search

JuliaGeodynamics / LaMEM.jl

Julia interface to LaMEM (Lithosphere and Mantle Evolution Model)
GNU General Public License v3.0
30 stars 13 forks source link

Error with MUMPS #57

Closed mthielma closed 1 month ago

mthielma commented 4 months ago

I tried to run the falling block example from the README (Section 3), but with the direct solver input file FallingBlock_DirectSolver.dat When I run it with:

run_lamem(ParamFile, 4,"-time_end 1")`

I get (I just quote the relevant stuff):

================================= STEP 1 =================================
--------------------------------------------------------------------------
Current time        : 0.00000000 [ ] 
Tentative time step : 10.00000000 [ ] 
--------------------------------------------------------------------------
  0 SNES Function norm 7.310266752999e+01 
  0 PICARD ||F||/||F0||=1.000000e+00 
  ** Instance Error 1 in DMUMPS_F77           0
application called MPI_Abort(MPI_COMM_WORLD, -99) - process 1
boriskaus commented 4 months ago

Which machine and julia version?

mthielma commented 4 months ago

Mac, Ventura 13.3, julia 1.10.3

mthielma commented 3 months ago

Update: running the tests also results in a fail of the parallel test. I tried with different direct solvers and they all fail. So I'm thinking it may not be LaMEM, but an issue with PETSc and the installed direct solvers?

boriskaus commented 1 month ago

PETSc_jll 3.19.6 is compiled with support for MUMPS and SuperLU_Dist on Mac and linux. So the next compilation of LaMEM (2.1.4) should fix this; it does come with the caveat that you need Julia >1.9

boriskaus commented 1 month ago

this should be solved in LaMEM.jl version 0.4.0 (where MUMPS works again on linux + mac).