def Eigen_Value_Solver(Hamiltonian, input_par, l, energy, Viewer):
number_of_eigenvalues = 1
k = np.sqrt(2 * energy)
EV_Solver = SLEPc.EPS().create(comm=PETSc.COMM_WORLD)
EV_Solver.setOperators(Hamiltonian) ##pass the hamiltonian to the
# EV_Solver.setType(SLEPc.EPS.Type.JD)
EV_Solver.setProblemType(SLEPc.EPS.ProblemType.NHEP)
EV_Solver.setTolerances(input_par["tolerance"], PETSc.DECIDE)
EV_Solver.setWhichEigenpairs(EV_Solver.Which.TARGET_REAL)
EV_Solver.setTarget(energy)
size_of_matrix = PETSc.Mat.getSize(Hamiltonian)
dimension_size = int(size_of_matrix[0]) * 0.1
EV_Solver.setDimensions(number_of_eigenvalues, PETSc.DECIDE,
dimension_size)
EV_Solver.solve() ##solve the eigenvalue problem
eigen_states = [[], []]
for i in range(number_of_eigenvalues):
eigen_vector = Hamiltonian.getVecLeft()
eigen_state = EV_Solver.getEigenpair(i, eigen_vector)
eigen_vector.setName("Psi_" + str(l) + "_" + str(i))
Viewer.view(eigen_vector)
energy = PETSc.Vec().createMPI(1, comm=PETSc.COMM_WORLD)
energy.setValue(0, eigen_state)
energy.setName("Energy_" + str(l) + "_" + str(i))
energy.assemblyBegin()
energy.assemblyEnd()
Viewer.view(energy)
def compute_inf_sup_constant(spaces, b, inner_products):
V, Q = spaces
m, n = inner_products
Z = V * Q
u, p = firedrake.TrialFunctions(Z)
v, q = firedrake.TestFunctions(Z)
A = assemble(-(m(u, v) + b(v, p) + b(u, q)), mat_type='aij').M.handle
M = assemble(n(p, q), mat_type='aij').M.handle
opts = PETSc.Options()
opts.setValue('eps_gen_hermitian', None)
opts.setValue('eps_target_real', None)
opts.setValue('eps_smallest_magnitude', None)
opts.setValue('st_type', 'sinvert')
opts.setValue('st_ksp_type', 'preonly')
opts.setValue('st_pc_type', 'lu')
opts.setValue('st_pc_factor_mat_solver_type', 'mumps')
opts.setValue('eps_tol', 1e-8)
num_values = 1
eigensolver = SLEPc.EPS().create(comm=firedrake.COMM_WORLD)
eigensolver.setDimensions(num_values)
eigensolver.setOperators(A, M)
eigensolver.setFromOptions()
eigensolver.solve()
Vr, Vi = A.getVecs()
λ = eigensolver.getEigenpair(0, Vr, Vi)
return λ
def slepc_eigh(A, B, Z=None, is_hermitian=True):
'''GEVP by SLEPc via lapack'''
A, B = as_backend_type(A).mat(), as_backend_type(B).mat()
# Setup the eigensolver
E = SLEPc.EPS().create()
E.setOperators(A, B)
E.setProblemType(SLEPc.EPS.ProblemType.GHEP)
E.setType('lapack')
print 'Z is', Z
if Z is not None:
print 'Set'
Z = [as_backend_type(z).vec() for z in Z]
E.setDeflationSpace(Z)
# type is -eps_type
E.solve()
eigw = []
for i in range(E.getConverged()):
eigw.append(E.getEigenvalue(i).real)
eigw = np.array(eigw)
idx = np.argsort(eigw)
eigw = eigw[idx]
v = A.createVecRight()
eigv = []
for i in idx:
E.getEigenvector(i, v)
eigv.append(v.array.real.tolist())
print v.dot(Z[0])
eigv = np.array(eigv)
return eigw, eigv.T
def main():
opts = PETSc.Options()
n = opts.getInt('n', 32)
m = opts.getInt('m', 32)
Print("2-D Laplacian Eigenproblem solved with contour integral, "
"N=%d (%dx%d grid)\n" % (m * n, m, n))
A = construct_operator(m, n)
# Solver object
E = SLEPc.EPS().create()
E.setOperators(A)
E.setProblemType(SLEPc.EPS.ProblemType.HEP)
E.setType(SLEPc.EPS.Type.CISS)
# Define region of interest
R = E.getRG()
R.setType(SLEPc.RG.Type.ELLIPSE)
R.setEllipseParameters(0.0, 0.2, 0.1)
E.setFromOptions()
# Compute solution
E.solve()
# Print solution
vw = PETSc.Viewer.STDOUT()
vw.pushFormat(PETSc.Viewer.Format.ASCII_INFO_DETAIL)
E.errorView(viewer=vw)
vw.popFormat()
def slep_powerit(Ptp, n):
Ptp.transpose()
E = SLEPc.EPS()
E.create()
E.setOperators(Ptp)
E.setProblemType(SLEPc.EPS.ProblemType.NHEP)
