hi all,
i've been working on implementing the Summation by Parts (SBP) operators on some evolution codes i have and i was seeing some instabilities when using the 4th order accurate 2nd derivative operator.
in order to try to understand the problem better, i've coded a very simple thorn that evolves the wave equation in first-order in time, second-order in space form. i can evolve the system just fine without any problems with the SBP operators of order 2, 6 and 8. with order 4, however, there are instabilities propagating from the boundary that crash the code very early on. has anyone seen something like this before? could there be a bug in the 4th order 2nd derivative SBP operators, or am i doing something wrong? i find it very strange that it only crashes with the 4th order accurate 2nd derivative SBP stencils (the code also seems to run fine with the parameter SummationByParts::sbp_2nd_deriv = no)...
i've uploaded this thorn (called IST_WaveToy) here: https://github.com/mzilhao/IST_WaveToy i tried to keep everything as simple as possible. i'm attaching the thornlist used to checkout and compile the code, and in the repository i've also added the parameter files i tried with the different SBP orders (under ./par).
any help or comments would be appreciated.
thanks, Miguel
Hi Miguel,
I have commented on the ticket you created about this. Please take a look at that comment and let me know what you think.
Cheers,
Peter
On Thursday 2020-04-09 13:08, Miguel Zilhão wrote:
Date: Thu, 9 Apr 2020 13:08:47 From: Miguel Zilhão miguel.zilhao.nogueira@tecnico.ulisboa.pt To: Einstein Toolkit Users users@einsteintoolkit.org Subject: [Users] Instabilities with Summation by Parts 4th order operator
hi all,
i've been working on implementing the Summation by Parts (SBP) operators on some evolution codes i have and i was seeing some instabilities when using the 4th order accurate 2nd derivative operator.
in order to try to understand the problem better, i've coded a very simple thorn that evolves the wave equation in first-order in time, second-order in space form. i can evolve the system just fine without any problems with the SBP operators of order 2, 6 and 8. with order 4, however, there are instabilities propagating from the boundary that crash the code very early on. has anyone seen something like this before? could there be a bug in the 4th order 2nd derivative SBP operators, or am i doing something wrong? i find it very strange that it only crashes with the 4th order accurate 2nd derivative SBP stencils (the code also seems to run fine with the parameter SummationByParts::sbp_2nd_deriv = no)...
i've uploaded this thorn (called IST_WaveToy) here: https://github.com/mzilhao/IST_WaveToy i tried to keep everything as simple as possible. i'm attaching the thornlist used to checkout and compile the code, and in the repository i've also added the parameter files i tried with the different SBP orders (under ./par).
any help or comments would be appreciated.
thanks, Miguel
hi Peter,
thanks for having a look. i've just replied to the ticket. indeed, setting SummationByParts::operator_type = "Minimal Bandwidth" does seem to lead to stable evolutions.
thanks, Miguel
On 07/05/20 14:59, Peter Diener wrote:
Hi Miguel,
I have commented on the ticket you created about this. Please take a look at that comment and let me know what you think.
Cheers,
Peter
On Thursday 2020-04-09 13:08, Miguel Zilhão wrote:
Date: Thu, 9 Apr 2020 13:08:47 From: Miguel Zilhão miguel.zilhao.nogueira@tecnico.ulisboa.pt To: Einstein Toolkit Users users@einsteintoolkit.org Subject: [Users] Instabilities with Summation by Parts 4th order operator
hi all,
i've been working on implementing the Summation by Parts (SBP) operators on some evolution codes i have and i was seeing some instabilities when using the 4th order accurate 2nd derivative operator.
in order to try to understand the problem better, i've coded a very simple thorn that evolves the wave equation in first-order in time, second-order in space form. i can evolve the system just fine without any problems with the SBP operators of order 2, 6 and 8. with order 4, however, there are instabilities propagating from the boundary that crash the code very early on. has anyone seen something like this before? could there be a bug in the 4th order 2nd derivative SBP operators, or am i doing something wrong? i find it very strange that it only crashes with the 4th order accurate 2nd derivative SBP stencils (the code also seems to run fine with the parameter SummationByParts::sbp_2nd_deriv = no)...
i've uploaded this thorn (called IST_WaveToy) here: https://github.com/mzilhao/IST_WaveToy i tried to keep everything as simple as possible. i'm attaching the thornlist used to checkout and compile the code, and in the repository i've also added the parameter files i tried with the different SBP orders (under ./par).
any help or comments would be appreciated.
thanks, Miguel
users@lists.einsteintoolkit.org
-
Miguel Zilhão -
Peter Diener