Coverage for pySDC/projects/Resilience/paper_plots.py: 0%
36 statements
« prev ^ index » next coverage.py v7.8.0, created at 2025-04-01 13:12 +0000
« prev ^ index » next coverage.py v7.8.0, created at 2025-04-01 13:12 +0000
1# script to make pretty plots for papers or talks
2import numpy as np
3import matplotlib as mpl
4import matplotlib.pyplot as plt
5from pySDC.projects.Resilience.fault_stats import (
6 FaultStats,
7 run_Lorenz,
8 run_Schroedinger,
9 run_vdp,
10 run_quench,
11 run_AC,
12 run_RBC,
13 run_GS,
14 RECOVERY_THRESH_ABS,
15)
16from pySDC.projects.Resilience.strategies import (
17 BaseStrategy,
18 AdaptivityStrategy,
19 IterateStrategy,
20 HotRodStrategy,
21 DIRKStrategy,
22 ERKStrategy,
23 AdaptivityPolynomialError,
24 cmap,
25)
26from pySDC.helpers.plot_helper import setup_mpl, figsize_by_journal
27from pySDC.helpers.stats_helper import get_sorted
30cm = 1 / 2.5
31TEXTWIDTH = 11.9446244611 * cm
32JOURNAL = 'Springer_Numerical_Algorithms'
33BASE_PATH = 'data/paper'
36def get_stats(problem, path='data/stats-jusuf', num_procs=1, strategy_type='SDC'):
37 """
38 Create a FaultStats object for a given problem to use for the plots.
39 Note that the statistics need to be already generated somewhere else, this function will only load them.
41 Args:
42 problem (function): A problem to run
43 path (str): Path to the associated stats for the problem
45 Returns:
46 FaultStats: Object to analyse resilience statistics from
47 """
48 if strategy_type == 'SDC':
49 strategies = [BaseStrategy(), AdaptivityStrategy(), IterateStrategy(), AdaptivityPolynomialError()]
50 if JOURNAL not in ['JSC_beamer']:
51 strategies += [HotRodStrategy()]
52 elif strategy_type == 'RK':
53 strategies = [DIRKStrategy()]
54 if problem.__name__ in ['run_Lorenz', 'run_vdp']:
55 strategies += [ERKStrategy()]
57 stats_analyser = FaultStats(
58 prob=problem,
59 strategies=strategies,
60 faults=[False, True],
61 reload=True,
62 recovery_thresh=1.1,
63 recovery_thresh_abs=RECOVERY_THRESH_ABS.get(problem, 0),
64 mode='default',
65 stats_path=path,
66 num_procs=num_procs,
67 )
68 stats_analyser.get_recovered()
69 return stats_analyser
72def my_setup_mpl(**kwargs):
73 setup_mpl(reset=True, font_size=8)
74 mpl.rcParams.update({'lines.markersize': 6})
77def savefig(fig, name, format='pdf', tight_layout=True): # pragma: no cover
78 """
79 Save a figure to some predefined location.
81 Args:
82 fig (Matplotlib.Figure): The figure of the plot
83 name (str): The name of the plot
84 tight_layout (bool): Apply tight layout or leave as is
85 Returns:
86 None
87 """
88 if tight_layout:
89 fig.tight_layout()
90 path = f'{BASE_PATH}/{name}.{format}'
91 fig.savefig(path, bbox_inches='tight', transparent=True, dpi=200)
92 print(f'saved "{path}"')
95def analyse_resilience(problem, path='data/stats', **kwargs): # pragma: no cover
96 """
97 Generate some stats for resilience / load them if already available and make some plots.
99 Args:
100 problem (function): A problem to run
101 path (str): Path to the associated stats for the problem
103 Returns:
104 None
105 """
107 stats_analyser = get_stats(problem, path)
108 stats_analyser.get_recovered()
110 strategy = IterateStrategy()
111 not_fixed = stats_analyser.get_mask(strategy=strategy, key='recovered', val=False)
112 not_overflow = stats_analyser.get_mask(strategy=strategy, key='bit', val=1, op='uneq', old_mask=not_fixed)
113 stats_analyser.print_faults(not_overflow)
115 compare_strategies(stats_analyser, **kwargs)
116 plot_recovery_rate(stats_analyser, **kwargs)
119def compare_strategies(stats_analyser, **kwargs): # pragma: no cover
120 """
121 Make a plot showing local error and iteration number of time for all strategies
123 Args:
124 stats_analyser (FaultStats): Fault stats object, which contains some stats
126 Returns:
127 None
128 """
129 my_setup_mpl()
130 fig, ax = plt.subplots(figsize=(TEXTWIDTH, 5 * cm))
131 stats_analyser.compare_strategies(ax=ax)
132 savefig(fig, 'compare_strategies', **kwargs)
135def plot_recovery_rate(stats_analyser, **kwargs): # pragma: no cover
136 """
