FAQ: Difference between revisions

From openpipeflow.org
Jump to navigation Jump to search
Line 31: Line 31:
== I get 'matrix inversion error 1' ==
== I get 'matrix inversion error 1' ==
An error in a #define prevented serial use with openpipeflow-1.10, which was a significant update to the double parallelisation!
An error in a #define prevented serial use with openpipeflow-1.10, which was a significant update to the double parallelisation!
Please update to openpipeflow-1.11 or later in this case.
Please [[download]] openpipeflow-1.11 or later in this case.

Revision as of 20:11, 4 August 2015

How do I set up the libraries?

If you haven't already, please check out Getting_started#Compiling_libraries.

Sometimes the curl library is necessary, sometimes not. Try commenting adding/commenting out -lcurl


Where do I get initial conditions for a simulation?

The best initial condition is almost always to restart from a run with as similar parameters as possible. See Database.

If such a state-file is not to hand, try utils/ic.f90. Artificial initial conditions tend to suffer far greater transients. Be prepared to use smaller timesteps and possible greater resolution to overcome the transient.


My job isn't starting properly, what's wrong?

If you started with a command like nohup ./main.out > OUT 2> OUT.err & , then check OUT.err and the head/tail of OUT first.

If you see an error from tim_check_cgce, it's likely that either an inappropriate initial condition was used, or that the resolution was inadequate.

Double-check all parameters were set correctly, see the main.info file. If you had a similar run that worked, check which parameters have changed: try e.g. diff main.info ../job005/main.info

How do I select resolution?

Best to start from a similar job and to check one of the spectrum files, e.g. vel_spec0010.dat. The values should fall by several orders, to approximately below 1e-5 in each dimension.


What is the spatial grid?

i_N radial points are non-equispaced, bunched near the walls, and very slightly so at the axis to ensure spatial accuracy where fewer points are available for the banded matrices. The points are equispaced in theta on 3*i_M points and in z on 3*i_K points. The domain has length L=2.pi/d_alpha.


I get 'matrix inversion error 1'

An error in a #define prevented serial use with openpipeflow-1.10, which was a significant update to the double parallelisation! Please download openpipeflow-1.11 or later in this case.