max(time_chunk, 1)
                a = ofile.create_dataset('statistics/moments/' + k,
                                     (1, 10, 4),
                                     chunks = (time_chunk, 10, 4),
                                     maxshape = (None, 10, 4),
                                     dtype = np.float64)
                time_chunk = 2**20//(8*4*self.parameters['histogram_bins'])
                time_chunk = max(time_chunk, 1)
                ofile.create_dataset('statistics/histograms/' + k,
                                     (1,
                                      self.parameters['histogram_bins'],
                                      4),
                                     chunks = (time_chunk,
                                               self.parameters['histogram_bins'],
                                               4),
                                     maxshape = (None,
                                                 self.parameters['histogram_bins'],
                                                 4),
                                     dtype = np.int64)
            ofile['checkpoint'] = int(0)
423
        if self.dns_type in ['NSVE', 'NSVE_no_output']:
424
425
426
427
428
429
430
431
432
433
434
            return None

        if type(particle_ic) == type(None):
            pbase_shape = (self.parameters['nparticles'],)
            number_of_particles = self.parameters['nparticles']
        else:
            pbase_shape = particle_ic.shape[:-1]
            assert(particle_ic.shape[-1] == 3)
            number_of_particles = 1
            for val in pbase_shape[1:]:
                number_of_particles *= val
435
436
437
        ncomponents = 3
        if self.dns_type in ['NSVEparticlesP2P']:
            ncomponents = 6
438
439
440
441
442
443
444
445
446
447
        with h5py.File(self.get_checkpoint_0_fname(), 'a') as ofile:
            s = 0
            ofile.create_group('tracers{0}'.format(s))
            ofile.create_group('tracers{0}/rhs'.format(s))
            ofile.create_group('tracers{0}/state'.format(s))
            ofile['tracers{0}/rhs'.format(s)].create_dataset(
                    '0',
                    shape = (
                        (self.parameters['tracers{0}_integration_steps'.format(s)],) +
                        pbase_shape +
448
                        (ncomponents,)),
449
450
451
452
453
                    dtype = np.float)
            ofile['tracers{0}/state'.format(s)].create_dataset(
                    '0',
                    shape = (
                        pbase_shape +
454
                        (ncomponents,)),
455
456
                    dtype = np.float)
        return None
457
    def job_parser_arguments(
458
459
            self,
            parser):
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
        parser.add_argument(
                '--ncpu',
                type = int,
                dest = 'ncpu',
                default = -1)
        parser.add_argument(
                '--np', '--nprocesses',
                metavar = 'NPROCESSES',
                help = 'number of mpi processes to use',
                type = int,
                dest = 'nb_processes',
                default = 4)
        parser.add_argument(
                '--ntpp', '--nthreads-per-process',
                type = int,
                dest = 'nb_threads_per_process',
                metavar = 'NTHREADS_PER_PROCESS',
                help = 'number of threads to use per MPI process',
                default = 1)
        parser.add_argument(
                '--no-submit',
                action = 'store_true',
                dest = 'no_submit')
        parser.add_argument(
                '--environment',
                type = str,
                dest = 'environment',
                default = None)
        parser.add_argument(
                '--minutes',
                type = int,
                dest = 'minutes',
                default = 5,
                help = 'If environment supports it, this is the requested wall-clock-limit.')
        parser.add_argument(
               '--njobs',
               type = int, dest = 'njobs',
               default = 1)
        return None
    def simulation_parser_arguments(
            self,
            parser):
        parser.add_argument(
                '--simname',
                type = str, dest = 'simname',
                default = 'test')
        parser.add_argument(
507
               '-n', '--grid-size',
508
509
510
511
512
               type = int,
               dest = 'n',
               default = 32,
               metavar = 'N',
               help = 'code is run by default in a grid of NxNxN')
513
514
515
516
517
518
519
520
        for coord in ['x', 'y', 'z']:
            parser.add_argument(
                   '--L{0}'.format(coord), '--box-length-{0}'.format(coord),
                   type = float,
                   dest = 'L{0}'.format(coord),
                   default = 2.0,
                   metavar = 'length{0}'.format(coord),
                   help = 'length of the box in the {0} direction will be `length{0} x pi`'.format(coord))
521
522
523
524
525
526
527
528
529
        parser.add_argument(
                '--wd',
                type = str, dest = 'work_dir',
                default = './')
        parser.add_argument(
                '--precision',
                choices = ['single', 'double'],
                type = str,
                default = 'single')
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
        parser.add_argument(
                '--src-wd',
                type = str,
                dest = 'src_work_dir',
                default = '')
        parser.add_argument(
                '--src-simname',
                type = str,
                dest = 'src_simname',
                default = '')
        parser.add_argument(
                '--src-iteration',
                type = int,
                dest = 'src_iteration',
                default = 0)
        parser.add_argument(
               '--kMeta',
               type = float,
               dest = 'kMeta',
               default = 2.0)
        parser.add_argument(
               '--dtfactor',
               type = float,
               dest = 'dtfactor',
               default = 0.5,
               help = 'dt is computed as DTFACTOR / N')
556
557
558
559
        return None
    def particle_parser_arguments(
            self,
            parser):
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
        parser.add_argument(
               '--particle-rand-seed',
               type = int,
               dest = 'particle_rand_seed',
               default = None)
        parser.add_argument(
               '--pclouds',
               type = int,
               dest = 'pclouds',
               default = 1,
               help = ('number of particle clouds. Particle "clouds" '
                       'consist of particles distributed according to '
                       'pcloud-type.'))
        parser.add_argument(
                '--pcloud-type',
                choices = ['random-cube',
                           'regular-cube'],
                dest = 'pcloud_type',
                default = 'random-cube')
        parser.add_argument(
               '--particle-cloud-size',
               type = float,
               dest = 'particle_cloud_size',
               default = 2*np.pi)
        return None
585
586
587
588
589
590
591
592
593
594
595
596
597
598
    def add_parser_arguments(
            self,
            parser):
        subparsers = parser.add_subparsers(
                dest = 'DNS_class',
                help = 'type of simulation to run')
        subparsers.required = True
        parser_NSVE = subparsers.add_parser(
                'NSVE',
                help = 'plain Navier-Stokes vorticity formulation')
        self.simulation_parser_arguments(parser_NSVE)
        self.job_parser_arguments(parser_NSVE)
        self.parameters_to_parser_arguments(parser_NSVE)