E.setTolerances(tol=1e-12)
E.setFromOptions()
E.solve()
Print = PETSc.Sys.Print
its = E.getIterationNumber()
Print("Number of iterations of the method: %d" % its)
eps_type = E.getType()
Print("Solution method: %s" % eps_type)
nconv = E.getConverged()
Print("Number of converged eigenpairs %d" % nconv)
tol, maxit = E.getTolerances()
Print("Stopping condition: tol=%.4g, maxit=%d" % (tol, maxit))
if nconv > 0:
vpost, wpost = Ptp.getVecs()
vposti, wposti = Ptp.getVecs()
k = E.getEigenpair(0, vpost, vposti)
Print("Valeur propre" + str(k))
return abs(vpost)
def _create_eigenproblem_solver(self, A, M, num, problem_type, method_type,
which, tol, maxit, shift_invert):
"""
Create a SLEPc eigenproblem solver with the operator
"""
# XXX TODO: This import should actually happen at the top, but on some
# systems it seems to be slightly non-trivial to install
# slepc4py, and since we don't use it for the default eigen-
# value methods, it's better to avoid raising an ImportError
# which forces users to try and install it. -- Max, 20.3.2014
from slepc4py import SLEPc
E = SLEPc.EPS()
E.create()
E.setOperators(A, M)
E.setProblemType(getattr(SLEPc.EPS.ProblemType, problem_type))
E.setType(getattr(SLEPc.EPS.Type, method_type))
E.setWhichEigenpairs(getattr(SLEPc.EPS.Which, which))
E.setDimensions(nev=num)
E.setTolerances(tol, maxit)
if shift_invert == True:
st = E.getST()
st.setType(SLEPc.ST.Type.SINVERT)
st.setShift(0.0)
return E
def eigensolver_setup(self, prefix=None):
E = SLEPc.EPS()
E.create()
E.setType(SLEPc.EPS.Type.KRYLOVSCHUR)
if hasattr(self, 'M'):
E.setProblemType(SLEPc.EPS.ProblemType.GHEP)
else:
E.setProblemType(SLEPc.EPS.ProblemType.HEP)
# if self.initial_guess:
# E.setInitialSpace(self.initial_guess)
E.setWhichEigenpairs(E.Which.TARGET_REAL)
E.setTarget(-.1)
st = E.getST()
st.setType('sinvert')
st.setShift(-1.e-3)
if prefix:
E.setOptionsPrefix(prefix)
self.set_options(self.slepc_options)
E.setFromOptions()
return E
def main():
args = docopt("""
Usage:
learn_spectral_embeddings.py [options] <adjacency_matrix_path> <type_of_laplacian> <output_path>
Options:
--pow NUM Every non-zero value in adjacency matrix will be scaled^{pow} [default: 1.0]
--max_iter NUM Maximum number of iterations of LOBPCG algorithm [default: 100]
--dim NUM Number of eigen-pairs to return [default: 500]
--verbosity NUM Verbosity level of LOBPCG solver [default: 0]
--pmi Turn adjacency matrix into PMI-based adjacency matrix
--neg NUM Negative sampling for PMI-based adjacency matrix [default: 1]
--scale_weights Scale weights of the adjacency matrix between 0 and 1
--tune_rhoB Solve quadratic eigenproblem to find better rhoB estimation
""")
start = time.time()
adjacency_matrix_path = args["<adjacency_matrix_path>"]
adjacency_matrix = load_npz(adjacency_matrix_path + ".adjacency.npz")
_, iw = load_vocabulary(adjacency_matrix_path + ".words.vocab")
### Build laplacian
n = adjacency_matrix.shape[0]
degrees = np.asarray(adjacency_matrix.sum(axis=1)).flatten()
D = scipy.sparse.spdiags(degrees, [0], n, n, format='csr')
L = D - adjacency_matrix
degrees_sqrt = np.sqrt(degrees)
DH = scipy.sparse.spdiags(1.0 / degrees_sqrt, [0], n, n, format='csr')
L = DH.dot(L.dot(DH))
### Build PETSc sparse matrix
A = PETSc.Mat().create()
A.setSizes([n, n])
A.setFromOptions()
A.setUp()
inds = L.nonzero()
for row_ind, col_ind in np.stack(inds, axis=1):
value = L[row_ind, col_ind]
assert value != 0
A[row_ind, col_ind] = value
A.assemble()
### Solve eigenvalue problem
E = SLEPc.EPS()
E.create()
