[MeshMoving] Fix Race Condition In MoveMesh #11811
Conversation
matekelemen
added
C++
Parallel-SMP
| block_for_each( | ||
| rModelPart.Nodes(), | ||
| [&rTransform, time](Node& rNode){ | ||
| rTransform, | ||
| [time](Node& rNode, ParametricAffineTransform& rTLSTransform){ | ||
| const array_1d<double,3>& initial_position = rNode.GetInitialPosition(); | ||
| noalias(rNode.GetSolutionStepValue(MESH_DISPLACEMENT)) = rTransform.Apply( | ||
| noalias(rNode.GetSolutionStepValue(MESH_DISPLACEMENT)) = rTLSTransform.Apply( | ||
| initial_position, | ||
| time, | ||
| rNode.X0(), |
There was a problem hiding this comment.
this is the main change, the rest is style/docs/renaming
| // Main Authors: Máté Kelemen | ||
| // | ||
|
|
||
| #ifndef KRATOS_MESH_MOVING_LINEAR_TRANSFORMATION_INCLUDED |
There was a problem hiding this comment.
Overall i am ok with the change, however i see you rename methods so this makes it an API changer.
are u sure those methods are not used by other people?
If that is so (no impact on other people) than just merge...otherwise let's discuss the way forward to minimize impact onto other people
I'm not sure how many people are using it. I originally wrote these ~3 years ago for @mpentek. Maybe he has a better idea of other potential users. |
matekelemen
dismissed stale reviews from philbucher and RiccardoRossi
via
4841532
Merge queue setting changed
roigcarlo
removed this pull request from the merge queue due to the queue being cleared
|
Can we merge? |
|
go ahead |
Main change
Move the
ParametricAffineTransforminblock_for_eachfrom a lambda capture to TLS, because it's not thread-safe. The related discussion is in #11804.Other changes
LinearTransformwas a misnomer (because it also involved a translation) so I renamed it toAffineTransform