ale3d #248
ale3d #248
Conversation
numbers as the fragment IDs. Using fragment IDs in evaluateDerivatives to turn off forces between fragments when they are moving away from each other.
This happens in FEusion/SPH when we update the smoothing lengths with the scaling factor.
This reverts commit 3bd393c.
…v:7999/ale/rzspheral into ale3d
|
|
||
| if(APPLE) | ||
| set(SHARED_EXT "dylib") | ||
| set(${lib_name}_libs libpolytope.${SHARED_EXT}) | ||
| endif() | ||
|
|
||
| if(ENABLE_STATIC_TPL) | ||
| string(REPLACE ".${SHARED_EXT}" ".a;" ${lib_name}_libs ${${lib_name}_libs}) | ||
| endif() |
There was a problem hiding this comment.
Can we make this a more generic Macro?
There was a problem hiding this comment.
Which macro do you want changed, the APPLE or SHARED_EXT?
There was a problem hiding this comment.
Not the CMake variable names but creating a CMake macro/function (we define some of them in spheral/cmake/spheral/) for switching out the library extensions to reduce code duplication in each tpl file.
src/Mesh/MeshInline.hh
Outdated
| @@ -369,7 +371,11 @@ minimumScale() const { | |||
| result = std::min(result, (face.position() - zonePosition).magnitude2()); | |||
| } | |||
| } | |||
| #ifdef USE_MPI | |||
There was a problem hiding this comment.
The allReduce interface is already guarded by USE_MPI, the changes guarding all of these calls shouldn't be necessary. I think the issue here is that our usual thin interface for communicator when building without MPI support is also guarded out. Communicator headers should be included with or without MPI support in order to support the current allReduce interface.
…fdefs with USE_MPI.
…tion on 1 process within a parallel run.
…suji1/loadbalance
…tlab.llnl.gov:7999/ale/rzspheral into ale3d
…tlab.llnl.gov:7999/ale/rzspheral into ale3d
|
@mdavis36 for some reason I can't respond to your comment about allReduce; I've moved the Communicator.hh stuff to Utilities, and have removed all of the unnecessary ifdefs with USE_MPI. |
Summary