GitLab

auto detect if we are running on a shared filesystem and remove the cmdline option



See merge request !654

fix PETSc build on operating systems where python defaults to python3



See merge request !646

Variable P with periodic boundary conditions

This MR extends the variable P functionality to simulations with periodic boundary conditions.

This was a lot easier than I expected, but it seems to be working fine.

See merge request !658

Fix FC verbose option when using --procid

This MR just replaces a bunch of `GetRank() == 0` by `TreatAsRankZero()` to fix the verbose option of FieldConvert when using `--procid`.

See merge request !648

Feature/interpolator

This MR introduces an Interpolator class which adds several features/improvements over how interpolation was handled by the PtsField class:

* The brute force search algorithm was swapped for boosts rtree algorithm. This change decreases the interpolation time from hours to seconds
* Besides the modified shepard and linear/quadratic interpolation, two new algorithms, nearest neighbor and Gaussian smoothing were added
* The new class can interpolate between ptFields, ptsFields and expansions and expansions. The latter is only useful when dealing with different meshes.
* Several bugfixes

FieldConvert and EquationSystem were ported to this new class to benefit from the speedup.

This MR also includes the changes what were originally in !582

The required boost version was increased to 1.56 as this is the first version that includes geometry::rtree which works on windows

See merge request !584

Fix boundary conditions after DG optimisation

This MR fixes issues from the DG optimisation introduced in MR !555 inside the compressible flow solver, whereby MPI deadlocks under the use of most boundary conditions due to the use of `ExtractTracePhys` inside the boundary condition evaluation. This is fixed by pulling these calls out to the parent function and passing them through to the boundary condition functions, which is in fact more efficient and avoids the use of these functions repeatedly for each RHS evaluation.

See merge request !657

Fixes for Euler 1D solvers



See merge request !565

Variable P in 3D

This is an attempt to extend the variable order functionality to 3D, partially addressing issue #6.

I tested this for all element types using the channel tests of the incompressible solver (each element having a fixed order in all 3 directions). All tests worked, even in parallel. It would be nice if someone who actually uses the 3D solver could test this with a larger case.

Once we have some confidence in this, I will probably extend DriverAdaptive to 3D, since that should be straightforward. 

See merge request !604

We dont use the partition info here because pts files should be
independant of any meshes/expansions. Instead, we use the file matching
the rank and in case n_ranks!=n_files, we load all pts files for all ranks.

Fix filters output with adaptive driver

This is a small change I forgot to make when introducing DriverAdaptive. Basically, in this case the filters writing text files need to append the file, since otherwise only the last run in the adaptive procedure would be obtained.

I changed the behaviour only when using the adaptive driver, to avoid causing confusion in other cases. However, this could also be used to facilitate restarting from the same directory (although in this case it would be desirable to edit the existing file to keep only times before the restart).

See merge request !588

Create HomogeneousStretch module for FieldConvert

This MR introduces a new module to FieldConvert, which extends the span of a 3DH1D field (only for integer factors). For example, we might want to double the span and obtain the correct duplicated periodic solution. 

This is achieved by moving coefficients between planes so that they are at the correct wave number in the extended domain.

See merge request !609

Fix C0Projection

This MR is a small fix for the C0Projection. I had previously forced FieldConvert to define continuous expansions using a valid variable, since wss and boundary extract require a boundary expansion. However, this conflicts with C0Projection, which requires an expansion without boundary conditions. I created a new flag to make things work in both modules. 

See merge request !644