E.setOperators(A)
E.setProblemType(SLEPc.EPS.ProblemType.HEP)
E.setFromOptions()
E.setWhichEigenpairs(SLEPc.EPS.Which.SMALLEST_MAGNITUDE)
print("Time elapsed %f" % (time.time() - start))
def get_extreme_eigw(A, B, Z, params, which):
'''Use SLEPC for fish for extremal eigenvalues of GEVP(A, B)'''
is_symmetric = is_hermitian(A, tol=1E-8 * A.size(0))
assert is_symmetric
# From monolithic matrices to petsc
A, B = (as_backend_type(x).mat() for x in (A, B))
# Wrap
opts = PETSc.Options()
for key, value in params.items():
opts.setValue(key, None if value == 'none' else value)
# Setup the eigensolver
E = SLEPc.EPS().create()
if Z is not None:
Z = [as_backend_type(z).vec() for z in Z]
E.setDeflationSpace(Z)
E.setOperators(A, B)
# type is -eps_type
if is_symmetric:
E.setProblemType(SLEPc.EPS.ProblemType.GHEP)
else:
assert False
E.setProblemType(SLEPc.EPS.ProblemType.GNHEP)
# Whch
E.setWhichEigenpairs(which)
# For 'small' system the shifted problems will use direct solver
flag = 'slepc_direct'
# Otherwise using shift and invert spectral transformation with zero shift
# itertively
# NOTE: these can ge configured from commandiline
if A.size[0] > 2E6:
flag = 'slepc_iter'
ST = E.getST()
ST.setType('sinvert')
KSP = ST.getKSP()
KSP.setType('cg') # How to invert the B matrix
PC = KSP.getPC()
# NOTE: with these setting the cost is as high as direct
PC.setType('lu')
PC.setFactorSolverPackage('mumps')
KSP.setFromOptions()
E.setFromOptions()
E.solve()
its = E.getIterationNumber()
nconv = E.getConverged()
eigw = [E.getEigenvalue(i).real for i in range(nconv)]
return eigw, flag
def eigenfunctions(self, matrix, number_modes):
import sys, slepc4py
slepc4py.init(sys.argv)
from petsc4py import PETSc
from slepc4py import SLEPc
E = SLEPc.EPS()
E.create()
E.setOperators(matrix.petScMatrix())
E.setProblemType(SLEPc.EPS.ProblemType.HEP)
#E.setType(SLEPc.EPS.Type.ARNOLDI)
E.setFromOptions()
E.setTolerances(tol=1e-9, max_it=200)
E.setDimensions(nev=number_modes)
E.solve()
Print = PETSc.Sys.Print
iterations = E.getIterationNumber()
self.log("Number of iterations of the method: %d" % iterations)
eps_type = E.getType()
self.log("Solution method: %s" % eps_type)
nev, ncv, mpd = E.getDimensions()
self.log("Number of requested eigenvalues: %d" % nev)
tol, maxit = E.getTolerances()
self.log("Stopping condition: tol=%.4g, maxit=%d" % (tol, maxit))
nconv = E.getConverged()
self.log("Number of converged eigenpairs %d" % nconv)
eigenvalues = np.zeros(nconv, dtype=np.complex128)
result_vector = ParallelVector(matrix.distributionPlan())
plan = DistributionPlan(mpi.COMM_WORLD,
n_columns=matrix.totalShape()[1],
n_rows=nconv)
eigenvectors_parallel = ParallelMatrix(plan)
# Create the results vectors
vr, wr = matrix.petScMatrix().getVecs()
vi, wi = matrix.petScMatrix().getVecs()
#
for i in range(nconv):
k = E.getEigenpair(i, vr, vi)
result_vector.setCollective(vr.getArray())
eigenvalues[i] = k
if i in eigenvectors_parallel.localRows():
eigenvectors_parallel.setRow(i, result_vector.fullData())
return eigenvalues, eigenvectors_parallel
def test_laplace_physical_ev(parallel=False):
try:
from slepc4py import SLEPc
except ImportError:
pytest.skip(msg="SLEPc unavailable, skipping eigenvalue test")
mesh = UnitSquareMesh(64, 64)
V = FunctionSpace(mesh, 'CG', 1)
u = TrialFunction(V)
v = TestFunction(V)
bc = DirichletBC(V, Constant(0.0), (1, 2, 3, 4))
# We just need the Stiffness and Mass matrix
a = inner(grad(u), grad(v)) * dx
m = inner(u, v) * dx
A = topetsc(assemble(a, bcs=[bc]))
M = topetsc(assemble(m, bcs=[bc]))
