def calc_spec_weight_pardiso(wmin, wmax):
wvals = []
gf0 = []
gf1 = []
gf2 = []
rhs0 = rspace.create_rhs_vec(0)
rhs1 = rspace.create_rhs_vec(1)
rhs2 = rspace.create_rhs_vec(2)
rhs = np.asfortranarray(np.vstack((rhs0, rhs1, rhs2)).T)
w = wmin
dw = prm.eta / 2.0
static_eom = rspace.create_static_eom_matrix(prm)
# static_eom.sort_indices()
ia = static_eom.indptr + 1 # fortran indexing starts at 1
ja = static_eom.indices + 1 # fortran indexing starts at 1
a = static_eom.data
pt = np.empty(64, dtype='i8', order='F')
iparm = np.empty(64, dtype='i4', order='F')
dparm = np.empty(64, dtype='f8', order='F')
iparm[2] = 1 # number of processors
error = pardiso.init(pt, iparm, dparm)
if error != 0:
print 'Error in pardiso.init(), ERROR={:d}'.format(error)
x, error = pardiso.sym_fact(pt, a, ia, ja, iparm, rhs, dparm)
if error != 0:
print 'Error in pardiso.sum_fact(), ERROR={:d}'.format(error)
while w < wmax:
eom = rspace.create_eom_matrix(static_eom, w)
a = eom.data
x, error = pardiso.solve(pt, a, ia, ja, iparm, rhs, dparm)
if error != 0:
print 'Error in pardiso.solve(), ERROR={:d}'.format(error)
gf0.append(x[rspace.get_index(0, 0, 0), 0])
gf1.append(x[rspace.get_index(0, 0, 1), 1])
gf2.append(x[rspace.get_index(0, 0, 2), 2])
wvals.append(w)
w += dw
def calc_spec_weight_bicg(wmin, wmax):
wvals = []
gf0 = []
gf1 = []
gf2 = []
rhs0 = rspace.create_rhs_vec(0)
rhs1 = rspace.create_rhs_vec(1)
rhs2 = rspace.create_rhs_vec(2)
w = wmin
dw = prm.eta / 2.0
static_eom = rspace.create_static_eom_matrix(prm)
sps.linalg.dsolve.use_solver( useUmfpack = True,\
assumeSortedIndices = True )
while w < wmax:
eom = rspace.create_eom_matrix(static_eom, w)
x0, info = sps.linalg.bicg(eom, rhs0)
x1, info = sps.linalg.bicg(eom, rhs1)
x2, info = sps.linalg.bicg(eom, rhs2)
gf0.append(x0[rspace.get_index(0, 0, 0)])
gf1.append(x1[rspace.get_index(0, 0, 1)])
gf2.append(x2[rspace.get_index(0, 0, 2)])
wvals.append(w)
w += dw
def calc_spec_weight_spsolve(wmin, wmax):
wvals = []
gf0 = []
gf1 = []
gf2 = []
rhs = rspace.create_rhs_matrix()
w = wmin
dw = prm.eta / 2.0
static_eom = rspace.create_static_eom_matrix(prm)
static_eom = static_eom.tocsc()
static_eom.sort_indices()
sps.linalg.dsolve.use_solver( useUmfpack = True,\
assumeSortedIndices = True )
while w < wmax:
eom = rspace.create_eom_matrix(static_eom, w)
x = sps.linalg.spsolve(eom.tocsc(), rhs).toarray()
gf0.append(x[rspace.get_index(0, 0, 0), 0])
gf1.append(x[rspace.get_index(0, 0, 1), 1])
gf2.append(x[rspace.get_index(0, 0, 2), 2])
wvals.append(w)
w += dw
print(w - wmin) / (wmax - wmin)