On Fri, 20 Jul 2018 at 12:34, Lawrence Mitchell <[email protected]> wrote:

Dear petsc-dev,

I'm once again revisiting doing "proper" checkpoint-restart cycles. I would like to leverage the existing PETSc stuff for this as much as possible, but I am a bit lost as to what is implemented, and what is missing.

I have:

- A (distributed) DMPlex defining the topology

Not sure if it will be really useful, but I have some code I can share with you to save/load DMPlex objects using PETSc's native binary viewers (either stdio or MPI-IO). A minor (or big?) issue is that loads are sequential, you have to distribute/overlap afterwards, but this is the way it works with most mesh readers. I should be really putting that code in PETSc, but... 1) It is most likely not 100% general, I'm not supporting the new DMField for coordinates. 2) Herr Prof. Matt seems to be in favor of all these fancy HDF5-based formats (https://xkcd.com/927/) -- Lisandro Dalcin ============ Research Scientist Computer, Electrical and Mathematical Sciences & Engineering (CEMSE) Extreme Computing Research Center (ECRC) King Abdullah University of Science and Technology (KAUST) http://ecrc.kaust.edu.sa/ 4700 King Abdullah University of Science and Technology al-Khawarizmi Bldg (Bldg 1), Office # 0109 Thuwal 23955-6900, Kingdom of Saudi Arabia http://www.kaust.edu.sa Office Phone: +966 12 808-0459