def evaluateError(cls, w, coeff_field, pde, f, zeta, gamma, ceta, cQ, newmi_add_maxm, maxh=0.1, quadrature_degree= -1, projection_degree_increase=1, refine_projection_mesh=1):
"""Evaluate EGSZ Error (7.5)."""
logger.debug("starting evaluateError")
# define store function for timings
from functools import partial
def _store_stats(val, key, stats):
stats[key] = val
timing_stats = {}
with timing(msg="ResidualEstimator.evaluateResidualEstimator", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-RESIDUAL", stats=timing_stats)):
resind, reserror = ResidualEstimator.evaluateResidualEstimator(w, coeff_field, pde, f, quadrature_degree)
logger.debug("starting evaluateProjectionEstimator")
with timing(msg="ResidualEstimator.evaluateProjectionError", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-PROJECTION", stats=timing_stats)):
projind, projerror = ResidualEstimator.evaluateProjectionError(w, coeff_field, pde, maxh, True, projection_degree_increase, refine_projection_mesh)
logger.debug("starting evaluateInactiveProjectionError")
with timing(msg="ResidualEstimator.evaluateInactiveMIProjectionError", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-INACTIVE-MI", stats=timing_stats)):
mierror = ResidualEstimator.evaluateInactiveMIProjectionError(w, coeff_field, pde, maxh, newmi_add_maxm)
eta = sum(reserror[mu] ** 2 for mu in reserror)
delta = sum(projerror[mu] ** 2 for mu in projerror)
delta_inactive_mi = sum(v[1] ** 2 for v in mierror)
est1 = ceta / sqrt(1 - gamma) * sqrt(eta)
est2 = cQ / sqrt(1 - gamma) * sqrt(delta + delta_inactive_mi)
est3 = cQ * sqrt(zeta / (1 - gamma))
est4 = zeta / (1 - gamma)
# xi = (ceta / sqrt(1 - gamma) * sqrt(eta) + cQ / sqrt(1 - gamma) * sqrt(delta)
# + cQ * sqrt(zeta / (1 - gamma))) ** 2 + zeta / (1 - gamma)
xi = (est1 + est2 + est3) ** 2 + est4
logger.info("Total Residual ERROR Factors: A1=%s A2=%s A3=%s A4=%s", ceta / sqrt(1 - gamma), cQ / sqrt(1 - gamma), cQ * sqrt(zeta / (1 - gamma)), zeta / (1 - gamma))
return (xi, resind, projind, mierror, (est1, est2, est3, est4), (eta, delta, zeta), timing_stats)
def get_projected_solution(w, mu, proj_basis):
# TODO: obfuscated method call since project is not obvious in the interface of MultiVector! This should be separated more clearly!
# print "sampling.get_projected_solution"
# print w[mu].num_sub_spaces
# print proj_basis.num_sub_spaces
with timing(msg="get_projected_solution (%s --- %i)" % (str(mu), w[mu].dim), logfunc=logger.debug):
w_proj = w.project(w[mu], proj_basis)
return w_proj
def compute_direct_sample_solution(pde, RV_samples, coeff_field, A, maxm, proj_basis, cache=None):
try:
A0 = cache.A
A_m = cache.A_m
b = cache.b
logger.debug("compute_direct_sample_solution: CACHE USED")
print "CACHE USED"
except AttributeError:
with timing(msg="direct_sample_sol: compute A_0, b", logfunc=logger.info):
a = coeff_field.mean_func
A0 = pde.assemble_lhs(basis=proj_basis, coeff=a, withDirichletBC=False)
b = pde.assemble_rhs(basis=proj_basis, coeff=a, withDirichletBC=False)
A_m = [None] * maxm
logger.debug("compute_direct_sample_solution: CACHE NOT USED")
print "CACHE NOT USED"
if cache is not None:
cache.A = A0
cache.A_m = A_m
cache.b = b
with timing(msg="direct_sample_sol: compute A_m", logfunc=logger.info):
A = A0.copy()
for m in range(maxm):
