Hi, Weixiong Zheng I am wondering where is the DG code written by Praveen C, using the slope limiter?
Thanks. Best, Yongchao Zhang 在 2015年11月4日星期三 UTC+8下午1:44:22，Weixiong Zheng写道： > > Thanks for the reply, > > > > 在 2015年11月3日星期二 UTC-6下午11:15:14，Praveen C写道： >> >> Hi >> >> I am using uniform cartesian mesh. In this case, do I still modify > anything (though I would think I don't need to)? > >> You can use FE_DGP on cartesian meshes only. >> >> That's what I am thinking about. My local face terms are identical in x > direction and y directions, respectively. So I may just need to assemble > once for two directions, respectively. > >> In your case, your assembly can be written as a matrix-vector product. If >> Ax, By are same in every cell, then you need to store just one matrix >> which can be used in every cell. This is assuming all your cells are >> identical. If not, then the matrix will be scaled by factors depending on >> dx, dy of the corresponding cell. >> >> One real quick question: when trying to see how to do slope limiter, I > visited your DG code, which is awesome, though it's a bit hard to follow > for me (newbie on DG and slope limiting). Do you have any document > explaining slope limiting with DGQ(1)? I just don't get the idea of > applying limiter for DGQ just because it has 4 DOFs per cell but limiter > can only directly give 3, equivalently. (1, kx, ky) > > Thanks > >> Best >> praveen >> >> On Wed, Nov 4, 2015 at 8:05 AM, Weixiong Zheng <[email protected]> wrote: >> >>> Thanks, Dr Bangerth! >>> >>> This is enlightening! >>> >>> My current equation is kinda like: du/dt + d(Ax*u)/dx + d(By*u)/dx + cu >>> = f. Ax and By are constant coefficient matrices and u (the solution) is a >>> vector. Every nonzero entry of Ax and By is a distinct constant number. >>> Though I don't know what exactly I can do 1 time assembly and store for >>> use, I can see what you say and will think about it. >>> >>> By the way, I am also thinking about transferring from FE_DGQ(1) to >>> FE_DGP(1). Do I need to do any modifications to my code or just do the >>> simple switch directly? I am asking this because I may need slope limiting, >>> for which DGQ is not that easy but DGP could be real straightforward. Also, >>> it has less dof counts. >>> >>> Thanks! >>> >>> 在 2015年11月3日星期二 UTC-6下午6:16:46，bangerth写道： >>>> >>>> >>>> Weixiong, >>>> >>>> > Last time, I got my serial DG code to run for a multi-component >>>> system. Yet, >>>> > the assembly is not affordable even with release mode (36 component >>>> is using >>>> > 1.5 hrs to assemble, yet a higher order approximation with 55 >>>> components takes >>>> > up to 6 hours to assemble (face assembly is the killer!)). So I am >>>> thinking >>>> > about parallelizing my code. >>>> >>>> In addition to Bruno's answer, I think you might want to rethink your >>>> general >>>> approach. If you have this many components, are all of the blocks in >>>> your >>>> matrix really different? For example, if some blocks of the matrix are >>>> the >>>> same, maybe there is no need to individually assemble them, but it >>>> would be >>>> enough if you assembled one and then copied it, or assemble it on the >>>> side and >>>> whenever you need to multiply with this block do so by hand. There are >>>> often >>>> ways to avoid work, rather than just using brute force to assemble >>>> everything. >>>> >>>> Best >>>> W. >>>> >>>> -- >>>> ------------------------------------------------------------------------ >>>> >>>> Wolfgang Bangerth email: [email protected] >>>> www: >>>> http://www.math.tamu.edu/~bangerth/ >>>> >>>> -- >>> The deal.II project is located at http://www.dealii.org/ >>> For mailing list/forum options, see >>> https://groups.google.com/d/forum/dealii?hl=en >>> --- >>> You received this message because you are subscribed to the Google >>> Groups "deal.II User Group" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to [email protected]. >>> For more options, visit https://groups.google.com/d/optout. >>> >> >> -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
