cwpearson/astaroth
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Forcing hotfix.

Browse Source 
Will need more investigation before scientific runs. Now just something to correct the obvious bug.
This commit is contained in:
Miikka Vaisala
2019-08-07 16:04:48 +08:00
parent cf6b75f82a
commit 7e6361a92a
1 changed files with 6 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
acc/mhd_solver/stencil_process.sps
View File

@@ -222,13 +222,16 @@ helical_forcing(Scalar magnitude, Vector k_force, Vector xx, Vector ff_re, Vecto
// JP: This looks wrong:
// 1) Should it be dsx * nx instead of dsx * ny?
// 2) Should you also use globalGrid.n instead of the local n?
// MV: You are rigth. Made a quickfix. I did not see the error because multigpu is split
// in z direction not y direction.
// 3) Also final point: can we do this with vectors/quaternions instead?
// Tringonometric functions are much more expensive and inaccurate/
// MV: Good idea. No an immediate priority.
// Fun related article:
// https://randomascii.wordpress.com/2014/10/09/intel-underestimates-error-bounds-by-1-3-quintillion/
xx.x = xx.x*(2.0*M_PI/(dsx*(DCONST_INT(AC_ny_max) - DCONST_INT(AC_ny_min))));
xx.y = xx.y*(2.0*M_PI/(dsy*(DCONST_INT(AC_ny_max) - DCONST_INT(AC_ny_min))));
xx.z = xx.z*(2.0*M_PI/(dsz*(DCONST_INT(AC_ny_max) - DCONST_INT(AC_ny_min))));
xx.x = xx.x*(2.0*M_PI/(dsx*globalGrid.n.x));
xx.y = xx.y*(2.0*M_PI/(dsy*globalGrid.n.y));
xx.z = xx.z*(2.0*M_PI/(dsz*globalGrid.n.z));
Scalar cos_phi = cos(phi);
Scalar sin_phi = sin(phi);