137 Make a plot showing recovery rate for all faults and only for those that can be recovered.
139 Args:
140 stats_analyser (FaultStats): Fault stats object, which contains some stats
142 Returns:
143 None
144 """
145 my_setup_mpl()
146 # fig, axs = plt.subplots(1, 2, figsize=(TEXTWIDTH, 5 * cm), sharex=True, sharey=True)
147 fig, axs = plt.subplots(1, 2, figsize=figsize_by_journal(JOURNAL, 1, 0.4), sharex=True)
148 stats_analyser.plot_things_per_things(
149 'recovered',
150 'bit',
151 False,
152 op=stats_analyser.rec_rate,
153 args={'ylabel': 'recovery rate'},
154 plotting_args={'markevery': 5},
155 ax=axs[0],
156 )
157 plot_recovery_rate_recoverable_only(stats_analyser, fig, axs[1], ylabel='')
158 axs[0].get_legend().remove()
159 axs[0].set_title('All faults')
160 axs[1].set_title('Only recoverable faults')
161 axs[0].set_ylim((-0.05, 1.05))
162 savefig(fig, 'recovery_rate_compared', **kwargs)
165def plot_recovery_rate_recoverable_only(stats_analyser, fig, ax, **kwargs): # pragma: no cover
166 """
167 Plot the recovery rate considering only faults that can be recovered theoretically.
169 Args:
170 stats_analyser (FaultStats): Fault stats object, which contains some stats
171 fig (matplotlib.pyplot.figure): Figure in which to plot
172 ax (matplotlib.pyplot.axes): Somewhere to plot
174 Returns:
175 None
176 """
177 for i in range(len(stats_analyser.strategies)):
178 fixable = stats_analyser.get_fixable_faults_only(strategy=stats_analyser.strategies[i])
180 stats_analyser.plot_things_per_things(
181 'recovered',
182 'bit',
183 False,
184 op=stats_analyser.rec_rate,
185 mask=fixable,
186 args={**kwargs},
187 ax=ax,
188 fig=fig,
189 strategies=[stats_analyser.strategies[i]],
190 plotting_args={'markevery': 10 if stats_analyser.prob.__name__ in ['run_RBC', 'run_Schroedinger'] else 5},
191 )
194def compare_recovery_rate_problems(target='resilience', **kwargs): # pragma: no cover
195 """
196 Compare the recovery rate for various problems.
197 Only faults that can be recovered are shown.
199 Returns:
200 None
201 """
202 if target == 'resilience':
203 problems = [run_Lorenz, run_Schroedinger, run_AC, run_RBC]
204 titles = ['Lorenz', r'Schr\"odinger', 'Allen-Cahn', 'Rayleigh-Benard']
205 elif target in ['thesis', 'talk']:
206 problems = [run_vdp, run_Lorenz, run_GS, run_RBC] # TODO: swap in Gray-Scott
207 titles = ['Van der Pol', 'Lorenz', 'Gray-Scott', 'Rayleigh-Benard']
208 else:
209 raise NotImplementedError()
211 stats = [get_stats(problem, **kwargs) for problem in problems]
213 my_setup_mpl()
214 fig, axs = plt.subplots(2, 2, figsize=figsize_by_journal(JOURNAL, 1, 0.8), sharey=True)
215 [
216 plot_recovery_rate_recoverable_only(stats[i], fig, axs.flatten()[i], ylabel='', title=titles[i])
217 for i in range(len(stats))
218 ]
220 for ax in axs.flatten():
221 ax.get_legend().remove()
223 if kwargs.get('strategy_type', 'SDC') == 'SDC':
224 axs[1, 0].legend(frameon=False, loc="lower right")
225 else:
226 axs[0, 1].legend(frameon=False, loc="lower right")
227 axs[0, 0].set_ylim((-0.05, 1.05))
228 axs[1, 0].set_ylabel('recovery rate')
229 axs[0, 0].set_ylabel('recovery rate')
231 if target == 'talk':
232 axs[0, 0].set_xlabel('')
233 axs[0, 1].set_xlabel('')
235 name = ''
236 for key, val in kwargs.items():
237 name = f'{name}_{key}-{val}'
239 savefig(fig, f'compare_equations{name}.pdf')
242def plot_recovery_rate_detailed_Lorenz(target='resilience'): # pragma: no cover
243 stats = get_stats(run_Lorenz, num_procs=1, strategy_type='SDC')
244 stats.get_recovered()
245 mask = None
247 for x in ['node', 'iteration', 'bit']:
248 if target == 'talk':
249 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.6, 0.6))
250 else:
251 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.8, 0.5))
253 stats.plot_things_per_things(
254 'recovered',
255 x,
256 False,
257 op=stats.rec_rate,
258 mask=mask,
259 args={'ylabel': 'recovery rate'},
260 ax=ax,
261 plotting_args={'markevery': 5 if x == 'bit' else 1},
262 )
263 ax.set_ylim((-0.05, 1.05))
265 if x == 'node':
266 ax.set_xticks([0, 1, 2, 3])
267 elif x == 'iteration':
268 ax.set_xticks([1, 2, 3, 4, 5])
270 savefig(fig, f'recovery_rate_Lorenz_{x}')
273def plot_adaptivity_stuff(): # pragma: no cover
274 """
275 Plot the solution for a van der Pol problem as well as the local error and cost associated with the base scheme and
276 adaptivity in k and dt in order to demonstrate that adaptivity is useful.