# This shifts the "1.0" Eigenvalues out of the spectrum
# of interest (which is around 0.0 in this case).
vals = np.repeat(1E8, len(bc.nodes))
A.setValuesLocalRCV(bc.nodes.reshape(-1, 1), bc.nodes.reshape(-1, 1),
vals.reshape(-1, 1))
A.assemble()
E = SLEPc.EPS()
E.create()
E.setOperators(A, M)
st = E.getST()
st.setType('sinvert')
kspE = st.getKSP()
kspE.setType('fgmres')
E.setDimensions(5, PETSc.DECIDE)
E.solve()
nconv = E.getConverged()
assert (nconv > 0)
# Create the results vectors
vr, wr = A.getVecs()
vi, wi = A.getVecs()
ev = []
for i in range(nconv):
k = E.getEigenpair(i, vr, vi)
ev.append(k.real)
# Exact eigenvalues are
ev_exact = np.array([
1**2 * np.pi**2 + 1**2 * np.pi**2, 2**2 * np.pi**2 + 1**2 * np.pi**2,
1**2 * np.pi**2 + 2**2 * np.pi**2
])
assert np.allclose(ev_exact, np.array(ev)[:3], atol=1e-1)
def __init__(self, A, B):
self.eps = SLEPc.EPS().create(wrapping.get_mpi_comm(A))
if B is not None:
self.A = wrapping.to_petsc4py(A)
self.B = wrapping.to_petsc4py(B)
self.eps.setOperators(self.A, self.B)
else:
self.A = wrapping.to_petsc4py(A)
self.B = None
self.eps.setOperators(self.A)
def getMinAbsEigenvalue(K):
import sys, slepc4py
slepc4py.init(sys.argv)
from petsc4py import PETSc
from slepc4py import SLEPc
N = len(K[0,:])
opts = PETSc.Options()
A = PETSc.Mat(); A.createDense((N,N))
# copy matrices to petsc, is there are more direct way ?
for i in range(N):
for j in range(N):
A[i,j] = K[i,j]
A.assemble()
E = SLEPc.EPS(); E.create()
E.setOperators(A)
# Non-hermite problem
E.setProblemType(SLEPc.EPS.ProblemType.NHEP)
E.setWhichEigenpairs(SLEPc.EPS.Which.SMALLEST_MAGNITUDE)
E.setTolerances(tol=1.e-11, max_it=1000)
E.setDimensions(1)
E.solve()
#nPrint = PETSc.Sys.Print
eps_type = E.getType()
#Print( "Solution method: %s" % eps_type )
nconv = E.getConverged()
Print( "Number of converged eigenpairs %d" % nconv )
val = []
# # Create the results vectors
vr, wr = A.getVecs()
vi, wi = A.getVecs()
# #
for i in range(nconv): val.append(E.getEigenpair(i, vr, vi))
# val.append(k)
# error = E.computeRelativeError(i)
# #E.getEigenvector(0, vr, vi)
# #return vr[:]+ vi[:]
# print "Eigenvalues: ", val
vals = array(val)
idx = argmin(abs(vals))
return vals[idx]
def eigsh_slepc(A, k, tol, max_iter):
### Setup matrix operator
n = A.shape[0]
mat = MatrixOperator(A)
A_operator = PETSc.Mat().createPython([n, n], mat)
A_operator.setUp()
### Solve eigenproblem
E = SLEPc.EPS()
E.create()
E.setOperators(A_operator)
E.setProblemType(SLEPc.EPS.ProblemType.HEP)
E.setDimensions(k)
E.setTolerances(tol, max_iter)
E.setWhichEigenpairs(SLEPc.EPS.Which.SMALLEST_REAL)
def monitor_fun(eps, iters, nconv, eigs, errors):
print("Current iteration: %d, number of converged eigenvalues: %d" % (iters, nconv))
E.setMonitor(monitor_fun)
E.solve()
print("Number of calls to Ax: %d" % mat.n_calls)
### Collect results
print("")
its = E.getIterationNumber()
print("Number of iterations of the method: %i" % its)
sol_type = E.getType()
print("Solution method: %s" % sol_type)
nev, ncv, mpd = E.getDimensions()
print("NEV %d NCV %d MPD %d" % (nev, ncv, mpd))
tol, maxit = E.getTolerances()