if A_m[m] is None:
a_m = coeff_field[m][0]
A_m[m] = pde.assemble_lhs(basis=proj_basis, coeff=a_m, withDirichletBC=False)
A += RV_samples[m] * A_m[m]
with timing(msg="direct_sample_sol: apply BCs", logfunc=logger.info):
A, b = pde.apply_dirichlet_bc(proj_basis._fefs, A, b)
with timing(msg="direct_sample_sol: solve linear system", logfunc=logger.info):
X = 0 * b
logger.info("compute_direct_sample_solution with %i dofs" % b.size())
solve(A, X, b)
return FEniCSVector(Function(proj_basis._fefs, X))
def compute_parametric_sample_solution(RV_samples, coeff_field, w, proj_basis, cache=None):
with timing(msg="parametric_sample_sol", logfunc=logger.info):
Lambda = w.active_indices()
sample_map, _ = coeff_field.sample_realization(Lambda, RV_samples)
# sum up (stochastic) solution vector on reference function space wrt samples
if cache is None:
sample_sol = sum(get_projected_solution(w, mu, proj_basis) * sample_map[mu] for mu in Lambda)
else:
try:
projected_sol = cache.projected_sol
except AttributeError:
projected_sol = {mu: get_projected_solution(w, mu, proj_basis) for mu in Lambda}
cache.projected_sol = projected_sol
sample_sol = sum(projected_sol[mu] * sample_map[mu] for mu in Lambda)
return sample_sol
def AdaptiveSolver(A, coeff_field, pde,
mis, w0, mesh0, degree,
gamma=0.9,
cQ=1.0,
ceta=6.0,
# marking parameters
theta_eta=0.4, # residual marking bulk parameter
theta_zeta=0.1, # projection marking threshold factor
min_zeta=1e-8, # minimal projection error to be considered
maxh=0.1, # maximal mesh width for projection maximum norm evaluation
newmi_add_maxm=20, # maximal search length for new new multiindices (to be added to max order of solution w)
theta_delta=10.0, # number new multiindex activation bound
max_Lambda_frac=1 / 10, # max fraction of |Lambda| for new multiindices
marking_strategy="SEPARATE with CELLPROJECTION", # separate (as initially in EGSZ) or relative marking wrt overall error, projection refinement based on cell or mesh errors
# residual error
quadrature_degree= -1,
# projection error
projection_degree_increase=1,
refine_projection_mesh=1,
# pcg solver
pcg_eps=1e-6,
pcg_maxiter=100,
# adaptive algorithm threshold
error_eps=1e-2,
# refinements
max_refinements=5,
max_dof=1e10,
do_refinement={"RES":True, "PROJ":True, "MI":False},
do_uniform_refinement=False,
w_history=None,
sim_stats=None):
# define store function for timings
from functools import partial
def _store_stats(val, key, stats):
stats[key] = val
# define tuple type
EstimatorData = namedtuple('EstimatorData', ['xi', 'gamma', 'cQ', 'ceta'])
# get rhs
f = pde.f
# setup w and statistics
w = w0
if sim_stats is None:
assert w_history is None or len(w_history) == 0
sim_stats = []
try:
start_iteration = max(len(sim_stats) - 1, 0)
except:
start_iteration = 0
logger.info("START/CONTINUE EXPERIMENT at iteration %i", start_iteration)
# data collection
import resource
refinement = None
for refinement in range(start_iteration, max_refinements + 1):
logger.info("************* REFINEMENT LOOP iteration %i (of %i or max_dof %i) *************", refinement, max_refinements, max_dof)
# memory usage info
logger.info("\n======================================\nMEMORY USED: " + str(resource.getrusage(resource.RUSAGE_SELF).ru_maxrss) + "\n======================================\n")
