Are you using multi-threading or multi-processing? The bundled UMFPACK solver is single threaded, the theoretical speedup is about 1.
I see 2 choices: (a) Use external multi-threaded direct solver like PARDISO, you can pass the reordered CSR data to it and (b) use MPI multi-processing and distributed parallel direct solver MUMPS or SuperLU_DIST from PETSc or Trilinos. 在2024年7月18日星期四 UTC+8 21:10:43<[email protected]> 写道： > Hi all, > > I am writing a parallel code for the Newton-Raphson scheme and testing the > speedup for different parts of the scheme. The performance is attached in > the figure below. It shows that the speed up of the direct solver is 1, > degrading the performance of the whole scheme. Other parts including > assembling the global system, updating internal variables exhibit linear > speedup. > > Has anyone met the same issue with the linear solver? are there any > solutions to gain better performance for using a linear solver? > > Thank you, > Giang > > -- 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]. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/fff1f05b-4497-42b8-9b17-c0a7c03a3249n%40googlegroups.com.
