Compound spaces for pde files

More
7 years 4 months ago #59 by ddrake
Compound spaces for pde files was created by ddrake
Hi,
Using version NETGEN-6.2-dev with NGSolve-6.2.1706-163-g0ed4994
pde files which use compound spaces like this:
Code:
fespace fs -type=compound -spaces=[fs1,fs2,fs3]
are now failing in
Code:
numproc bvp
...add grid_function euq
.The error message is "Exception in NGS_LoadPDE. don't know my dimension"
Here is an example pde file that is failing this way.
Code:
shared = "../libDPG" constant heapsize = 10000000 geometry = square.in2d mesh = square2.vol.gz constant one = 1 constant minus = -1.0 coefficient lam (1+I) # Source (exact solution unknown) coefficient f ( (1+I)*exp( -100.0*(x*x+y*y) ) ) # Compound finite element space: fespace fs1 -type=h1ho -order=3 -dirichlet=[1] -complex fespace fs2 -type=hdivho -order=2 -orderinner=1 -complex fespace fs3 -type=l2ho -order=4 -complex fespace fs -type=compound -spaces=[fs1,fs2,fs3] -complex # Forms: bilinearform dpg -fespace=fs -eliminate_internal gradgrad (1) (3) lam # (grad u, grad v) + Hermitian transpose flxtrc (2) (3) minus # - << q.n, v >> + Hermitian transpose laplace one -comp=3 # (grad e, grad v) mass one -comp=3 # (e,v) linearform lf -fespace=fs source f -comp=3 # Solve: gridfunction uqe -fespace=fs numproc bvp n2 -bilinearform=dpg -linearform=lf -gridfunction=uqe -solver=direct
I'm attaching a pdf file with a gdb stack trace and some source code context.

I will be porting these pde files to python, but please let me know if there is some modification to the pde syntax that may get them working again.

Thanks!
Dow Drake
Attachments:
More
7 years 4 months ago #61 by joachim
Replied by joachim on topic Compound spaces for pde files
fixed with the coming nightly release
(commit 08d09652024c5ebae10e717c731e7d93dbb3c3d0)
The following user(s) said Thank You: ddrake
Time to create page: 0.097 seconds