# pcg solve
# ---------
stats = {}
with timing(msg="pcg_solve", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-PCG", stats=stats)):
w, zeta = pcg_solve(A, w, coeff_field, pde, stats, pcg_eps, pcg_maxiter)
logger.info("DIM of w = %s", w.dim)
if w_history is not None and (refinement == 0 or start_iteration < refinement):
w_history.append(w)
# error evaluation
# ----------------
# residual and projection errors
logger.debug("evaluating ResidualEstimator.evaluateError")
with timing(msg="ResidualEstimator.evaluateError", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-ESTIMATOR", stats=stats)):
xi, resind, projind, mierror, estparts, errors, timing_stats = ResidualEstimator.evaluateError(w, coeff_field, pde, f, zeta, gamma, ceta, cQ,
newmi_add_maxm, maxh, quadrature_degree, projection_degree_increase,
refine_projection_mesh)
reserrmu = [(mu, sqrt(sum(resind[mu].coeffs ** 2))) for mu in resind.keys()]
projerrmu = [(mu, sqrt(sum(projind[mu].coeffs ** 2))) for mu in projind.keys()]
res_part, proj_part, pcg_part = estparts[0], estparts[1], estparts[2]
err_res, err_proj, err_pcg = errors[0], errors[1], errors[2]
logger.info("Overall Estimator Error xi = %s while residual error is %s, projection error is %s, pcg error is %s", xi, res_part, proj_part, pcg_part)
stats.update(timing_stats)
stats["EST"] = xi
stats["RES-PART"] = res_part
stats["PROJ-PART"] = proj_part
stats["PCG-PART"] = pcg_part
stats["ERR-RES"] = err_res
stats["ERR-PROJ"] = err_proj
stats["ERR-PCG"] = err_pcg
stats["ETA-ERR"] = errors[0]
stats["DELTA-ERR"] = errors[1]
stats["ZETA-ERR"] = errors[2]
stats["RES-mu"] = reserrmu
stats["PROJ-mu"] = projerrmu
stats["PROJ-MAX-ZETA"] = 0
stats["PROJ-MAX-INACTIVE-ZETA"] = 0
stats["MARKING-RES"] = 0
stats["MARKING-PROJ"] = 0
stats["MARKING-MI"] = 0
stats["TIME-MARKING"] = 0
stats["MI"] = [(mu, vec.basis.dim) for mu, vec in w.iteritems()]
if refinement == 0 or start_iteration < refinement:
sim_stats.append(stats)
# print "SIM_STATS:", sim_stats[refinement]
logger.debug("squared error components: eta=%s delta=%s zeta=%", errors[0], errors[1], errors[2])
# exit when either error threshold or max_refinements or max_dof is reached
if refinement > max_refinements:
logger.info("SKIPPING REFINEMENT after FINAL SOLUTION in ITERATION %i", refinement)
break
if sim_stats[refinement]["DOFS"] >= max_dof:
logger.info("REACHED %i DOFS, EXITING refinement loop", sim_stats[refinement]["DOFS"])
break
if xi <= error_eps:
logger.info("error reached requested accuracy, xi=%f", xi)
break
# marking
# -------
if refinement < max_refinements:
if not do_uniform_refinement:
logger.debug("starting Marking.mark")
estimator_data = EstimatorData(xi=xi, gamma=gamma, cQ=cQ, ceta=ceta)
mesh_markers_R, mesh_markers_P, new_multiindices, proj_zeta, new_multiindices_all = Marking.mark(resind, projind, mierror, w.max_order,
theta_eta, theta_zeta, theta_delta,
min_zeta, maxh, max_Lambda_frac,
estimator_data, marking_strategy)
sim_stats[-1]["PROJ-MAX-ZETA"] = proj_zeta[0]
sim_stats[-1]["PROJ-MAX-INACTIVE-ZETA"] = proj_zeta[1]
sim_stats[-1]["PROJ-INACTIVE-ZETA"] = new_multiindices_all
# assert len(new_multiindices_all) == 0 or proj_zeta[1] == max([v for v in new_multiindices_all.values()])
logger.info("PROJECTION error values: max_zeta = %s and max_inactive_zeta = %s with threshold factor theta_zeta = %s (=%s)",
