> On Oct 24, 2016, at 11:15 AM, Anton Popov <po...@uni-mainz.de> wrote: > > Our code breaks after compiling with recent branch > (barry/fix-matmpixxxsetpreallocation-reentrant) > > I figured out it's because of this: > > On 09/23/2016 05:19 AM, Barry Smith wrote: >> branch barry/fix-dmdacreate in next for testing pulls the DMSetFromOptions() >> and DMSetUp() call out of the DMDACreateNd(). It also removes any special >> handling of negative M,N, and P arguments. > > Is it going to be the standard behavior of DMDACreateNd() in future releases?
Yes, the new code fixed some issues with multiple DMDAs and is more like the rest of the PETSc design. > > Anton >
