Re: [deal.II] unstable computation when using parallel computation (PETSC)
Thanks for the rapid response. I check the residual, which turns the same as the previous. But after solving the linear system, the solution contains NAN. So I guess the linear solver doesn't work properly when I change the number of cores. Best, Chenyi On Sunday, January 9, 2022 at 3:14:48 PM UTC+1 Wolfgang Bangerth wrote: > On 1/9/22 6:57 AM, Chenyi LUO wrote: > > > > The codes ran successfully before by either 128, 256 or 512 cores.After > one > > day, I find it doesn't work by 128, 256 cores but still works for 512 > cores. > > > > Did similar situation occur before? Is there a clue for possible reasons? > > A good starting point is always to elaborate on what "it doesn't work" > precisely is. What happens? > > Best > W. > > > -- > > Wolfgang Bangerth email: bang...@colostate.edu > www: http://www.math.colostate.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 dealii+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/988804c2-f927-4c51-b932-e38a234f21b4n%40googlegroups.com.
[deal.II] unstable computation when using parallel computation (PETSC)
Dear Dealii group, I'm now trying to solve a poromechanical problem by using petsc library. Recently, I met a problem, which I cannot figure out the reason. The codes ran successfully before by either 128, 256 or 512 cores.After one day, I find it doesn't work by 128, 256 cores but still works for 512 cores. Did similar situation occur before? Is there a clue for possible reasons? Moreover, another confusion comes from the computational time for, e.g. linear solver. I didn't change the code. But the required time, for example today and yesterday, might change by 10 - 30 times. Best, Chenyi -- 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 dealii+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/19066ee5-29e1-4281-afbc-4daa56e37d04n%40googlegroups.com.
[deal.II]
Dear dealii group, I have a question regarding the computational efficiency of two styles of codes. For example, when computing the residual of momentum balance, one can realize d_epsilon \cdot sigma (the dot production between the test function of strain and stress) by 1. compute multiplications one by one and sum them up for (const unsigned int i : fe_values_u_p.dof_indices()) { for (unsigned int m=0; m bfem_i = ShapeTools::elastic_strain_shape(fe_values,i,q); for (const unsigned int i : fe_values_u_p.dof_indices()) { cell_rhs(i) = contract2(bfem_i, stress) * fe_values_u_p.JxW(q); } Do both perform the same? or I should prevent the loop style in dealii to speed up the computation? Best, Chenyi -- 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 dealii+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/CANmBX4QdwnBiJx8yA1XTo%3DOkSuoqa7JjO4Hsc7os%3D9s-wSw_iA%40mail.gmail.com.
[deal.II] initial values of different-type data
Hi dealii developers, I would like to ask(confirm)whether zero is the initial value of different-type data, e.g. for double, int, Vector and Tensor. For example, if I just define double x;. Does x have an initial value of 0.0? Moreover, in my task, I set initial values of a variable, e.g. "a", to be one in the whole domain. When I assemble the residual vector, I need to first compute the value of "a" at quadrature points (which turns out not exactly 1.0). Consequently, when I compute, e.g. "3*a^4-2*a^3+a^2", the answer is not zero. Is it avoidable? Best, Chenyi -- 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 dealii+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/CANmBX4TfES8vYsWN4uiEd5BTGJrC8Mkpaw%3DHzTO%3Ddo8-7VJD0g%40mail.gmail.com.