proj_zeta[0], proj_zeta[1], theta_zeta, theta_zeta * proj_zeta[0])
logger.info("MARKING will be carried out with %s (res) + %s (proj) cells and %s new multiindices",
sum([len(cell_ids) for cell_ids in mesh_markers_R.itervalues()]),
sum([len(cell_ids) for cell_ids in mesh_markers_P.itervalues()]), len(new_multiindices))
stats["MARKING-RES"] = sum([len(cell_ids) for cell_ids in mesh_markers_R.itervalues()])
stats["MARKING-PROJ"] = sum([len(cell_ids) for cell_ids in mesh_markers_P.itervalues()])
stats["MARKING-MI"] = len(new_multiindices)
if do_refinement["RES"]:
mesh_markers = mesh_markers_R.copy()
else:
mesh_markers = defaultdict(set)
logger.info("SKIP residual refinement")
if do_refinement["PROJ"]:
for mu, cells in mesh_markers_P.iteritems():
if len(cells) > 0:
mesh_markers[mu] = mesh_markers[mu].union(cells)
else:
logger.info("SKIP projection refinement")
if not do_refinement["MI"] or refinement == max_refinements:
new_multiindices = {}
logger.info("SKIP new multiindex refinement")
else:
logger.info("UNIFORM REFINEMENT active")
mesh_markers = {}
for mu, vec in w.iteritems():
from dolfin import cells
mesh_markers[mu] = list([c.index() for c in cells(vec._fefunc.function_space().mesh())])
new_multiindices = {}
# carry out refinement of meshes
with timing(msg="Marking.refine", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-MARKING", stats=stats)):
Marking.refine(w, mesh_markers, new_multiindices.keys(), partial(setup_vector, pde=pde, mesh=mesh0, degree=degree))
if refinement:
logger.info("ENDED refinement loop after %i of %i refinements with %i dofs and %i active multiindices",
refinement, max_refinements, sim_stats[refinement]["DOFS"], len(sim_stats[refinement]["MI"]))
# except Exception as ex:
# import pickle
# logger.error("EXCEPTION during AdaptiveSolver: %s", str(ex))
# print "DIM of w:", w.dim
# if not w_history is None:
# w_history.append(w)
# wname = "W-PCG-FAILED.pkl"
# try:
# with open(wname, 'wb') as fout:
# pickle.dump(w, fout)
# except Exception as ex:
# logger.error("NEXT EXCEPTION %s", str(ex))
# logger.info("exported last multivector w to %s", wname)
# finally:
return w, sim_stats
def run_mc(err, w, pde, A, coeff_field, mesh0, ref_maxm, MC_N, MC_HMAX, param_sol_cache=None, direct_sol_cache=None, stored_rv_samples=None, quadrature_degree = -1):
# create reference mesh and function space
sub_spaces = w[Multiindex()].basis.num_sub_spaces
degree = w[Multiindex()].basis.degree
projection_basis = get_projection_basis(mesh0, mesh_refinements=0, degree=degree, sub_spaces=sub_spaces)
logger.info("projection_basis dim = %i \t hmin of mi[0] = %s, reference mesh = (%s, %s)", projection_basis.dim, w[Multiindex()].basis.minh, projection_basis.minh, projection_basis.maxh)
# get realization of coefficient field
err_L2, err_H1 = 0, 0
# set quadrature degree
if quadrature_degree > -1:
quadrature_degree_old = parameters["form_compiler"]["quadrature_degree"]
parameters["form_compiler"]["quadrature_degree"] = quadrature_degree
logger.debug("MC error sampling quadrature order = " + str(quadrature_degree))
# setup caches for sample solutions
# param_sol_cache = None #param_sol_cache or MCCache()
# direct_sol_cache = None #direct_sol_cache or MCCache()
logger.info("---- MC caches %s/%s ----", param_sol_cache, direct_sol_cache)