278 Returns:
279 None
280 """
281 from pySDC.implementations.hooks.log_errors import LogLocalErrorPostStep
282 from pySDC.implementations.hooks.log_work import LogWork
283 from pySDC.projects.Resilience.hook import LogData
284 import pickle
286 my_setup_mpl()
287 scale = 0.5 if JOURNAL == 'JSC_beamer' else 1.0
288 fig, axs = plt.subplots(3, 1, figsize=figsize_by_journal(JOURNAL, scale, 1), sharex=True, sharey=False)
290 def plot_error(stats, ax, iter_ax, strategy, **kwargs):
291 """
292 Plot global error and cumulative sum of iterations
294 Args:
295 stats (dict): Stats from pySDC run
296 ax (Matplotlib.pyplot.axes): Somewhere to plot the error
297 iter_ax (Matplotlib.pyplot.axes): Somewhere to plot the iterations
298 strategy (pySDC.projects.Resilience.fault_stats.Strategy): The resilience strategy
300 Returns:
301 None
302 """
303 markevery = 1 if type(strategy) in [AdaptivityStrategy, AdaptivityPolynomialError] else 10000
304 e = stats['e_local_post_step']
305 ax.plot([me[0] for me in e], [me[1] for me in e], markevery=markevery, **strategy.style, **kwargs)
306 k = stats['work_newton']
307 iter_ax.plot(
308 [me[0] for me in k], np.cumsum([me[1] for me in k]), **strategy.style, markevery=markevery, **kwargs
309 )
310 ax.set_yscale('log')
311 ax.set_ylabel('local error')
312 iter_ax.set_ylabel(r'Newton iterations')
314 run = False
315 for strategy in [BaseStrategy, IterateStrategy, AdaptivityStrategy, AdaptivityPolynomialError]:
316 S = strategy(newton_inexactness=False)
317 desc = S.get_custom_description(problem=run_vdp, num_procs=1)
318 desc['problem_params']['mu'] = 1000
319 desc['problem_params']['u0'] = (1.1, 0)
320 if strategy in [AdaptivityStrategy, BaseStrategy]:
321 desc['step_params']['maxiter'] = 5
322 if strategy in [BaseStrategy, IterateStrategy]:
323 desc['level_params']['dt'] = 1e-4
324 desc['sweeper_params']['QI'] = 'LU'
325 if strategy in [IterateStrategy]:
326 desc['step_params']['maxiter'] = 99
327 desc['level_params']['restol'] = 1e-10
329 path = f'./data/adaptivity_paper_plot_data_{strategy.__name__}.pickle'
330 if run:
331 stats, _, _ = run_vdp(
332 custom_description=desc,
333 Tend=20,
334 hook_class=[LogLocalErrorPostStep, LogWork, LogData],
335 custom_controller_params={'logger_level': 15},
336 )
338 data = {
339 'u': get_sorted(stats, type='u', recomputed=False),
340 'e_local_post_step': get_sorted(stats, type='e_local_post_step', recomputed=False),
341 'work_newton': get_sorted(stats, type='work_newton', recomputed=None),
342 }
343 with open(path, 'wb') as file:
344 pickle.dump(data, file)
345 else:
346 with open(path, 'rb') as file:
347 data = pickle.load(file)
349 plot_error(data, axs[1], axs[2], strategy())
351 if strategy == BaseStrategy or True:
352 u = data['u']
353 axs[0].plot([me[0] for me in u], [me[1][0] for me in u], color='black', label=r'$u$')
355 axs[2].set_xlabel(r'$t$')
356 axs[0].set_ylabel('solution')
357 axs[2].legend(frameon=JOURNAL == 'JSC_beamer')
358 axs[1].legend(frameon=True)
359 axs[2].set_yscale('log')
360 savefig(fig, 'adaptivity')
363def plot_fault_vdp(bit=0): # pragma: no cover
364 """