599
600
601
602
603
604
605
606
607
608
609
610
611
612
        parser_NSVE_no_output = subparsers.add_parser(
                'NSVE_no_output',
                help = 'plain Navier-Stokes vorticity formulation, checkpoints are NOT SAVED')
        self.simulation_parser_arguments(parser_NSVE_no_output)
        self.job_parser_arguments(parser_NSVE_no_output)
        self.parameters_to_parser_arguments(parser_NSVE_no_output)

        parser_NSVEparticles_no_output = subparsers.add_parser(
                'NSVEparticles_no_output',
                help = 'plain Navier-Stokes vorticity formulation, with basic fluid tracers, checkpoints are NOT SAVED')
        self.simulation_parser_arguments(parser_NSVEparticles_no_output)
        self.job_parser_arguments(parser_NSVEparticles_no_output)
        self.particle_parser_arguments(parser_NSVEparticles_no_output)
        self.parameters_to_parser_arguments(parser_NSVEparticles_no_output)
613
        self.parameters_to_parser_arguments(
614
                parser_NSVEparticles_no_output,
615
                self.NSVEp_extra_parameters)
616
617
618
619
620
621
622
623
624
625
626

        parser_NSVEp2 = subparsers.add_parser(
                'NSVEparticles',
                help = 'plain Navier-Stokes vorticity formulation, with basic fluid tracers')
        self.simulation_parser_arguments(parser_NSVEp2)
        self.job_parser_arguments(parser_NSVEp2)
        self.particle_parser_arguments(parser_NSVEp2)
        self.parameters_to_parser_arguments(parser_NSVEp2)
        self.parameters_to_parser_arguments(
                parser_NSVEp2,
                self.NSVEp_extra_parameters)
627
628
629
630
631
632
633
634
635
636
637