print("Stopping condition: tol=%.4g, maxit=%d" % (tol, maxit))
nconv = E.getConverged()
print("Number of converged eigenpairs: %d" % nconv)
nconv = min(nconv, k)
if nconv < k:
raise ZeroDivisionError("Failed to converge for requested number of k with maxiter=%d" % max_iter)
vecs = np.zeros([n, nconv])
vals = np.zeros(nconv)
xr, tmp = A_operator.getVecs()
xi, tmp = A_operator.getVecs()
if nconv > 0:
for i in range(nconv):
k = E.getEigenpair(i, xr, xi)
vals[i] = k.real
vecs[:, i] = xr
return vals, vecs
def __init__(self, Mat, n_eigvals=1, n_initial=1):
self.eps = SLEPc.EPS()
self.Mat = Mat
self.eps.create(self.Mat.getComm())
self.eps.setOperators(self.Mat)
self.n_eigvals = n_eigvals
self.eps.setDimensions(nev=self.n_eigvals, ncv=PETSc.DECIDE, mpd=PETSc.DECIDE)
self.eps.setProblemType(SLEPc.EPS.ProblemType.HEP)
self.eps.setType(SLEPc.EPS.Type.JD)
self.eps.setWhichEigenpairs(SLEPc.EPS.Which.SMALLEST_REAL)
self.eps.max_it = 99
def __init__(self, PType, SType):
self.E = spc.EPS().create()
if PType == "GHEP":
self.E.setProblemType(spc.EPS.ProblemType.GHEP)
elif PType == "HEP":
self.E.setProblemType(spc.EPS.ProblemType.HEP)
elif PType == "NHEP":
self.E.setProblemType(spc.EPS.ProblemType.NHEP)
elif PType == "GNHEP":
self.E.setProblemType(spc.EPS.ProblemType.GNHEP)
self.E.setType(SType)
def solve_eigensystem(A, B, problem_type=SLEPc.EPS.ProblemType.GHEP):
# Create the results vectors
xr, tmp = A.getVecs()
xi, tmp = A.getVecs()
pc = PETSc.PC().create()
# pc.setType(pc.Type.HYPRE)
pc.setType(pc.Type.BJACOBI)
ksp = PETSc.KSP().create()
ksp.setType(ksp.Type.PREONLY)
ksp.setPC(pc)
F = SLEPc.ST().create()
F.setType(F.Type.PRECOND)
F.setKSP(ksp)
F.setShift(0)
# Setup the eigensolver
E = SLEPc.EPS().create()
E.setST(F)
E.setOperators(A, B)
E.setType(E.Type.LOBPCG)
E.setDimensions(10, PETSc.DECIDE)
E.setWhichEigenpairs(E.Which.SMALLEST_REAL)
E.setProblemType(problem_type)
E.setFromOptions()
# Solve the eigensystem
E.solve()
Print("")
its = E.getIterationNumber()
Print("Number of iterations of the method: %i" % its)
sol_type = E.getType()
Print("Solution method: %s" % sol_type)
nev, ncv, mpd = E.getDimensions()
Print("Number of requested eigenvalues: %i" % nev)
tol, maxit = E.getTolerances()
Print("Stopping condition: tol=%.4g, maxit=%d" % (tol, maxit))
nconv = E.getConverged()
Print("Number of converged eigenpairs: %d" % nconv)
if nconv > 0:
Print("")
Print(" k ||Ax-kx||/||kx|| ")
Print("----------------- ------------------")
for i in range(nconv):
k = E.getEigenpair(i, xr, xi)
error = E.computeError(i)
if k.imag != 0.0:
Print(" %9f%+9f j %12g" % (k.real, k.imag, error))
else:
Print(" %12f %12g" % (k.real, error))
Print("")
def eigensolver_setup(self):
E = SLEPc.EPS()
E.create()
E.setType(SLEPc.EPS.Type.KRYLOVSCHUR)
E.setProblemType(SLEPc.EPS.ProblemType.GHEP)
#E.setWhichEigenpairs(E.Which.SMALLEST_MAGNITUDE)
E.setWhichEigenpairs(E.Which.TARGET_MAGNITUDE)
E.setTarget(0)
st = E.getST()
st.setType('sinvert')
return E
def eigensolver_setup(self):
E = SLEPc.EPS()
E.create()
E.setType(SLEPc.EPS.Type.KRYLOVSCHUR)
E.setProblemType(SLEPc.EPS.ProblemType.HEP)
