Closed totto82 closed 6 years ago
To me this looks like a quite large change - more than a post merge cleanup? At least the transfer through the MessageTypeBuffer
comes a bit out of the blue?!
To me this looks like a quite large change - more than a post merge cleanup?
It is supposed to be more than a post merge cleanup. It is more like the final step in my work of re-distributing the work for output between opm-output, ewoms and opm-simulator. With this the output code almost scales (only tested up to 8 cores on my machine). Before it used to increase with increased number of nodes.
[...] It is more like the final step in my work of re-distributing the work for output between opm-output, ewoms and opm-simulator.
OK
With this the output code almost scales (only tested up to 8 cores on my machine). Before it used to increase with increased number of nodes.
Well - that is certainly impressive!
jenkins build this please
OK; with test added to test_summary
I am happy with this.
1) Add well communication to data::wells. 2) Pass cell data used by RFT via the data::wells::completion. Motivation: Avoid communication of all cell data. 3) Bugfix. Store global index instead of active index in completions. RFT data is wrong for MPI simulations on current master.
Merge together with downstream PRs.