        parser_NSVEp2p = subparsers.add_parser(
                'NSVEparticlesP2P',
                help = 'plain Navier-Stokes vorticity formulation, with basic fluid tracers')
        self.simulation_parser_arguments(parser_NSVEp2p)
        self.job_parser_arguments(parser_NSVEp2p)
        self.particle_parser_arguments(parser_NSVEp2p)
        self.parameters_to_parser_arguments(parser_NSVEp2p)
        self.parameters_to_parser_arguments(
                parser_NSVEp2p,
                self.NSVEp_extra_parameters)
638
        return None
639
640
    def prepare_launch(
            self,
641
642
            args = [],
            extra_parameters = None):
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
        """Set up reasonable parameters.

        With the default Lundgren forcing applied in the band [2, 4],
        we can estimate the dissipation, therefore we can estimate
        :math:`k_M \\eta_K` and constrain the viscosity.

        In brief, the command line parameter :math:`k_M \\eta_K` is
        used in the following formula for :math:`\\nu` (:math:`N` is the
        number of real space grid points per coordinate):

        .. math::

            \\nu = \\left(\\frac{2 k_M \\eta_K}{N} \\right)^{4/3}

        With this choice, the average dissipation :math:`\\varepsilon`
        will be close to 0.4, and the integral scale velocity will be
        close to 0.77, yielding the approximate value for the Taylor
        microscale and corresponding Reynolds number:

        .. math::

            \\lambda \\approx 4.75\\left(\\frac{2 k_M \\eta_K}{N} \\right)^{4/6}, \\hskip .5in
            R_\\lambda \\approx 3.7 \\left(\\frac{N}{2 k_M \\eta_K} \\right)^{4/6}

