Lines of Code
[root]/CCA/Components/Wasatch/Operators
| Author | Changes | Lines of Code | Lines per Change |
|---|---|---|---|
| Totals | 57 (100.0%) | 973 (100.0%) | 17.0 |
| tsaad | 30 (52.6%) | 762 (78.3%) | 25.4 |
| jsutherland | 27 (47.4%) | 211 (21.7%) | 7.8 |
Remove references to Stencil2.h, s2detail, etc.
0 lines of code changed in 8 files:
Rework the way we are building the windowed "src" fields to be a bit more robust.
0 lines of code changed in 4 files:
Silence a few compiler warnings by removing a few unused variables.
0 lines of code changed in 2 files:
Runtime ghosts in Wasatch.
As of now, this still results in hard-coded ghost cells per field - no variation is allowed. It is the first step toward variable ghosts per task though.
62 lines of code changed in 3 files:
Implement thread locks on flux limiter and upwind operators so that the use of the advective velocity is thread safe.
Also remove a const qualifier from the FluxLimiter apply_to_field() method since it is going to modify the mutex and reset the advective velocity member variable.
61 lines of code changed in 4 files:
Use consistent device information on advecting velocity in UpwindInterpolant.cc
5 lines of code changed in 1 file:
Fix some const correctness issues in UpwindInterpolant.cc
5 lines of code changed in 1 file:
Fix some const correctness with the FluxLimiter stuff.
73 lines of code changed in 2 files:
Minor cleanup; initialize some variables.
3 lines of code changed in 1 file:
Refactor the FluxLimiter to use SpatialFields exclusively. This will allow us to employ the nebo GPU backend. Also rip-out obsolete staggered FluxLimiter declarations.
234 lines of code changed in 5 files:
use memorytype and device id in flux limiters and upwind.
44 lines of code changed in 3 files:
fix a nasty bug in the flux limiter calculation. Account for the plus staggered face on physical boundaries when calculating fluxes.
6 lines of code changed in 1 file:
remove a cout statement. Also note that my previous commit will break the RT on all turbulence tests because of the turbulent-viscosity extrapolation.
0 lines of code changed in 1 file:
use appropriate field references when creating source/destination fields in the flux-limiters. This will pave the way for implementing nebo-assignments for the flux-limiters and provide a hybrid evaluation backend. NOTE: the new field creation still requires appropriate arguments (i.e. consumerMemType and consumerDeviceID).
19 lines of code changed in 2 files:
Make the extrapolant even more robust by allowing developers to specify the minimum and maximum allowable values for extrapolated quantities. I also should clarify the comment on my previous commit: I meant to say that allowing the extrapolant to skip physical boundaries will help in avoiding negative values for the viscosity which are causing Warches issues. With this commit however, we can have a more robust control over extrapolation, even at physical boundaries.
26 lines of code changed in 2 files:
allow extrapolant to skip physical boundaries. This caused problems with Warches when calculating the turbulent viscosity at the boundary because it was leading to negative values in the extra-cells.
50 lines of code changed in 3 files:
refactor the UpwindInterpolant to use nebo assignment.
11 lines of code changed in 2 files:
small cosmetic change to the FluxLimiter Interpolant.
2 lines of code changed in 1 file:
convert the flux limiter interpolant interpolation to use windows. this is a more elgant implementation and will hopefully bring us one step further to using anisotropic nebo stencils.
160 lines of code changed in 2 files:
constify some stuff.
2 lines of code changed in 1 file:
make patch-boundary extrapolant work on S,X,Y,ZVol fields.
3 lines of code changed in 1 file:
This is a major Wasatch commit. First, the viscosity is removed from the stress tensor calculation so that it becomes a strain tensor. This separation helps us reuse the strain tensor in LES models yielding a noticeable speedup. Second, as the viscosity is moved into the momentum RHS calculation, one must exrapolate its values at all patch boundaries - this lead to the development of an extrapolant operator that takes care of this extrapolation. One could of course perform and MPI communication on the turbulent viscosity but this was twice as costly as the extrapolation. Furthermore, the extrapolant will be used in the dynamic smagorinsky model to fill in ghost-cell values before filtering operations. Finally, a few cosmetic changes here and there to make the code look nice. NOTE: this commit will break every Wasatch test that solves the momentum equations - as expected - due to the change in the algebraic operations order. All differences are of the order of machine precision except for the turbulence tests due to the extrapolation.
207 lines of code changed in 7 files: