def check_grad_mat():
import pyublas
if not pyublas.has_sparse_wrappers():
return
grad_mat = op.grad_matrix()
#print len(discr), grad_mat.nnz, type(grad_mat)
for i in range(10):
u = numpy.random.randn(len(discr))
mat_result = grad_mat * u
op_result = numpy.hstack(op.grad(u))
err = la.norm(mat_result-op_result)*la.norm(op_result)
assert err < 1e-5
def check_grad_mat():
import pyublas
if not pyublas.has_sparse_wrappers():
return
grad_mat = op.grad_matrix()
#print len(discr), grad_mat.nnz, type(grad_mat)
for i in range(10):
u = numpy.random.randn(len(discr))
mat_result = grad_mat * u
op_result = numpy.hstack(op.grad(u))
err = la.norm(mat_result - op_result) * la.norm(op_result)
assert err < 1e-5
def do_test_umfpack(self, dtype):
if not pyublas.has_sparse_wrappers():
return
if not pyublasext.has_umfpack():
return
size = 100
#A = make_random_matrix(size, dtype, numpy.SparseExecuteMatrix)
#b = make_random_vector(size, dtype)
A = tmd.umf_a[dtype]
b = tmd.umf_b[dtype]
umf_op = pyublasext.UMFPACKOperator.make(A)
x = numpy.zeros((size,), dtype)
umf_op.apply(b, x)
self.assert_(la.norm(b - A * x) < 1e-10)
def test_add_scattered(self):
if not pyublas.has_sparse_wrappers():
return
self.for_all_dtypes(self.do_test_add_scattered)
def test_sparse_operators(self):
if not pyublas.has_sparse_wrappers():
return
for flavor in [pyublas.SparseBuildMatrix, pyublas.SparseExecuteMatrix]:
self.do_test_sparse_operators(flavor)
def test_sparse(self):
if not pyublas.has_sparse_wrappers():
return
self.for_all_dtypes(self.do_test_sparse)
def test_arpack_generalized(self):
if not pyublas.has_sparse_wrappers():
return
self.for_all_dtypes(self.do_test_arpack_generalized)