        """
        opt = _code.prepare_launch(self, args = args)
669
670
671
672
        self.set_precision(opt.precision)
        self.dns_type = opt.DNS_class
        self.name = self.dns_type + '-' + self.fluid_precision + '-v' + bfps.__version__
        # merge parameters if needed
673
        if self.dns_type in ['NSVEparticles', 'NSVEparticlesP2P', 'NSVEparticles_no_output']:
674
675
            for k in self.NSVEp_extra_parameters.keys():
                self.parameters[k] = self.NSVEp_extra_parameters[k]
676
677
678
679
        if type(extra_parameters) != type(None):
            if self.dns_type in extra_parameters.keys():
                for k in extra_parameters[self.dns_type].keys():
                    self.parameters[k] = extra_parameters[self.dns_type][k]
680
681
682
683
684
685
686
687
688
689
690
        self.parameters['nu'] = (opt.kMeta * 2 / opt.n)**(4./3)
        self.parameters['dt'] = (opt.dtfactor / opt.n)
        # custom famplitude for 288 and 576
        if opt.n == 288:
            self.parameters['famplitude'] = 0.45
        elif opt.n == 576:
            self.parameters['famplitude'] = 0.47
        if ((self.parameters['niter_todo'] % self.parameters['niter_out']) != 0):
            self.parameters['niter_out'] = self.parameters['niter_todo']
        if len(opt.src_work_dir) == 0:
            opt.src_work_dir = os.path.realpath(opt.work_dir)
691
692
693
694
695
696
        if type(opt.dkx) == type(None):
            opt.dkx = 2. / opt.Lx
        if type(opt.dky) == type(None):
            opt.dky = 2. / opt.Ly
        if type(opt.dkx) == type(None):
            opt.dkz = 2. / opt.Lz
697
698
699
700
701
702
703
704
705
706
        if type(opt.nx) == type(None):
            opt.nx = opt.n
        if type(opt.ny) == type(None):
            opt.ny = opt.n
        if type(opt.nz) == type(None):
            opt.nz = opt.n
        if type(opt.checkpoints_per_file) == type(None):
            # hardcoded FFTW complex representation size
            field_size = 3*(opt.nx+2)*opt.ny*opt.nz*self.fluid_dtype.itemsize
            checkpoint_size = field_size
707
            if self.dns_type in ['NSVEparticles', 'NSVEparticlesP2P', 'NSVEparticles_no_output']:
708
709
710
                rhs_size = self.parameters['tracers0_integration_steps']
                if type(opt.tracers0_integration_steps) != type(None):
                    rhs_size = opt.tracers0_integration_steps
711
712
713
714
                nparticles = opt.nparticles
                if type(nparticles) == type(None):
                    nparticles = self.NSVEp_extra_parameters['nparticles']
                particle_size = (1+rhs_size)*3*nparticles*8
715
716
717
                checkpoint_size += particle_size
            if checkpoint_size < 1e9:
                opt.checkpoints_per_file = int(1e9 / checkpoint_size)
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
        self.pars_from_namespace(opt)
        return opt
    def launch(
            self,
            args = [],
            **kwargs):
        opt = self.prepare_launch(args = args)
        self.launch_jobs(opt = opt, **kwargs)
        return None
    def get_checkpoint_0_fname(self):
        return os.path.join(
                    self.work_dir,
                    self.simname + '_checkpoint_0.h5')
    def generate_tracer_state(
            self,
            rseed = None,
734
735
736
737
            species = 0):
        with h5py.File(self.get_checkpoint_0_fname(), 'a') as data_file:
            dset = data_file[
                'tracers{0}/state/0'.format(species)]
738
739
            if not type(rseed) == type(None):
                np.random.seed(rseed)
740
741
742
            nn = self.parameters['nparticles']
            cc = int(0)
            batch_size = int(1e6)
743
744
745
746
747
748
749
750
            def get_random_phases(npoints):
                return np.random.random(
                            (npoints, 3))*2*np.pi
            def get_random_versors(npoints):
                bla = np.random.normal(
                        size = (npoints, 3))
                bla  /= np.sum(bla**2, axis = 1)[:, None]
                return bla
751
752
            while nn > 0:
                if nn > batch_size:
753
754
755
                    dset[cc*batch_size:(cc+1)*batch_size, :3] = get_random_phases(batch_size)
                    if dset.shape[1] == 6:
                        dset[cc*batch_size:(cc+1)*batch_size, 3:] = get_random_versors(batch_size)
756
757
                    nn -= batch_size
                else:
758
759
760
                    dset[cc*batch_size:cc*batch_size+nn, :3] = get_random_phases(nn)
                    if dset.shape[1] == 6:
                        dset[cc*batch_size:cc*batch_size+nn, 3:] = get_random_versors(nn)
761
762
763
                    nn = 0
                cc += 1
        return None
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
    def generate_vector_field(
            self,
            rseed = 7547,
            spectra_slope = 1.,
            amplitude = 1.,
            iteration = 0,
            field_name = 'vorticity',
            write_to_file = False,
            # to switch to constant field, use generate_data_3D_uniform
            # for scalar_generator
            scalar_generator = tools.generate_data_3D):
        """generate vector field.

        The generated field is not divergence free, but it has the proper
        shape.

        :param rseed: seed for random number generator
        :param spectra_slope: spectrum of field will look like k^(-p)
        :param amplitude: all amplitudes are multiplied with this value
        :param iteration: the field is written at this iteration
        :param field_name: the name of the field being generated
        :param write_to_file: should we write the field to file?
        :param scalar_generator: which function to use for generating the
            individual components.
            Possible values: bfps.tools.generate_data_3D,
            bfps.tools.generate_data_3D_uniform
        :type rseed: int
        :type spectra_slope: float
        :type amplitude: float
        :type iteration: int
        :type field_name: str
        :type write_to_file: bool
        :type scalar_generator: function

