[Getdp] Post processing and split results bugs
Daryl Van Vorst
dvpub at telus.net
Wed Mar 8 08:54:56 CET 2006
Hi,
I've been doing some runs with many time steps and have noticed that
occasionally the post processing results at a particular time instant
will be bogus (extremely large values), but are fine just before and
just after the problem time step. This appears to only happen when using
the -bin option.
I've been able to recreate this using a very slightly modified got.pro
(from the microwave example on the wiki). Just change the line:
dt = 0.05/Freq;
to:
dt = 0.005/Freq;
And process it with:
$getdp got -pre MW_e_2Dt -cal -bin
$getdp got -pos e
Then load the results into gmsh.
If you don't use "-bin", then the problem doesn't occur. I find that
time step 65 (shown as 64 in gmsh) is invalid. I've tried this on two
linux PC's and get the same result. This also happens with the -split
option (also only when -bin is used). I should add that this problem
isn't restricted to .pos format files. I'm having this problem in
another simulation which saves post-processing results in a "Format
TimeTable" text file.
Then I figured it might be informative to find out what would happen if
I restarted the simulation using the questionable result file (to narrow
down where the cause of the above problem is) and ran into another
problem. (At least I think this is a problem... maybe I'm just using the
program incorrectly.) If you process the same problem using -split and
then try to re-start the processing at some time step the program
segfaults. The segfault isn't related to using what might be a bad
result file (it happens with any set of initial result files).
$getdp got -msh got.msh -name out/got -pre MW_e_2Dt -cal -bin -split
$getdp got -msh got.msh -name out/got -res out/got-00064.res
out/got-00065.res -restart
...
Info : Newmark Time = 1.5821268e-10 s (TimeStep 66)
Operation : Generate[A]
Solver : Loading parameter file 'out/solver.par'
Info : Setting System {A,b} to zero
Error : Segmentation violation (invalid memory reference)
I'm using getdp version 1.1.2 compiled from source. I just tried the
precompiled version from the getdp website and both problems occur, but
the first problem happens at a different time step (7). It also happens
with an unmodified got.pro at step 9 using the precompiled getdp.
Regards,
Daryl.