365 Make a plot showing the impact of a fault on van der Pol without any resilience.
366 The faults are inserted in the last iteration in the last node in u_t such that you can best see the impact.
368 Args:
369 bit (int): The bit that you want to flip
371 Returns:
372 None
373 """
374 from pySDC.projects.Resilience.fault_stats import (
375 FaultStats,
376 BaseStrategy,
377 )
378 from pySDC.projects.Resilience.hook import LogData
380 stats_analyser = FaultStats(
381 prob=run_vdp,
382 strategies=[BaseStrategy()],
383 faults=[False, True],
384 reload=True,
385 recovery_thresh=1.1,
386 num_procs=1,
387 mode='combination',
388 )
390 my_setup_mpl()
391 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.8, 0.5))
392 colors = ['blue', 'red', 'magenta']
393 ls = ['--', '-']
394 markers = ['*', '^']
395 do_faults = [False, True]
396 superscripts = ['*', '']
397 subscripts = ['', 't', '']
399 run = 779 + 12 * bit # for faults in u_t
400 # run = 11 + 12 * bit # for faults in u
402 for i in range(len(do_faults)):
403 stats, controller, Tend = stats_analyser.single_run(
404 strategy=BaseStrategy(),
405 run=run,
406 faults=do_faults[i],
407 hook_class=[LogData],
408 )
409 u = get_sorted(stats, type='u')
410 faults = get_sorted(stats, type='bitflip')
411 for j in [0, 1]:
412 ax.plot(
413 [me[0] for me in u],
414 [me[1][j] for me in u],
415 ls=ls[i],
416 color=colors[j],
417 label=rf'$u^{ {superscripts[i]}} _{ {subscripts[j]}} $',
418 marker=markers[j],
419 markevery=60,
420 )
421 for idx in range(len(faults)):
422 ax.axvline(faults[idx][0], color='black', label='Fault', ls=':')
423 print(
424 f'Fault at t={faults[idx][0]:.2e}, iter={faults[idx][1][1]}, node={faults[idx][1][2]}, space={faults[idx][1][3]}, bit={faults[idx][1][4]}'
425 )
426 ax.set_title(f'Fault in bit {faults[idx][1][4]}')
428 ax.legend(frameon=True, loc='lower left')
429 ax.set_xlabel(r'$t$')
430 savefig(fig, f'fault_bit_{bit}')
433def plot_fault_Lorenz(bit=0, target='resilience'): # pragma: no cover
434 """
435 Make a plot showing the impact of a fault on the Lorenz attractor without any resilience.
436 The faults are inserted in the last iteration in the last node in x such that you can best see the impact.
438 Args:
439 bit (int): The bit that you want to flip
441 Returns:
442 None
443 """
444 from pySDC.projects.Resilience.fault_stats import (
445 FaultStats,
446 BaseStrategy,
447 )
448 from pySDC.projects.Resilience.hook import LogData
450 stats_analyser = FaultStats(
451 prob=run_Lorenz,
452 strategies=[BaseStrategy()],
453 faults=[False, True],
454 reload=True,
455 recovery_thresh=1.1,
456 num_procs=1,
457 mode='combination',
458 )
460 strategy = BaseStrategy()
462 my_setup_mpl()
463 if target == 'resilience':
464 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.4, 0.6))
465 else:
466 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.8, 0.5))
467 colors = ['grey', strategy.color, 'magenta']
468 ls = ['--', '-']
469 markers = [None, strategy.marker]
470 do_faults = [False, True]
471 superscripts = [r'\mathrm{no~faults}', '']
472 labels = ['x', 'x']
474 run = 19 + 20 * bit
476 for i in range(len(do_faults)):
477 stats, controller, Tend = stats_analyser.single_run(
478 strategy=BaseStrategy(),
479 run=run,
480 faults=do_faults[i],
481 hook_class=[LogData],
482 )
483 u = get_sorted(stats, type='u')
484 faults = get_sorted(stats, type='bitflip')
485 ax.plot(
486 [me[0] for me in u],
487 [me[1][0] for me in u],
488 ls=ls[i],
489 color=colors[i],
490 label=rf'${ {labels[i]}} _{ {superscripts[i]}} $',
491 marker=markers[i],
492 markevery=500,
493 )
494 for idx in range(len(faults)):
495 ax.axvline(faults[idx][0], color='black', label='Fault', ls=':')
496 print(
497 f'Fault at t={faults[idx][0]:.2e}, iter={faults[idx][1][1]}, node={faults[idx][1][2]}, space={faults[idx][1][3]}, bit={faults[idx][1][4]}'