E.setWhichEigenpairs(E.Which.TARGET_REAL)
E.setTarget(-.1)
st = E.getST()
st.setType('sinvert')
st.setShift(-1.e-3)
return E
def OptimalForcingsSLEPc(FR, shell, nev):
comm = MPI.COMM_WORLD
rank = comm.Get_rank()
f = FR.getVecRight()
q = shell.L.getVecRight()
tmp = shell.L.getVecRight()
ndof_f = f.getSize()
ndof_q = q.getSize()
# Setup EPS
Print(' - Setting the EPS')
S = SLEPc.EPS()
S.create(comm)
S.setOperators(FR)
S.setDimensions(nev=nev, ncv=10)
S.setTolerances(tol=1e-6, max_it=100)
S.setFromOptions()
# Solve the EVP
Print(" - Solving the EPS")
S.solve()
its = S.getIterationNumber()
nconv = S.getConverged()
if rank == 0:
eigvals = zeros(nconv, 'complex')
fs = zeros([ndof_f, nconv], 'complex')
qs = zeros([ndof_q, nconv], 'complex')
else:
eigvals = None
fs = None
qs = None
for i in range(nconv):
eigval = S.getEigenpair(i, f)
shell.Pu.mult(f, q)
shell.B2.mult(q, tmp)
shell.ksp.solve(tmp, q)
q.scale(-1.0)
vf = Vec2DOF(f)
vq = Vec2DOF(q)
if rank == 0:
eigvals[i] = eigval
fs[:, i] = vf
qs[:, i] = vq
return eigvals, fs, qs
def eigensolver_setup(self):
E = SLEPc.EPS()
E.create()
E.setType(SLEPc.EPS.Type.KRYLOVSCHUR)
#E.setType(SLEPc.EPS.Type.LANCZOS)
E.setProblemType(SLEPc.EPS.ProblemType.GHIEP)
#E.setWhichEigenpairs(E.Which.LARGEST_MAGNITUDE)
E.setWhichEigenpairs(E.Which.TARGET_MAGNITUDE)
E.setTarget(0.)
st = E.getST()
st.setType('sinvert')
st.setShift(1.e-3)
return E
def solve_fancy(P, beta_trial, E_trial=None, neigs=1):
"""
Solves eigenproblem with fancier tools
"""
from petsc4py import PETSc
from slepc4py import SLEPc
t = time.time()
# convert eigenproblem into a form that petsc can understand
fancy_P = PETSc.Mat().createAIJ(size=P.shape,
csr=(P.indptr, P.indices, P.data))
# initalise solver object
E = SLEPc.EPS()
E.create()
# lets set the solver options
E.setOperators(fancy_P)
E.setProblemType(SLEPc.EPS.ProblemType.NHEP)
E.setDimensions(nev=neigs)
if E_trial != None:
E.setInitialSpace(E_trial)
# now we set up the spectral region to look in
E.setTarget(beta_trial**2)
E.setWhichEigenpairs(7) #look for closest in absoult value
#beta_trial**2)
print('Solving eigenmodes using fancy solver')
E.solve()
#now we can start unpacking
nconv = E.getConverged()
beta = []
Ex = []
Ey = []
vr, wr = fancy_P.getVecs()
vi, wi = fancy_P.getVecs()
for i in range(nconv):
beta.append(E.getEigenpair(i, vr, vi)**0.5)
#beta.append(E.getEigenvalue(i)**0.5)
Exr, Eyr = np.split(np.array(vr), 2)
Exi, Eyi = np.split(np.array(vi), 2)
Ex.append(Exr + 1j * Exi)
Ey.append(Eyr + 1j * Eyi)
E.destroy()
fancy_P.destroy()
return beta, Ex, Ey
def projectCovToMesh(self, num_kl, cov_expr):
"""TODO: Docstring for projectCovToMesh. Solves CX = BX where C is the covariance matrix
:num_kl : number of kl exapansion terms needed
:returns: TODO
"""
# turn the flag to true
self.flag = True
# get C,B matrices
C = self.getCovMat(cov_expr)
B = self._getBMat()
# Solve the generalized eigenvalue problem
eigensolver = SLEPc.EPS()
eigensolver.create()
eigensolver.setOperators(C, B)
eigensolver.setDimensions(num_kl)
eigensolver.setProblemType(SLEPc.EPS.ProblemType.GHEP)
eigensolver.setFromOptions()
eigensolver.solve()
