Coverage for pySDC/projects/Resilience/extrapolation_within_Q.py: 0%

53 statements  

« prev     ^ index     » next       coverage.py v7.5.0, created at 2024-04-29 09:02 +0000

1import matplotlib.pyplot as plt 

2import numpy as np 

3 

4from pySDC.implementations.convergence_controller_classes.estimate_extrapolation_error import ( 

5 EstimateExtrapolationErrorWithinQ, 

6) 

7from pySDC.implementations.hooks.log_errors import LogLocalErrorPostStep 

8from pySDC.helpers.stats_helper import get_sorted 

9 

10from pySDC.projects.Resilience.piline import run_piline 

11from pySDC.projects.Resilience.advection import run_advection 

12from pySDC.projects.Resilience.vdp import run_vdp 

13 

14 

15def multiple_runs(prob, dts, num_nodes, quad_type='RADAU-RIGHT', QI='LU', useMPI=False): 

16 """ 

17 Make multiple runs of a specific problem and record vital error information 

18 

19 Args: 

20 prob (function): A problem from the resilience project to run 

21 dts (list): The step sizes to run with 

22 num_nodes (int): Number of nodes 

23 quad_type (str): Type of nodes 

24 

25 Returns: 

26 dict: Errors for multiple runs 

27 int: Order of the collocation problem 

28 """ 

29 description = {} 

30 description['level_params'] = {'restol': 1e-10} 

31 description['step_params'] = {'maxiter': 99} 

32 description['sweeper_params'] = {'num_nodes': num_nodes, 'quad_type': quad_type} 

33 description['convergence_controllers'] = {EstimateExtrapolationErrorWithinQ: {}} 

34 

35 if useMPI: 

36 from pySDC.implementations.sweeper_classes.generic_implicit_MPI import generic_implicit_MPI, MPI 

37 

38 description['sweeper_class'] = generic_implicit_MPI 

39 description['sweeper_params']['comm'] = MPI.COMM_WORLD.Split(MPI.COMM_WORLD.rank < num_nodes) 

40 if MPI.COMM_WORLD.rank > num_nodes: 

41 return None 

42 

43 if prob.__name__ == 'run_advection': 

44 description['problem_params'] = {'order': 6, 'stencil_type': 'center'} 

45 

46 res = {} 

47 

48 for dt in dts: 

49 description['level_params']['dt'] = dt 

50 

51 stats, controller, _ = prob(custom_description=description, Tend=5.0 * dt, hook_class=LogLocalErrorPostStep) 

52 

53 res[dt] = {} 

54 res[dt]['e_loc'] = max([me[1] for me in get_sorted(stats, type='e_local_post_step')]) 

55 res[dt]['e_ex'] = max([me[1] for me in get_sorted(stats, type='error_extrapolation_estimate')]) 

56 

57 coll_order = controller.MS[0].levels[0].sweep.coll.order 

58 return res, coll_order 

59 

60 

61def plot_and_compute_order(ax, res, num_nodes, coll_order): 

62 """ 

63 Plot and compute the order from the multiple runs ran with `multiple_runs`. Also, it is tested if the expected order 

64 is reached for the respective errors. 

65 

66 Args: 

67 ax (Matplotlib.pyplot.axes): Somewhere to plot 

68 res (dict): Result from `multiple_runs` 

69 num_nodes (int): Number of nodes 

70 coll_order (int): Order of the collocation problem 

71 

72 Returns: 

73 None 

74 """ 

75 dts = np.array(list(res.keys())) 

76 keys = list(res[dts[0]].keys()) 

77 

78 # local error is one order higher than global error 

79 expected_order = { 

80 'e_loc': coll_order + 1, 

81 'e_ex': num_nodes + 1, 

82 } 

83 

84 for key in keys: 

85 errors = np.array([res[dt][key] for dt in dts]) 

86 

87 mask = np.logical_and(errors < 1e-3, errors > 1e-10) 

88 order = np.log(errors[mask][1:] / errors[mask][:-1]) / np.log(dts[mask][1:] / dts[mask][:-1]) 

89 

90 if ax is not None: 

91 ax.loglog(dts, errors, label=f'{key}: order={np.mean(order):.2f}') 

92 

93 assert np.isclose( 

94 np.mean(order), expected_order[key], atol=0.5 

95 ), f'Expected order {expected_order[key]} for {key}, but got {np.mean(order):.2e}!' 

96 

97 if ax is not None: 

98 ax.legend(frameon=False) 

99 

100 

101def check_order(ax, prob, dts, num_nodes, quad_type, **kwargs): 

102 """ 

103 Check the order by calling `multiple_runs` and then `plot_and_compute_order`. 

104 

105 Args: 

106 ax (Matplotlib.pyplot.axes): Somewhere to plot 

107 prob (function): A problem from the resilience project to run 

108 dts (list): The step sizes to run with 

109 num_nodes (int): Number of nodes 

110 quad_type (str): Type of nodes 

111 """ 

112 res, coll_order = multiple_runs(prob, dts, num_nodes, quad_type, **kwargs) 

113 plot_and_compute_order(ax, res, num_nodes, coll_order) 

114 

115 

116def main(): 

117 fig, ax = plt.subplots() 

118 num_nodes = 3 

119 quad_type = 'RADAU-RIGHT' 

120 check_order(ax, run_advection, [5e-1, 1e-1, 5e-2, 1e-2], num_nodes, quad_type, QI='MIN', useMPI=True) 

121 plt.show() 

122 

123 

124if __name__ == "__main__": 

125 main()