498 )
499 ax.set_title(f'Fault in bit {faults[idx][1][4]}')
501 ax.set_xlabel(r'$t$')
503 h, l = ax.get_legend_handles_labels()
504 fig.legend(
505 h,
506 l,
507 loc='outside lower center',
508 ncols=3,
509 frameon=False,
510 fancybox=True,
511 borderaxespad=0.01,
512 )
514 savefig(fig, f'fault_bit_{bit}')
517def plot_Lorenz_solution(): # pragma: no cover
518 my_setup_mpl()
520 fig, axs = plt.subplots(1, 2, figsize=figsize_by_journal(JOURNAL, 1, 0.4), sharex=True)
522 strategy = BaseStrategy()
523 desc = strategy.get_custom_description(run_Lorenz, num_procs=1)
524 stats, controller, _ = run_Lorenz(custom_description=desc, Tend=strategy.get_Tend(run_Lorenz))
526 u = get_sorted(stats, recomputed=False, type='u')
528 axs[0].plot([me[1][0] for me in u], [me[1][2] for me in u])
529 axs[0].set_ylabel('$z$')
530 axs[0].set_xlabel('$x$')
532 axs[1].plot([me[1][0] for me in u], [me[1][1] for me in u])
533 axs[1].set_ylabel('$y$')
534 axs[1].set_xlabel('$x$')
536 for ax in axs:
537 ax.set_box_aspect(1.0)
539 path = 'data/paper/Lorenz_sol.pdf'
540 fig.savefig(path, bbox_inches='tight', transparent=True, dpi=200)
543def plot_quench_solution(): # pragma: no cover
544 """
545 Plot the solution of Quench problem over time
547 Returns:
548 None
549 """
550 my_setup_mpl()
551 if JOURNAL == 'JSC_beamer':
552 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.5, 0.9))
553 else:
554 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 1.0, 0.45))
556 strategy = BaseStrategy()
558 custom_description = strategy.get_custom_description(run_quench, num_procs=1)
560 stats, controller, _ = run_quench(custom_description=custom_description, Tend=strategy.get_Tend(run_quench))
562 prob = controller.MS[0].levels[0].prob
564 u = get_sorted(stats, type='u', recomputed=False)
566 ax.plot([me[0] for me in u], [max(me[1]) for me in u], color='black', label='$T$')
567 ax.axhline(prob.u_thresh, label=r'$T_\mathrm{thresh}$', ls='--', color='grey', zorder=-1)
568 ax.axhline(prob.u_max, label=r'$T_\mathrm{max}$', ls=':', color='grey', zorder=-1)
570 ax.set_xlabel(r'$t$')
571 ax.legend(frameon=False)
572 savefig(fig, 'quench_sol')
575def plot_RBC_solution(setup='resilience'): # pragma: no cover
576 """
577 Plot solution of Rayleigh-Benard convection
578 """
579 my_setup_mpl()
581 from mpl_toolkits.axes_grid1 import make_axes_locatable
583 nplots = 3 if setup == 'thesis_intro' else 2
584 aspect = 0.8 if nplots == 3 else 0.5
585 plt.rcParams['figure.constrained_layout.use'] = True
586 fig, axs = plt.subplots(nplots, 1, sharex=True, sharey=True, figsize=figsize_by_journal(JOURNAL, 1.0, aspect))
587 caxs = []
588 for ax in axs:
589 divider = make_axes_locatable(ax)
590 caxs += [divider.append_axes('right', size='3%', pad=0.03)]
592 from pySDC.projects.Resilience.RBC import RayleighBenard, PROBLEM_PARAMS
594 prob = RayleighBenard(**PROBLEM_PARAMS)
596 def _plot(t, ax, cax):
597 u_hat = prob.u_exact(t)
598 u = prob.itransform(u_hat)
599 im = ax.pcolormesh(prob.X, prob.Z, u[prob.index('T')], rasterized=True, cmap='plasma')
600 fig.colorbar(im, cax, label=f'$T(t={ {t}} )$')
602 if setup == 'resilience':
603 _plot(0, axs[0], caxs[0])
604 _plot(21, axs[1], caxs[1])
605 elif setup == 'work-precision':
606 _plot(10, axs[0], caxs[0])
607 _plot(16, axs[1], caxs[1])
608 elif setup == 'resilience_thesis':
609 _plot(20, axs[0], caxs[0])
610 _plot(21, axs[1], caxs[1])
611 elif setup == 'thesis_intro':
612 _plot(0, axs[0], caxs[0])
613 _plot(18, axs[1], caxs[1])
614 _plot(30, axs[2], caxs[2])
616 for ax in axs:
617 ax.set_ylabel('$z$')
618 ax.set_aspect(1)
619 axs[-1].set_xlabel('$x$')
621 savefig(fig, f'RBC_sol_{setup}', tight_layout=False)
624def plot_GS_solution(tend=500): # pragma: no cover
625 my_setup_mpl()
627 fig, axs = plt.subplots(1, 2, figsize=figsize_by_journal(JOURNAL, 1.0, 0.45), sharex=True, sharey=True)