# Get the number of eigen values that converged.
#nconv = eigensolver.get_number_converged()
# Get N eigenpairs where N is the number of KL expansion and check if N < nconv otherwise you had
# really bad matrix
# create numpy array of vectors and eigenvalues
self.eigen_funcs = np.empty((num_kl), dtype=object)
self.eigen_vals = np.empty((num_kl), dtype=float)
# store the eigenvalues and eigen functions
V = FunctionSpace(self._mesh, "CG", 1)
x_real = PETSc.Vec().create()
x_real.setSizes(self.domain.getNodes())
x_real.setUp()
x_real.setValues(np.arange(0, self.domain.getNodes()),
np.zeros(self.domain.getNodes()))
# for i in range(0, self.domain.getNodes()):
# x_real.setValue(i, 0, 0.)
for eigen_pairs in range(0, num_kl):
lam = eigensolver.getEigenpair(eigen_pairs, x_real)
self.eigen_funcs[eigen_pairs] = Function(V)
# use dof_to_vertex map to map values to the function space
self.eigen_funcs[eigen_pairs].vector()[:] = x_real.getValues(
dof_to_vertex_map(V).astype('int32'))
# divide by norm to make the unit norm again
self.eigen_funcs[eigen_pairs].vector()[:] = self.eigen_funcs[eigen_pairs].vector()[:] / \
norm(self.eigen_funcs[eigen_pairs])
self.eigen_vals[eigen_pairs] = lam.real
def Eigen_Value_Solver(Hamiltonian, number_of_eigenvalues, input_par, l, Viewer):
EV_Solver = SLEPc.EPS().create(comm=PETSc.COMM_WORLD)
EV_Solver.setOperators(Hamiltonian)
# EV_Solver.setType(SLEPc.EPS.Type.JD)
EV_Solver.setProblemType(SLEPc.EPS.ProblemType.NHEP)
EV_Solver.setTolerances(input_par["tolerance"], PETSc.DECIDE)
EV_Solver.setWhichEigenpairs(EV_Solver.Which.SMALLEST_REAL)
size_of_matrix = PETSc.Mat.getSize(Hamiltonian)
dimension_size = int(size_of_matrix[0]) * 0.1
EV_Solver.setDimensions(number_of_eigenvalues, PETSc.DECIDE, dimension_size)
EV_Solver.solve()
num_of_cont = 0
if rank == 0:
print("Number of eigenvalues requested and converged")
print(number_of_eigenvalues, EV_Solver.getConverged(), "\n")
converged = EV_Solver.getConverged()
for i in range(converged):
eigen_vector = Hamiltonian.getVecLeft()
eigen_state = EV_Solver.getEigenpair(i, eigen_vector)
if eigen_state.real < 0:
eigen_vector.setName("BS_Psi_" + str(l) + "_" + str(i + l + 1))
Viewer.view(eigen_vector)
energy = PETSc.Vec().createMPI(1, comm=PETSc.COMM_WORLD)
energy.setValue(0,eigen_state)
energy.setName("BS_Energy_" + str(l) + "_" + str(i + l + 1))
energy.assemblyBegin()
energy.assemblyEnd()
Viewer.view(energy)
else:
eigen_vector.setName("CS_Psi_" + str(l) + "_" + str(num_of_cont))
Viewer.view(eigen_vector)
energy = PETSc.Vec().createMPI(1, comm=PETSc.COMM_WORLD)
energy.setValue(0,eigen_state)
energy.setName("CS_Energy_" + str(l) + "_" + str(num_of_cont))
energy.assemblyBegin()
energy.assemblyEnd()
Viewer.view(energy)
num_of_cont += 1
return num_of_cont
def compute_spectrum(mat, filename):
from slepc4py import SLEPc
eps = SLEPc.EPS().create()
eps.setDimensions(10)
eps.setOperators(mat)
# Compute 10 eigenvalues with smallest magnitude
eps.setWhichEigenpairs(eps.Which.SMALLEST_MAGNITUDE)
eps.solve()
smallest = np.array([eps.getEigenvalue(i) for i in range(eps.getConverged())])
np.save(filename + '_smallest.npy', smallest)
# Compute 10 eigenvalues with largest magnitude
eps.setWhichEigenpairs(eps.Which.LARGEST_MAGNITUDE)
eps.solve()
largest = np.array([eps.getEigenvalue(i) for i in range(eps.getConverged())])
