Closed Alexey-Voronin closed 6 months ago
I've just taken a look at this. I think the key error message I found was:
Traceback (most recent call last):
File "/home/connor/Code/firedrake-dev1/src/tsfc/gem/node.py", line 226, in __call__
return self.cache[cache_key]
This shows us that the crash is happening inside TSFC and so I think our form compiler is struggling with something to do with your form/discretisation. I tried increasing this line by a factor of 100 (overkill) and it yielded a new error:
ModuleNotFoundError: No module named 'phmg'
I think it's unrelated so I suspect that increasing the stack limit might get your code to work.
Can you try increasing sys.setrecursionlimit
and let us know if that stops the crash? It naturally is still an issue that we are doing so much recursion in our code generation, but at least it will verify where the problem is.
Upping the recursion limit fixed the problem. Thanks!
resolved
I'm experiencing crashes during the setup phase of a monolithic h-multigrid solver for the 3D lid-driven cavity problem using Taylor-Hood elements. This occurs when solving with polynomial approximation degrees $k$ in the range $[2,10]$ for $Pk/P{k-1}$ elements, specifically at $k=8$. Below are the steps to reproduce the issue for $k=8$ and the error snapshot.
Changing the mesh size (
N
) does not change the outcome.Steps to Reproduce
python code.py
Expected behavior Expected convergence in around 10-11 iterations
Error message
Environment:
firedrake-status
Status of components:
|Package |Branch |Revision |Modified |
|FInAT |master |e2805c4 |False | |PyOP2 |master |fbde61f9 |False | |fiat |master |e7b2909 |False | |firedrake |master |c5e939dde |False | |h5py |firedrake |4c01efa9 |False | |libspatialindex |master |4768bf3 |True | |libsupermesh |master |dbe226b |False | |loopy |main |8158afdb |False | |petsc |firedrake |09f36907a6|False | |pyadjoint |master |f194553 |False | |pytest-mpi |main |a478bc8 |False | |tsfc |master |799191d |False | |ufl |master |054b0617 |False |