        :returns: ``Kdata``, a complex valued 4D ``numpy.array`` that uses the
            transposed FFTW layout.
            Kdata[ky, kz, kx, i] is the amplitude of mode (kx, ky, kz) for
            the i-th component of the field.
            (i.e. x is the fastest index and z the slowest index in the
            real-space representation).
        """
        np.random.seed(rseed)
        Kdata00 = scalar_generator(
                self.parameters['nz']//2,
                self.parameters['ny']//2,
                self.parameters['nx']//2,
                p = spectra_slope,
                amplitude = amplitude).astype(self.ctype)
        Kdata01 = scalar_generator(
                self.parameters['nz']//2,
                self.parameters['ny']//2,
                self.parameters['nx']//2,
                p = spectra_slope,
                amplitude = amplitude).astype(self.ctype)
        Kdata02 = scalar_generator(
                self.parameters['nz']//2,
                self.parameters['ny']//2,
                self.parameters['nx']//2,
                p = spectra_slope,
                amplitude = amplitude).astype(self.ctype)
        Kdata0 = np.zeros(
                Kdata00.shape + (3,),
                Kdata00.dtype)
        Kdata0[..., 0] = Kdata00
        Kdata0[..., 1] = Kdata01
        Kdata0[..., 2] = Kdata02
        Kdata1 = tools.padd_with_zeros(
                Kdata0,
                self.parameters['nz'],
                self.parameters['ny'],
                self.parameters['nx'])
        if write_to_file:
            Kdata1.tofile(
                    os.path.join(self.work_dir,
                                 self.simname + "_c{0}_i{1:0>5x}".format(field_name, iteration)))
        return Kdata1
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
    def copy_complex_field(
            self,
            src_file_name,
            src_dset_name,
            dst_file,
            dst_dset_name,
            make_link = True):
        # I define a min_shape thingie, but for now I only trust this method for
        # the case of increasing/decreasing by the same factor in all directions.
        # in principle we could write something more generic, but i'm not sure
        # how complicated that would be
        dst_shape = (self.parameters['nz'],
                     self.parameters['ny'],
                     (self.parameters['nx']+2) // 2,
                     3)
        src_file = h5py.File(src_file_name, 'r')
        if (src_file[src_dset_name].shape == dst_shape):
            if make_link and (src_file[src_dset_name].dtype == self.ctype):
                dst_file[dst_dset_name] = h5py.ExternalLink(
                        src_file_name,
                        src_dset_name)
            else:
                dst_file.create_dataset(
                        dst_dset_name,
                        shape = dst_shape,
                        dtype = self.ctype,
                        fillvalue = 0.0)
                for kz in range(src_file[src_dset_name].shape[0]):
                    dst_file[dst_dset_name][kz] = src_file[src_dset_name][kz]
        else:
            min_shape = (min(dst_shape[0], src_file[src_dset_name].shape[0]),
                         min(dst_shape[1], src_file[src_dset_name].shape[1]),
                         min(dst_shape[2], src_file[src_dset_name].shape[2]),
                         3)
874
            src_shape = src_file[src_dset_name].shape
875
876
877
            dst_file.create_dataset(
                    dst_dset_name,
                    shape = dst_shape,
878
879
                    dtype = np.dtype(self.ctype),
                    fillvalue = complex(0))
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
            for kz in range(min_shape[0]//2):
                dst_file[dst_dset_name][kz,:min_shape[1]//2, :min_shape[2]] = \
                        src_file[src_dset_name][kz, :min_shape[1]//2, :min_shape[2]]
                dst_file[dst_dset_name][kz,
                                        dst_shape[1] - min_shape[1]//2+1:,
                                        :min_shape[2]] = \
                        src_file[src_dset_name][kz,
                                                src_shape[1] - min_shape[1]//2+1,
                                                :min_shape[2]]
                if kz > 0:
                    dst_file[dst_dset_name][-kz,:min_shape[1]//2, :min_shape[2]] = \
                            src_file[src_dset_name][-kz, :min_shape[1]//2, :min_shape[2]]
                    dst_file[dst_dset_name][-kz,
                                            dst_shape[1] - min_shape[1]//2+1:,
                                            :min_shape[2]] = \
                            src_file[src_dset_name][-kz,
                                                    src_shape[1] - min_shape[1]//2+1,