629 from mpl_toolkits.axes_grid1 import make_axes_locatable
631 plt.rcParams['figure.constrained_layout.use'] = True
632 cax = []
633 divider = make_axes_locatable(axs[0])
634 cax += [divider.append_axes('right', size='5%', pad=0.05)]
635 divider2 = make_axes_locatable(axs[1])
636 cax += [divider2.append_axes('right', size='5%', pad=0.05)]
638 from pySDC.projects.Resilience.GS import grayscott_imex_diffusion
640 problem_params = {
641 'num_blobs': -48,
642 'L': 2,
643 'nvars': (128,) * 2,
644 'A': 0.062,
645 'B': 0.1229,
646 'Du': 2e-5,
647 'Dv': 1e-5,
648 }
649 P = grayscott_imex_diffusion(**problem_params)
650 Tend = tend
651 im = axs[0].pcolormesh(*P.X, P.u_exact(0)[1], rasterized=True, cmap='binary')
652 im1 = axs[1].pcolormesh(*P.X, P.u_exact(Tend)[1], rasterized=True, cmap='binary')
654 fig.colorbar(im, cax=cax[0])
655 fig.colorbar(im1, cax=cax[1])
656 axs[0].set_title(r'$v(t=0)$')
657 axs[1].set_title(rf'$v(t={ {Tend}} )$')
658 for ax in axs:
659 ax.set_aspect(1)
660 ax.set_xlabel('$x$')
661 ax.set_ylabel('$y$')
662 savefig(fig, f'GrayScott_sol{f"_{tend}" if tend != 500 else ""}')
665def plot_Schroedinger_solution(): # pragma: no cover
666 from pySDC.implementations.problem_classes.NonlinearSchroedinger_MPIFFT import nonlinearschroedinger_imex
668 my_setup_mpl()
669 if JOURNAL == 'JSC_beamer':
670 raise NotImplementedError
671 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.5, 0.9))
672 else:
673 fig, axs = plt.subplots(1, 2, figsize=figsize_by_journal(JOURNAL, 1.0, 0.45), sharex=True, sharey=True)
675 from mpl_toolkits.axes_grid1 import make_axes_locatable
677 plt.rcParams['figure.constrained_layout.use'] = True
678 cax = []
679 divider = make_axes_locatable(axs[0])
680 cax += [divider.append_axes('right', size='5%', pad=0.05)]
681 divider2 = make_axes_locatable(axs[1])
682 cax += [divider2.append_axes('right', size='5%', pad=0.05)]
684 problem_params = dict()
685 problem_params['nvars'] = (256, 256)
686 problem_params['spectral'] = False
687 problem_params['c'] = 1.0
688 description = {'problem_params': problem_params}
689 stats, _, _ = run_Schroedinger(Tend=1.0e0, custom_description=description)
691 P = nonlinearschroedinger_imex(**problem_params)
692 u = get_sorted(stats, type='u')
694 im = axs[0].pcolormesh(*P.X, np.abs(u[0][1]), rasterized=True)
695 im1 = axs[1].pcolormesh(*P.X, np.abs(u[-1][1]), rasterized=True)
697 fig.colorbar(im, cax=cax[0])
698 fig.colorbar(im1, cax=cax[1])
699 axs[0].set_title(r'$\|u(t=0)\|$')
700 axs[1].set_title(r'$\|u(t=1)\|$')
701 for ax in axs:
702 ax.set_aspect(1)
703 ax.set_xlabel('$x$')
704 ax.set_ylabel('$y$')
705 savefig(fig, 'Schroedinger_sol')
708def plot_AC_solution(): # pragma: no cover
709 from pySDC.projects.Resilience.AC import monitor
711 my_setup_mpl()
712 if JOURNAL == 'JSC_beamer':
713 raise NotImplementedError
714 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.5, 0.9))
715 else:
716 fig, axs = plt.subplots(1, 2, figsize=figsize_by_journal(JOURNAL, 1.0, 0.45))
718 description = {'problem_params': {'nvars': (256, 256)}}
719 stats, _, _ = run_AC(Tend=0.032, hook_class=monitor, custom_description=description)
721 u = get_sorted(stats, type='u')
723 computed_radius = get_sorted(stats, type='computed_radius')
724 axs[1].plot([me[0] for me in computed_radius], [me[1] for me in computed_radius], ls='-')
725 axs[1].axvline(0.025, ls=':', label=r'$t=0.025$', color='grey')
726 axs[1].set_title('Radius over time')
727 axs[1].set_xlabel('$t$')
728 axs[1].legend(frameon=False)
730 im = axs[0].imshow(u[0][1], extent=(-0.5, 0.5, -0.5, 0.5))
731 fig.colorbar(im)
732 axs[0].set_title(r'$u_0$')
733 axs[0].set_xlabel('$x$')
734 axs[0].set_ylabel('$y$')
735 savefig(fig, 'AC_sol')
738def plot_vdp_solution(setup='adaptivity'): # pragma: no cover
739 """
