Re: [deal.II] step-40. petsc with openmp
Thank you for the answer Regards, Kartik On Tuesday, July 11, 2017 at 5:53:48 PM UTC+2, Wolfgang Bangerth wrote: > > On 07/11/2017 08:33 AM, Kartik Jujare wrote: > > > > Does this still hold true of the Petsc wrappers not being thread-safe? > > Yes. But it's not the wrappers that are the problem, it's that PETSc > itself is > not thread safe. > > 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. For more options, visit https://groups.google.com/d/optout.
Re: [deal.II] step-40. petsc with openmp
On 07/11/2017 08:33 AM, Kartik Jujare wrote: Does this still hold true of the Petsc wrappers not being thread-safe? Yes. But it's not the wrappers that are the problem, it's that PETSc itself is not thread safe. Best W. -- Wolfgang Bangerth email: bange...@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. For more options, visit https://groups.google.com/d/optout.
Re: [deal.II] step-40. petsc with openmp
Hi Timo, Does this still hold true of the Petsc wrappers not being thread-safe? Regards, Kartik Jujare On Friday, March 14, 2014 at 3:24:18 PM UTC+1, Timo Heister wrote: > > > I am trying to use openmp in the element assemble procedure of step-40. > Has > > anyone done this before? Any advice is appreciated. > > Our implementation of PETScWrappers is not thread-safe, so you can not > write into matrices/vectors concurrently. That means using TBB (what > we use inside deal.II) or OpenMP won't help you much, because it would > require locking. > > Our Trilinos wrappers are thread-safe, though. > > Even if you change the implementation of PETScWrappers to allow this > (it wouldn't be too difficult, ask me if you want to know more), you > still have the problem that anything in the linear solver (matrix > vector products, preconditioners, ...) are likely not running > multi-threaded. > > > My purpose is to increase the assembly time by making use of > multi-threads > > of each core. Is this feasible? > > You know you can run one MPI task per core, right? > > -- > Timo Heister > http://www.math.clemson.edu/~heister/ > -- 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. For more options, visit https://groups.google.com/d/optout.
