Created by: mhoemmen
@trilinos/belos @trilinos/tpetra @cgcgcg @vbrunini
Description
- Fix #4639 (closed) (a Tpetra bug)
- Fix #4633 (closed) (a Tpetra not quite bug but weird wrinkle)
- Fix #4626 (closed)
The solution for #4626 (closed) needs an optimization (always allocate contiguous DualViews, by making them nonowning from a 1-D memory pool), but at least should avoid the performance issue of repeatedly allocating and freeing CUDA UVM allocations.
Related Issues
- Closes #4626 (closed), #4639 (closed), #4633 (closed)