# main MC loop
for i in range(MC_N):
logger.info("---- MC Iteration %i/%i ----", i + 1 , MC_N)
# create new samples if required or reuse existing samples
if stored_rv_samples is None or len(stored_rv_samples) <= i:
sample_rvs = coeff_field.sample_rvs()
RV_samples = [sample_rvs[j] for j in range(max(w.max_order, ref_maxm))]
if stored_rv_samples is not None:
stored_rv_samples.append(RV_samples)
else:
RV_samples = stored_rv_samples[i]
logger.info("-- RV_samples: %s", RV_samples)
# evaluate solutions
with timing(msg="parameteric sample solution", logfunc=logger.info):
sample_sol_param = compute_parametric_sample_solution(RV_samples, coeff_field, w, projection_basis, param_sol_cache)
with timing(msg="direct sample solution", logfunc=logger.info):
sample_sol_direct = compute_direct_sample_solution(pde, RV_samples, coeff_field, A, ref_maxm, projection_basis, direct_sol_cache)
# evaluate errors
with timing(msg="L2_err_1", logfunc=logger.info):
cerr_L2 = error_norm(sample_sol_param._fefunc, sample_sol_direct._fefunc, "L2")
with timing(msg="H1A_err_1", logfunc=logger.info):
cerr_H1 = error_norm(sample_sol_param._fefunc, sample_sol_direct._fefunc, pde.energy_norm)
# cerr_H1 = errornorm(sample_sol_param._fefunc, sample_sol_direct._fefunc, "H1")
logger.debug("-- current error L2 = %s H1A = %s", cerr_L2, cerr_H1)
err_L2 += 1.0 / MC_N * cerr_L2
err_H1 += 1.0 / MC_N * cerr_H1
if i + 1 == MC_N:
# deterministic part
with timing(msg="direct a0", logfunc=logger.info):
sample_sol_direct_a0 = compute_direct_sample_solution(pde, RV_samples, coeff_field, A, 0, projection_basis, direct_sol_cache)
with timing(msg="L2_err_2", logfunc=logger.info):
L2_a0 = error_norm(sample_sol_param._fefunc, sample_sol_direct_a0._fefunc, "L2")
with timing(msg="H1A_err_2", logfunc=logger.info):
H1_a0 = error_norm(sample_sol_param._fefunc, sample_sol_direct_a0._fefunc, pde.energy_norm)
# H1_a0 = errornorm(sample_sol_param._fefunc, sample_sol_direct_a0._fefunc, "H1")
logger.debug("-- DETERMINISTIC error L2 = %s H1A = %s", L2_a0, H1_a0)
# stochastic part
sample_sol_direct_am = sample_sol_direct - sample_sol_direct_a0
logger.debug("-- STOCHASTIC norm L2 = %s H1 = %s", sample_sol_direct_am.norm("L2"), sample_sol_direct_am.norm("H1"))
# restore quadrature degree
if quadrature_degree > -1:
parameters["form_compiler"]["quadrature_degree"] = quadrature_degree_old
logger.info("MC Error: L2: %s, H1A: %s", err_L2, err_H1)
err.append((err_L2, err_H1, L2_a0, H1_a0))
def AdaptiveSolver(A, coeff_field, pde,
mis, w0, mesh0, degree,
# marking parameters
rho=1.0, # tail factor
theta_x=0.4, # residual marking bulk parameter
theta_y=0.4, # tail bound marking bulk paramter
maxh=0.1, # maximal mesh width for coefficient maximum norm evaluation
add_maxm=100, # maximal search length for new new multiindices (to be added to max order of solution w)
# estimator
estimator_type = "RESIDUAL",
quadrature_degree= -1,
# pcg solver
pcg_eps=1e-6,
pcg_maxiter=100,
# adaptive algorithm threshold
error_eps=1e-2,
# refinements
max_refinements=5,
max_dof=1e10,
do_refinement={"RES":True, "TAIL":True, "OSC":True},
do_uniform_refinement=False,
refine_osc_factor=1.0,
w_history=None,
sim_stats=None):
# define store function for timings
def _store_stats(val, key, stats):