                                                    :min_shape[2]]
898
        return None
899
900
901
902
903
904
905
906
907
    def generate_particle_data(
            self,
            opt = None):
        if self.parameters['nparticles'] > 0:
            self.generate_tracer_state(
                    species = 0,
                    rseed = opt.particle_rand_seed)
            if not os.path.exists(self.get_particle_file_name()):
                with h5py.File(self.get_particle_file_name(), 'w') as particle_file:
Cristian Lalescu's avatar
Cristian Lalescu committed
908
                    particle_file.create_group('tracers0/position')
909
910
911
                    particle_file.create_group('tracers0/velocity')
                    particle_file.create_group('tracers0/acceleration')
        return None
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
    def launch_jobs(
            self,
            opt = None,
            particle_initial_condition = None):
        if not os.path.exists(os.path.join(self.work_dir, self.simname + '.h5')):
            # take care of fields' initial condition
            if not os.path.exists(self.get_checkpoint_0_fname()):
                f = h5py.File(self.get_checkpoint_0_fname(), 'w')
                if len(opt.src_simname) > 0:
                    source_cp = 0
                    src_file = 'not_a_file'
                    while True:
                        src_file = os.path.join(
                            os.path.realpath(opt.src_work_dir),
                            opt.src_simname + '_checkpoint_{0}.h5'.format(source_cp))
                        f0 = h5py.File(src_file, 'r')
                        if '{0}'.format(opt.src_iteration) in f0['vorticity/complex'].keys():
                            f0.close()
                            break
                        source_cp += 1
932
                    self.copy_complex_field(
933
                            src_file,
934
935
936
                            'vorticity/complex/{0}'.format(opt.src_iteration),
                            f,
                            'vorticity/complex/{0}'.format(0))
937
938
939
940
941
942
943
                else:
                    data = self.generate_vector_field(
                           write_to_file = False,
                           spectra_slope = 2.0,
                           amplitude = 0.05)
                    f['vorticity/complex/{0}'.format(0)] = data
                f.close()
944
            ## take care of particles' initial condition
945
            #if self.dns_type in ['NSVEparticles', 'NSVEparticles_no_output']:
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
            #    if opt.pclouds > 1:
            #        np.random.seed(opt.particle_rand_seed)
            #        if opt.pcloud_type == 'random-cube':
            #            particle_initial_condition = (
            #                np.random.random((opt.pclouds, 1, 3))*2*np.pi +
            #                np.random.random((1, self.parameters['nparticles'], 3))*opt.particle_cloud_size)
            #        elif opt.pcloud_type == 'regular-cube':
            #            onedarray = np.linspace(
            #                    -opt.particle_cloud_size/2,
            #                    opt.particle_cloud_size/2,
            #                    self.parameters['nparticles'])
            #            particle_initial_condition = np.zeros(
            #                    (opt.pclouds,
            #                     self.parameters['nparticles'],
            #                     self.parameters['nparticles'],
            #                     self.parameters['nparticles'], 3),
            #                    dtype = np.float64)
            #            particle_initial_condition[:] = \
            #                np.random.random((opt.pclouds, 1, 1, 1, 3))*2*np.pi
            #            particle_initial_condition[..., 0] += onedarray[None, None, None, :]
            #            particle_initial_condition[..., 1] += onedarray[None, None, :, None]
            #            particle_initial_condition[..., 2] += onedarray[None, :, None, None]
968
            self.write_par(
969
                    particle_ic = None)
970
            if self.dns_type in ['NSVEparticles', 'NSVEparticlesP2P', 'NSVEparticles_no_output']:
971
                self.generate_particle_data(opt = opt)
972
973
974
975
976
977
978
979
        self.run(
                nb_processes = opt.nb_processes,
                nb_threads_per_process = opt.nb_threads_per_process,
                njobs = opt.njobs,
                hours = opt.minutes // 60,
                minutes = opt.minutes % 60,
                no_submit = opt.no_submit)
        return None
Cristian Lalescu's avatar
Cristian Lalescu committed
980