add changes to fix the order #5630
Conversation
enabling reordering as default
ElyesAhmed
force-pushed
the
fix_sfc_order_alugrid
branch
from
59a6952
to
6c2e957
Compare
|
Note that flipping these defines in opm code does nothing. You are entirely at the mercy of however dune-alugrid itself was built as the defines are not used in headers but in translation units. |
This is an important point I was not aware of until today. Is there a way to detect either compile time or run time how dune-alugrid was built? |
|
No there is no indication of this. It's just an #ifdef in its gridfactory.cc that exits early. What makes it even more fun is that gridfactory.cc can be built inline (ie the .cc file is included in the .hh file) if you build dune-alugrid with COMPILE_ALUGRID_INLINE define. But you can't just set this for a library that was built without it as you'd get duplicate symbols. So as I said, at the mercy of how the library is built. |
|
jenkins build this serial please |
|
jenkins build this serial failure_report please |
|
can we try again! |
|
jenkins build this serial failure_report update_data please |
Reason: PR OPM/opm-simulators#5630 opm-common = 2a46477d267446991485e6ff8a13d2865c66d519 opm-grid = 36069f06e5fe321c2dd161a7614dc95be5d05978 opm-models = 6a0cb4b092f82c2a90193f1a16094c0fbdf503c3 opm-simulators = 6a0cb4b092f82c2a90193f1a16094c0fbdf503c3 ### Changed Tests ### * spe12_alugrid
|
jenkins build this serial opm-tests=1234 please |
|
jenkins build this serial failure_report update_data please |
|
Existing PR OPM/opm-tests#1234 was updated |
Reason: PR OPM/opm-simulators#5630 opm-common = 12334dc673e68d09b94427e020632dfb7798b87d opm-grid = 36069f06e5fe321c2dd161a7614dc95be5d05978 opm-models = 0abd925fd2a73fa8dabefd0579ea9875456f54be opm-simulators = 0abd925fd2a73fa8dabefd0579ea9875456f54be ### Changed Tests ### * spe12_alugrid
|
jenkins build this serial opm-tests=1234 please |
Automatic Reference Data Update for PR OPM/opm-simulators#5630
|
Data update merged, will merge this once the data update Jenkins job has started. |
The new reference solutions were installed 30 minutes ago. Did you accidentally forget to merge this? |
Yes, had to go to a meeting, sorry! |
|
Just a note. According to my failed tests (because of missing rebase). The changes to the well solutions are quite large: https://ci.opm-project.org/job/opm-common-PR-builder/7522/testReport/(root)/mpi/compareECLFiles_flow_blackoil_alugrid_SPE1CASE2/ |
We are aware of this, it is because the previous solution was wrong. The bugfixes in this PR apparently had a big influence. |
|
Well, our code should not rely on internal ordering. If it does then we are doomed (not only for ALUGrid...). Maybe somebody could describe what this is supposed to fix. Where does this ordering matter? |
This PR enables SFC (Space-Filling Curve) reordering in AluGrid as the default choice. The changes has been tested both with and without MPI across multiple cases from the opm-tests and opm-data!