stats[key] = val
# get rhs
f = pde.f
# setup w and statistics
w = w0
if sim_stats is None:
assert w_history is None or len(w_history) == 0
sim_stats = []
try:
start_iteration = max(len(sim_stats) - 1, 0)
except:
start_iteration = 0
logger.info("START/CONTINUE EXPERIMENT at iteration %i", start_iteration)
# data collection
import resource
refinement = None
for refinement in range(start_iteration, max_refinements + 1):
logger.info("************* REFINEMENT LOOP iteration {0} (of {1} or max dofs {2}) *************".format(refinement, max_refinements, max_dof))
# memory usage info
logger.info("\n======================================\nMEMORY USED: " + str(resource.getrusage(resource.RUSAGE_SELF).ru_maxrss) + "\n======================================\n")
# ---------
# pcg solve
# ---------
stats = {}
with timing(msg="pcg_solve", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-PCG", stats=stats)):
w, zeta = pcg_solve(A, w, coeff_field, pde, stats, pcg_eps, pcg_maxiter)
logger.info("DIM of w = %s", w.dim)
if w_history is not None and (refinement == 0 or start_iteration < refinement):
w_history.append(w)
# -------------------
# evaluate estimators
# -------------------
# evaluate estimate_y
logger.debug("evaluating upper tail bound")
with timing(msg="ResidualEstimator.evaluateUpperTailBound", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-TAIL", stats=stats)):
global_zeta, zeta, zeta_bar, eval_zeta_m = ResidualEstimator.evaluateUpperTailBound(w, coeff_field, pde, maxh, add_maxm)
# evaluate estimate_x
if estimator_type.upper() == "RESIDUAL":
# evaluate estimate_x
logger.debug("evaluating residual bound (residual)")
with timing(msg="ResidualEstimator.evaluateResidualEstimator", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-RES", stats=stats)):
global_eta, eta, eta_local = ResidualEstimator.evaluateResidualEstimator(w, coeff_field, pde, f, quadrature_degree)
elif estimator_type.upper() == "EQUILIBRATION_GLOBAL":
logger.debug("evaluating residual bound (global equilibration)")
with timing(msg="GlobalEquilibrationEstimator.evaluateEstimator", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-RES", stats=stats)):
global_eta, eta, eta_local, osc_global, osc_local = GlobalEquilibrationEstimator.evaluateEstimator\
(w, coeff_field, pde, f, quadrature_degree)
elif estimator_type.upper() == "EQUILIBRATION_LOCAL":
logger.debug("evaluating residual bound (global equilibration)")
with timing(msg="GlobalEquilibrationEstimator.evaluateEstimator", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-RES", stats=stats)):
global_eta, eta, eta_local, osc_global, osc_local = LocalEquilibrationEstimator.evaluateEstimator\
(w, coeff_field, pde, f, quadrature_degree)
else:
raise TypeError("invalid estimator type %s" %estimator_type.upper())
# set overall error
xi = sqrt(global_eta ** 2 + global_zeta ** 2)
logger.info("Overall Estimator Error xi = %s while spatial error is %s and tail error is %s", xi, global_eta, global_zeta)
# store simulation data
stats["ERROR-EST"] = xi
stats["ERROR-RES"] = global_eta
stats["ERROR-TAIL"] = global_zeta
stats["MARKING-RES"] = 0
stats["MARKING-MI"] = 0
# stats["MARKING-OSC"] = 0
stats["CADELTA"] = 0
stats["TIME-MARK-RES"] = 0
stats["TIME-REFINE-RES"] = 0
stats["TIME-MARK-TAIL"] = 0
