PETSc matrix convertion

Question

Hi,

I'm having trouble converting a FEniCS matrix into a PETSc matrix. For the smaller dimensional I have no problems but for the larger dimensional problems I get the following errors

As = AA.sparray()
A = PETSc.Mat().createAIJ(size=As.shape, csr=(As.indptr, As.indices, As.data))

gives the error

     24     else:
     25         if AA.size(1) == AA.size(0):
---> 26             As = AA.sparray()
     27             A = PETSc.Mat().createAIJ(size=As.shape, csr=(As.indptr, As.indices, As.data))
     28             #row, col, value = AA.data()

/ubc/cs/research/scl/people/mwathen/.hastdist/bld/profile/3kneeoad7jpp/lib/python2.7/site-packages/dolfin/cpp/la.py in sparray(self)
    649         from scipy.sparse import csr_matrix
    650         data = self.data(deepcopy=True)
--> 651         C = csr_matrix((data[2], data[1], data[0]))
    652         return C
    653

/ubc/cs/research/scl/people/mwathen/install/lib/python2.7/site-packages/scipy/sparse/compressed.pyc in __init__(self, arg1, shape, dtype, copy)
     86             self.data = np.asarray(self.data, dtype=dtype)
     87
---> 88         self.check_format(full_check=False)
     89
     90     def getnnz(self, axis=None):

/ubc/cs/research/scl/people/mwathen/install/lib/python2.7/site-packages/scipy/sparse/compressed.pyc in check_format(self, full_check)
    167             raise ValueError("indices and data should have the same size")
    168         if (self.indptr[-1] > len(self.indices)):
--> 169             raise ValueError("Last value of index pointer should be less than "
    170                                 "the size of index and data arrays")
    171

ValueError: Last value of index pointer should be less than the size of index and data arrays

.

row, col, value = AA.data()
A = PETSc.Mat().createAIJ(size=(AA.size(0),AA.size(1)),csr=(row.astype('int32'), col.astype('int32'), value))

gives the error

Segmentation fault

.

A = as_backend_type(AA).mat()

[0]PETSC ERROR: ------------------------------------------------------------------------
[0]PETSC ERROR: Caught signal number 11 SEGV: Segmentation Violation, probably memory access out of range
[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger
[0]PETSC ERROR: or see http://www.mcs.anl.gov/petsc/documentation/faq.html#valgrind[0]PETSC ERROR: or try http://valgrind.org on GNU/linux and Apple Mac OS X to find memory corruption errors
[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run
[0]PETSC ERROR: to get more information on the crash.
application called MPI_Abort(MPI_COMM_WORLD, 59) - process 0

Is there any other way to convert the matrix into a PETSc format? The machine I am running this on has 256G of RAM so memory isn't an issue.

Comments

Is there any particular reason to not start with a PETSc matrix right away?

---

Do you mean using

A = PETScMatrix()

This function is not available in FEniCS 1.5....

---

Hi, maybe you could use the mat() method of PETScMatrix. Works fine for 1.5.0

from dolfin import *

mesh = UnitSquareMesh(10, 10)
V = FunctionSpace(mesh, 'CG', 1)
u = TrialFunction(V)
v = TestFunction(V)

M = PETScMatrix()
assemble(inner(u, v)*dx, tensor=M)

Mmat = M.mat()
print type(Mmat)

Mcopy = Mmat.duplicate()
Mcopy.zeroEntries()

print Mcopy.getRow(3)
print Mmat.getRow(3)
print M.getrow(3)