740 Plot the solution of van der Pol problem over time to illustrate the varying time scales.
742 Returns:
743 None
744 """
745 from pySDC.implementations.convergence_controller_classes.adaptivity import Adaptivity
747 my_setup_mpl()
748 if JOURNAL == 'JSC_beamer':
749 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.5, 0.9))
750 else:
751 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 1.0, 0.33))
753 if setup == 'adaptivity':
754 custom_description = {
755 'convergence_controllers': {Adaptivity: {'e_tol': 1e-7, 'dt_max': 1e0}},
756 'problem_params': {'mu': 1000, 'crash_at_maxiter': False},
757 'level_params': {'dt': 1e-3},
758 }
759 Tend = 2000
760 elif setup == 'resilience':
761 custom_description = {
762 'convergence_controllers': {Adaptivity: {'e_tol': 1e-7, 'dt_max': 1e0}},
763 'problem_params': {'mu': 5, 'crash_at_maxiter': False},
764 'level_params': {'dt': 1e-3},
765 }
766 Tend = 50
768 stats, _, _ = run_vdp(custom_description=custom_description, Tend=Tend)
770 u = get_sorted(stats, type='u', recomputed=False)
771 _u = np.array([me[1][0] for me in u])
772 _x = np.array([me[0] for me in u])
774 name = ''
775 if setup == 'adaptivity':
776 x1 = _x[abs(_u - 1.1) < 1e-2][0]
777 ax.plot(_x, _u, color='black')
778 ax.axvspan(x1, x1 + 20, alpha=0.4)
779 elif setup == 'resilience':
780 x1 = _x[abs(_u - 2.0) < 1e-2][0]
781 ax.plot(_x, _u, color='black')
782 ax.axvspan(x1, x1 + 11.5, alpha=0.4)
783 name = '_resilience'
785 ax.set_ylabel(r'$u$')
786 ax.set_xlabel(r'$t$')
787 savefig(fig, f'vdp_sol{name}')
790def work_precision(): # pragma: no cover
791 from pySDC.projects.Resilience.work_precision import (
792 all_problems,
793 )
795 all_params = {
796 'record': False,
797 'work_key': 't',
798 'precision_key': 'e_global_rel',
799 'plotting': True,
800 'base_path': 'data/paper',
801 }
803 for mode in ['compare_strategies', 'parallel_efficiency', 'RK_comp']:
804 all_problems(**all_params, mode=mode)
805 all_problems(**{**all_params, 'work_key': 'param'}, mode='compare_strategies')
808def plot_recovery_rate_per_acceptance_threshold(problem, target='resilience'): # pragma no cover
809 stats_analyser = get_stats(problem)
811 if target == 'talk':
812 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.4, 0.6))
813 else:
814 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.8, 0.5))
816 stats_analyser.plot_recovery_thresholds(thresh_range=np.logspace(-1, 4, 500), recoverable_only=False, ax=ax)
817 ax.set_xscale('log')
818 ax.set_ylim((-0.05, 1.05))
819 ax.set_xlabel('relative threshold')
820 savefig(fig, 'recovery_rate_per_thresh')
823def make_plots_for_TIME_X_website(): # pragma: no cover
824 global JOURNAL, BASE_PATH
825 JOURNAL = 'JSC_beamer'
826 BASE_PATH = 'data/paper/time-x_website'
828 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.5, 2.0 / 3.0))
829 plot_recovery_rate_recoverable_only(get_stats(run_vdp), fig, ax)
830 savefig(fig, 'recovery_rate', format='png')
832 from pySDC.projects.Resilience.work_precision import vdp_stiffness_plot
834 vdp_stiffness_plot(base_path=BASE_PATH, format='png')
837def make_plots_for_SIAM_CSE23(): # pragma: no cover
838 """
839 Make plots for the SIAM talk
840 """
841 global JOURNAL, BASE_PATH
842 JOURNAL = 'JSC_beamer'
843 BASE_PATH = 'data/paper/SIAMCSE23'
845 fig, ax = plt.subplots(figsize=figsize_by_journal(JOURNAL, 0.5, 3.0 / 4.0))
846 plot_recovery_rate_recoverable_only(get_stats(run_vdp), fig, ax)
847 savefig(fig, 'recovery_rate')
849 plot_adaptivity_stuff()
850 compare_recovery_rate_problems()
851 plot_vdp_solution()
854def make_plots_for_adaptivity_paper(): # pragma: no cover
855 """