stats["TIME-REFINE-TAIL"] = 0
stats["TIME-REFINE-OSC"] = 0
stats["MI"] = [mu for mu in w.active_indices()]
stats["DIM"] = w.dim
if refinement == 0 or start_iteration < refinement:
sim_stats.append(stats)
print "SIM_STATS:", sim_stats[refinement]
# exit when either error threshold or max_refinements or max_dof is reached
if refinement > max_refinements:
logger.info("SKIPPING REFINEMENT after FINAL SOLUTION in ITERATION %i", refinement)
break
if sim_stats[refinement]["DOFS"] >= max_dof:
logger.info("REACHED %i DOFS, EXITING refinement loop", sim_stats[refinement]["DOFS"])
break
if xi <= error_eps:
logger.info("SKIPPING REFINEMENT since ERROR REACHED requested ACCURACY, xi=%f", xi)
break
# -----------------------------------
# mark and refine and activate new mi
# -----------------------------------
if refinement < max_refinements:
logger.debug("START marking === %s", str(do_refinement))
# === mark x ===
res_marked = False
if do_refinement["RES"]:
cell_ids = []
if not do_uniform_refinement:
if global_eta > rho * global_zeta or not do_refinement["TAIL"]:
logger.info("REFINE RES")
with timing(msg="Marking.mark_x", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-MARK-RES", stats=stats)):
cell_ids = Marking.mark_x(global_eta, eta_local, theta_x)
res_marked = True
else:
logger.info("SKIP REFINE RES -> mark stochastic modes instead")
else:
# uniformly refine mesh
logger.info("UNIFORM refinement RES")
cell_ids = [c.index() for c in cells(w.basis._fefs.mesh())]
res_marked = True
else:
logger.info("SKIP residual refinement")
# refine mesh
if res_marked:
logger.debug("w.dim BEFORE refinement: %s", w.dim)
with timing(msg="Marking.refine_x", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-REFINE-RES", stats=stats)):
w = Marking.refine_x(w, cell_ids)
logger.debug("w.dim AFTER refinement: %s", w.dim)
# === mark y ===
if do_refinement["TAIL"] and not res_marked:
logger.info("REFINE TAIL")
with timing(msg="Marking.mark_y", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-MARK-TAIL", stats=stats)):
new_mi = Marking.mark_y(w.active_indices(), zeta, eval_zeta_m, theta_y, add_maxm)
# add new multiindices
with timing(msg="Marking.refine_y", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-REFINE-TAIL", stats=stats)):
Marking.refine_y(w, new_mi)
else:
new_mi = []
logger.info("SKIP tail refinement")
# === uniformly refine for coefficient function oscillations ===
if do_refinement["OSC"]:
logger.info("REFINE OSC")
with timing(msg="Marking.refine_osc", logfunc=logger.info, store_func=partial(_store_stats, key="TIME-REFINE-OSC", stats=stats)):
w, maxh, Cadelta = Marking.refine_osc(w, coeff_field, refine_osc_factor)
logger.info("coefficient oscillations require maxh %f with current mesh maxh %f and Cadelta %f" % (maxh, w.basis.basis.mesh.hmax(), Cadelta))
stats["CADELTA"] = Cadelta
else:
logger.info("SKIP oscillation refinement")
logger.info("MARKING was carried out with %s (res) cells and %s (mi) new multiindices", len(cell_ids), len(new_mi))
stats["MARKING-RES"] = len(cell_ids)
stats["MARKING-MI"] = len(new_mi)
if refinement:
logger.info("ENDED refinement loop after %i of (max) %i refinements with %i dofs and %i active multiindices",
refinement, max_refinements, sim_stats[refinement]["DOFS"], len(sim_stats[refinement]["MI"]))
return w, sim_stats