- removed comment
Boundary thorn
i was trying to do a (unigrid) run with "scalar" boundary conditions on some variables but with uneven boundary width, ie width of 2 grid points on the x and y axis and 0 on the z axis. in the documentation of the Boundary thorn it states that one should create a table passing this information in an array called "BOUNDARY_WIDTH":
"The table handle identifies a table which holds extra arguments for the particular boundary condition that is requested. For example, if a negative value is passed for the boundary width, then the boundary condition will look in this table for a 2d-element integer array, which holds the width of each face of the boundary (for a d dimensional grid variable). (The first element of the array holds the width of the ‘-x’ face, the second the ‘+x’ face, the third the ‘-y’ face, etc.)"
so i've accordingly created the following table
call Util_TableCreateFromString(param_table_handle, "BOUNDARY_WIDTH = { 2 2 2 2 0 0 }")
and then registered the variables with
ierr = Boundary_SelectGroupForBC(cctkGH, CCTK_ALL_FACES, -one, &
param_table_handle, "ScalarBase::phi", "scalar")
however, i was getting errors like the following:
Boundary/src/Check.c:130: BndSanityCheckWidths: Assertion `dim < (int)sizeof(dims)' failed.
inspecting that file, this is only triggered if (boundary_widths[i] > 100 || boundary_widths[i] < 0)
which meant that my BOUNDARY_WIDTH array was likely not being parsed correctly. digging a little bit deeper, i've found the following in ScalarBoundary.c:138 (and analogous for the other files under CactusBase/Boundary/src):
/* Determine boundary width on all faces */
/* allocate memory for buffer */
gdim = CCTK_GroupDimI(gi);
if (gdim > max_gdim) {
width_alldirs =
(CCTK_INT *)realloc(width_alldirs, 2 * gdim * sizeof(CCTK_INT));
max_gdim = gdim;
}
/* fill it with values, either from table or the boundary_width
parameter */
if (widths[i] < 0) {
err = Util_TableGetIntArray(tables[i], gdim, width_alldirs,
"BOUNDARY_WIDTH");
it seems to me that this last line should be instead
err = Util_TableGetIntArray(tables[i], 2 * gdim, width_alldirs,
"BOUNDARY_WIDTH");
for consistency with the rest of the file and with the documentation, right? indeed, with this change the errors disappeared. i've attached a simple patch that applies this on this file, but i guess an equivalent change would be needed also for the rest of the Boundary files...
if this patch is correct, could this be ported to the current release? and if it's not correct, is there anything i'm missing, in order to register the boundary conditions?
Keyword: Boundary
Comments (11)
-
-
- removed comment
-
- changed status to open
- removed comment
-
- removed comment
Miguel, Erik: is there any progress on this?
-
- removed comment
Miguel, Erik: since you have already proposed a solution, would you like to propose a patch so that it could potentially be included in the next release?
-
reporter - removed comment
There was already one patch proposed, but it only fixed one file. I've just added another patch, fixing all the files affected.
-
- changed status to open
- removed comment
-
- changed status to open
- removed comment
The patch looks fine to me.
Please apply.
Or let me know if you do not have commit rights and I will apply the patch with you as the author.
-
reporter - removed comment
I do not have commit rights... Could you take care of it?
-
- changed status to resolved
- removed comment
Applied as git hash 08315df "Boundary: correct size of width_alldirs data array" of cactusbase.
Thank you.
I did not mention this, but one could have also created a pull request on bitbucket which anyone of the developers could then merge in (ideally using the fast-forward merge method to avoid having one additional merge commit for every pull requested commit).
-
- edited description
- changed status to closed
- Log in to comment
Yes, this looks like an error in the code.
We will also need to have a look at the other files in the same directory, as they might contain the same error.