I have been using version 0.4.X.X of dolfinx built from source. Only version 0.3.X.X is available via spack. I successfully installed this spack version on a hpc cluster. My only modification to the installation procedure at dolfinx/README.md at main · FEniCS/dolfinx · GitHub is to pre-install gcc@10.3.0 (since only 8.3.0 is available in the compiler list).
git clone https://github.com/spack/spack.git
. ./spack/share/spack/setup-env.sh
spack env create fenicsx-env
spack env activate fenicsx-env
spack add gcc@10.3.0
spack install
spack load gcc@10.3.0
spack location -i gcc@10.3.0 | xargs spack compiler add
spack add py-fenics-dolfinx%gcc@10.3.0 cflags="-O3" fflags="-O3"
spack install
This version throws some errors that I’m not sure if they’re caused by any platform dependent installs or just bugs in the old version of code. For example (version 0.3.0) of dolfinx:
import dolfinx
import numpy as np
import ufl
from dolfinx.fem import (Constant, dirichletbc as DirichletBC, Function, FunctionSpace, locate_dofs_topological)
from dolfinx.fem import LinearProblem
# from dolfinx.fem.petsc import LinearProblem
from dolfinx.io import XDMFFile
from dolfinx.mesh import locate_entities_boundary
from mpi4py import MPI
from petsc4py import PETSc
from petsc4py.PETSc import ScalarType
from ufl import ds, dx, grad, inner
# many other lines of code
x0bc = DirichletBC.DirichletBC(u0, locate_dofs_topological(V, 0, x0facet))
x1bc = DirichletBC.DirichletBC(u1, locate_dofs_topological(V, 0, x1facet))
# Define variational problem
u = ufl.TrialFunction(V)
v = ufl.TestFunction(V)
x = ufl.SpatialCoordinate(mesh)
f = Constant(mesh, ScalarType(0))
g = Constant(mesh, ScalarType(0))
a = inner(grad(u), grad(v)) * dx
L = inner(f, v) * dx + inner(g, v) * ds
problem = LinearProblem(a, L, bcs=[x0bc, x1bc], petsc_options={"ksp_type": "preonly", "pc_type": "lu"})
calling LinearProblem raises the following error:
Traceback (most recent call last):
File "/storage/coda1/p-tf74/0/hpcuser/ssb/transport.py", line 75, in <module>
problem = LinearProblem(a, L, bcs=[x0bc, x1bc], petsc_options={"ksp_type": "preonly", "pc_type": "lu"})
File "/storage/coda1/p-tf74/0/hpcuser/spack/opt/spack/linux-rhel7-cascadelake/gcc-10.3.0/py-fenics-dolfinx-0.3.0-gknuzu6dvtixtdpicjmxc576yy6dgbi5/lib/python3.9/site-packages/dolfinx/fem/problem.py", line 58, in __init__
self._A = fem.create_matrix(self._a)
File "/storage/coda1/p-tf74/0/hpcuser/spack/opt/spack/linux-rhel7-cascadelake/gcc-10.3.0/py-fenics-dolfinx-0.3.0-gknuzu6dvtixtdpicjmxc576yy6dgbi5/lib/python3.9/site-packages/dolfinx/fem/assemble.py", line 60, in create_matrix
return cpp.fem.create_matrix(_create_cpp_form(a))
ValueError: vector::reserve
In version 0.4.2, LinearProblem is imported from dolfinx.fem.petsc.
While building a cpp version of dolfinx code using a CMakeLists.txt, I get the error:
[ 25%] Compile conduction.py using FFCx
ERROR:ffcx:Expecting a UFL form file (.ufl).
make[2]: *** [conduction.c] Error 1
make[1]: *** [CMakeFiles/conduction.dir/all] Error 2
make: *** [all] Error 2
I can run a python version of the code and build/run a cpp translation of the same code on a separate machine where I built out dolfinx from source.
Q: Am I to treat the errors as issues that were addressed in v0.4.2 or is there something silently faulty with my spack installation of py-fenics-dolfinx?