Dear all,
I am currently trying to set up an easy GRMHD simulation using axial symmetry. In the documentation I find the thorn Cartoon2D recommended for general axisymmetric problems. Is the use of Cartoon2D possible with evolutions in GRHydro? Doing my setup I do find a lot of NaNs in the Prim2Con routines.
Thank you and my best from Valencia
Jens
Jens
Yes, Cartoon2D should be possible with GRHydro. Apart from certain technical difficulties (choosing correct coordinates, number of ghost points, etc.), the largest problem I expect is that Cartoon2D will interpolate the solution, and interpolation is not a conservative operation. Thus I'd expect problems near shocks, discontinuities, and surfaces -- problems in the sense of "the hydro scheme is inaccurate", not in the sense of "segfault" or "nan".
Most likely -- and I have no hard information on which to base my opinion -- what is necessary is a very careful debugging of your parameter file. It's difficult to explain this in an email (and I might also be wrong) -- feel free to call in next Monday for more in-person advice.
-erik
On Thu, Dec 1, 2016 at 5:51 AM, Jens Mahlmann jensmahl@alumni.uv.es wrote:
Dear all,
I am currently trying to set up an easy GRMHD simulation using axial symmetry. In the documentation I find the thorn Cartoon2D recommended for general axisymmetric problems. Is the use of Cartoon2D possible with evolutions in GRHydro? Doing my setup I do find a lot of NaNs in the Prim2Con routines.
Thank you and my best from Valencia
Jens
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Hello Jens, Erik,
if possible, it would also be good if you could attach the content of the Cactus log file (*.out when using simfactory, the file that has the Cactus logo in it) so that one can check that all routines are called in their correct order (since uninitialized variables can show up as NaNs).
Yours, Roland
Jens
Yes, Cartoon2D should be possible with GRHydro. Apart from certain technical difficulties (choosing correct coordinates, number of ghost points, etc.), the largest problem I expect is that Cartoon2D will interpolate the solution, and interpolation is not a conservative operation. Thus I'd expect problems near shocks, discontinuities, and surfaces -- problems in the sense of "the hydro scheme is inaccurate", not in the sense of "segfault" or "nan".
Most likely -- and I have no hard information on which to base my opinion -- what is necessary is a very careful debugging of your parameter file. It's difficult to explain this in an email (and I might also be wrong) -- feel free to call in next Monday for more in-person advice.
-erik
On Thu, Dec 1, 2016 at 5:51 AM, Jens Mahlmann jensmahl@alumni.uv.es wrote:
Dear all,
I am currently trying to set up an easy GRMHD simulation using axial symmetry. In the documentation I find the thorn Cartoon2D recommended for general axisymmetric problems. Is the use of Cartoon2D possible with evolutions in GRHydro? Doing my setup I do find a lot of NaNs in the Prim2Con routines.
Thank you and my best from Valencia
Jens
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Dear Roland, dear Erik,
thank you for your reply. I have attached the Cactus log file. I have found that some problems arise from the table value sdetg(i,j,k) given to the Prim2ConM subroutine in GRHydro (This problem with GRHydro_SqrtSpatialDeterminant I have encountered more than once lately). Using a point-wise derivation of the determinant at runtime I can avoid this. The remaining (few) NaNs I relate so far (as Erik said) to a bad choice of my initial data which has excision and therefore jumps and, thus, may yield problems with the interpolation. The attached file is from after the fixing of the determinant issue - however, the scheduling was not changed.
Thank you and my best
Jens
On 12/01/2016 03:21 PM, Roland Haas wrote:
Hello Jens, Erik,
if possible, it would also be good if you could attach the content of the Cactus log file (*.out when using simfactory, the file that has the Cactus logo in it) so that one can check that all routines are called in their correct order (since uninitialized variables can show up as NaNs).
Yours, Roland
Jens
Yes, Cartoon2D should be possible with GRHydro. Apart from certain technical difficulties (choosing correct coordinates, number of ghost points, etc.), the largest problem I expect is that Cartoon2D will interpolate the solution, and interpolation is not a conservative operation. Thus I'd expect problems near shocks, discontinuities, and surfaces -- problems in the sense of "the hydro scheme is inaccurate", not in the sense of "segfault" or "nan".
Most likely -- and I have no hard information on which to base my opinion -- what is necessary is a very careful debugging of your parameter file. It's difficult to explain this in an email (and I might also be wrong) -- feel free to call in next Monday for more in-person advice.
-erik
On Thu, Dec 1, 2016 at 5:51 AM, Jens Mahlmann jensmahl@alumni.uv.es wrote:
Dear all,
I am currently trying to set up an easy GRMHD simulation using axial symmetry. In the documentation I find the thorn Cartoon2D recommended for general axisymmetric problems. Is the use of Cartoon2D possible with evolutions in GRHydro? Doing my setup I do find a lot of NaNs in the Prim2Con routines.
Thank you and my best from Valencia
Jens
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
users@lists.einsteintoolkit.org
-
Erik Schnetter -
Jens Mahlmann -
Roland Haas