856 Make plots that are supposed to go in the paper.
857 """
858 global JOURNAL, BASE_PATH
859 JOURNAL = 'Springer_Numerical_Algorithms'
860 BASE_PATH = 'data/paper'
862 plot_adaptivity_stuff()
864 work_precision()
866 plot_vdp_solution()
867 plot_AC_solution()
868 plot_Schroedinger_solution()
869 plot_quench_solution()
872def make_plots_for_resilience_paper(): # pragma: no cover
873 global JOURNAL
874 JOURNAL = 'Springer_proceedings'
876 plot_Lorenz_solution()
877 plot_RBC_solution()
878 plot_AC_solution()
879 plot_Schroedinger_solution()
881 plot_fault_Lorenz(0)
882 plot_fault_Lorenz(20)
883 compare_recovery_rate_problems(target='resilience', num_procs=1, strategy_type='SDC')
884 plot_recovery_rate(get_stats(run_Lorenz))
885 plot_recovery_rate_detailed_Lorenz()
886 plot_recovery_rate_per_acceptance_threshold(run_Lorenz)
887 plt.show()
890def make_plots_for_notes(): # pragma: no cover
891 """
892 Make plots for the notes for the website / GitHub
893 """
894 global JOURNAL, BASE_PATH
895 JOURNAL = 'Springer_Numerical_Algorithms'
896 BASE_PATH = 'notes/Lorenz'
898 analyse_resilience(run_Lorenz, format='png')
899 analyse_resilience(run_quench, format='png')
902def make_plots_for_thesis(): # pragma: no cover
903 global JOURNAL
904 JOURNAL = 'TUHH_thesis'
905 for setup in ['thesis_intro', 'resilience_thesis', 'work_precision']:
906 plot_RBC_solution(setup)
908 from pySDC.projects.Resilience.RBC import plot_factorizations_over_time
910 plot_factorizations_over_time(t0=0, Tend=50)
912 from pySDC.projects.Resilience.work_precision import all_problems, single_problem
914 all_params = {
915 'record': False,
916 'work_key': 't',
917 'precision_key': 'e_global_rel',
918 'plotting': True,
919 'base_path': 'data/paper',
920 'target': 'thesis',
921 }
923 for mode in ['compare_strategies', 'parallel_efficiency_dt_k', 'parallel_efficiency_dt', 'RK_comp']:
924 all_problems(**all_params, mode=mode)
925 all_problems(**{**all_params, 'work_key': 'param'}, mode='compare_strategies')
926 single_problem(**all_params, mode='RK_comp_high_order_RBC', problem=run_RBC)
928 for tend in [500, 2000]:
929 plot_GS_solution(tend=tend)
930 for setup in ['resilience', 'adaptivity']:
931 plot_vdp_solution(setup=setup)
933 plot_adaptivity_stuff()
935 plot_fault_Lorenz(0)
936 plot_fault_Lorenz(20)
937 compare_recovery_rate_problems(target='thesis', num_procs=1, strategy_type='SDC')
938 plot_recovery_rate_per_acceptance_threshold(run_Lorenz)
939 plot_recovery_rate(get_stats(run_Lorenz))
940 plot_recovery_rate_detailed_Lorenz()
943def make_plots_for_TUHH_seminar(): # pragma: no cover
944 global JOURNAL
945 JOURNAL = 'JSC_beamer'
947 from pySDC.projects.Resilience.work_precision import (
948 all_problems,
949 )
951 all_params = {
952 'record': False,
953 'work_key': 't',
954 'precision_key': 'e_global_rel',
955 'plotting': True,
956 'base_path': 'data/paper',
957 'target': 'talk',
958 }
960 for mode in ['compare_strategies', 'parallel_efficiency_dt_k', 'parallel_efficiency_dt', 'RK_comp']:
961 all_problems(**all_params, mode=mode)
962 all_problems(**{**all_params, 'work_key': 'param'}, mode='compare_strategies')
964 plot_GS_solution()
965 for setup in ['resilience_thesis', 'work_precision']:
966 plot_RBC_solution(setup)
967 for setup in ['resilience', 'adaptivity']:
968 plot_vdp_solution(setup=setup)
970 plot_adaptivity_stuff()
972 plot_fault_Lorenz(20, target='talk')
973 compare_recovery_rate_problems(target='talk', num_procs=1, strategy_type='SDC')
974 plot_recovery_rate_per_acceptance_threshold(run_Lorenz, target='talk')
975 plot_recovery_rate_detailed_Lorenz(target='talk')
978if __name__ == "__main__":
979 import argparse
981 parser = argparse.ArgumentParser()
982 parser.add_argument(
983 '--target',
984 choices=['adaptivity', 'resilience', 'thesis', 'notes', 'SIAM_CSE23', 'TIME_X_website', 'TUHH_seminar'],
985 type=str,
986 )
987 args = parser.parse_args()
989 if args.target == 'adaptivity':
990 make_plots_for_adaptivity_paper()
991 elif args.target == 'resilience':
992 make_plots_for_resilience_paper()
993 elif args.target == 'thesis':
994 make_plots_for_thesis()
995 elif args.target == 'notes':
996 make_plots_for_notes()
997 elif args.target == 'SIAM_CSE23':
998 make_plots_for_SIAM_CSE23()
999 elif args.target == 'TIME_X_website':
1000 make_plots_for_TIME_X_website()
1001 elif args.target == 'TUHH_seminar':
1002 make_plots_for_TUHH_seminar()
1003 else:
1004 raise NotImplementedError(f'Don\'t know how to make plots for target {args.target}')