np.save(filename + '_largest.npy', largest)
def find_eigval(H):
print("Importing Stuff")
import sys, slepc4py
from petsc4py import PETSc
from slepc4py import SLEPc
import numpy
A = PETSc.Mat().create()
print(H.shape)
A.setSizes(H.shape)
A.setFromOptions()
A.setUp()
A[:, :] = np.array(H, dtype=np.complex64)[:, :]
A.assemble()
E = SLEPc.EPS()
E.create()
E.setOperators(A)
E.setProblemType(SLEPc.EPS.ProblemType.HEP)
E.setDimensions(nev=H.shape[0])
E.setTolerances(1e-8, 1000)
E.setFromOptions()
E.solve()
Print = PETSc.Sys.Print
Print()
Print("******************************")
Print("*** SLEPc Solution Results ***")
Print("******************************")
Print()
its = E.getIterationNumber()
Print("Number of iterations of the method: %d" % its)
eps_type = E.getType()
Print("Solution method: %s" % eps_type)
nev, ncv, mpd = E.getDimensions()
Print("Number of requested eigenvalues: %d" % nev)
Print(
" the maximum dimension of the subspace to be used by the solver: %d "
% ncv)
Print(" the maximum dimension allowed for the projected problem: %d" % mpd)
#tol, maxit = E.getTolerances()
#Print("Stopping condition: tol=%.4g, maxit=%d" % (tol, maxit))
nconv = E.getConverged()
ans = []
for i in range(nconv):
ans.append(E.getEigenvalue(i))
return ans
def configure_solver(self):
eps = SLEPc.EPS().create()
st = eps.getST()
st.setType("sinvert")
# Set up the linear solver
ksp = st.getKSP()
ksp.setType("preonly")
pc = ksp.getPC()
pc.setType("lu")
pc.setFactorSolverType("mumps")
# Set up number of (default) modes to converge
eps.setDimensions(self.nof_modes_to_converge, SLEPc.DECIDE)
return eps
def eigenvalues(self):
import slepc4py
slepc4py.init()
from slepc4py import SLEPc
E = SLEPc.EPS(); E.create()
E.setOperators(self.A)
E.setProblemType(SLEPc.EPS.ProblemType.NHEP)
E.setDimensions(1)
E.setWhichEigenpairs(SLEPc.EPS.Which.SMALLEST_REAL)
E.setFromOptions()
E.solve()
n = E.getConverged()
for i in range(0, n):
print(E.getEigenvalue(0))
return
def krylovschur(A, n, massmatrix=None, onseperation="continue"):
if massmatrix is not None:
raise NotImplementedError
if onseperation != "continue":
raise NotImplementedError
from petsc4py import PETSc
from slepc4py import SLEPc
M = petsc_matrix(A)
E = SLEPc.EPS().create()
E.setOperators(M)
E.setDimensions(nev=n)
E.setWhichEigenpairs(E.Which.LARGEST_REAL)
E.solve()
X = np.column_stack([x.array for x in E.getInvariantSubspace()])
return X[:, :n]
def DirectModeSLEPc(ffdisc, shift, nev):
from petsc4py import PETSc
from slepc4py import SLEPc
# Operators
print 'Set operators'
Lmat = CSR2Mat(ffdisc.L)
Bmat = CSR2Mat(ffdisc.B)
# Setup EPS
print 'Set solver'
S = SLEPc.EPS()
S.create()
S.setTarget(shift)
S.setWhichEigenpairs(SLEPc.EPS.Which.TARGET_MAGNITUDE)
SI = SLEPc.ST().create()
SI.setType(SLEPc.ST.Type.SINVERT)
SI.setOperators(Lmat, Bmat)
SI.setShift(shift)
S.setST(SI)
S.setDimensions(nev=nev, ncv=60)
S.setTolerances(tol=tol_ev, max_it=100)
S.setFromOptions()
# Solve the EVP
print 'Solving EVP'
S.solve()
its = S.getIterationNumber()
nconv = S.getConverged()
omega = zeros(nconv, 'complex')
modes = zeros([ffdisc.ndof, nconv], 'complex')
ev = Lmat.getVecRight()
for i in range(nconv):
eigval = S.getEigenpair(i, ev)
v = Vec2DOF(ev)
omega[i] = eigval / (-1j)
modes[:, i] = v
return omega, modes
