Temporary change to how tractions are assembled into the global force vector. Currently, tractions have to be pre-integrated in the config file and attributed to their nodes, so we can just directly apply the integrated point loads into the force vector.
Why?
Previously used a for-loop that performed unnecessary iterations. Caused significant computational expense in fine meshes.
How?
Previously used a for-loop that looped through all elements and nodes to see if the traction node was part of the element, now we just directly assemble the forces.
Testing?
RunSim MasterConfig at high element counts, RunTests
Change reduces Fext computational time from 38s to 1.4s at 3e5 dofs.
Anything Else?
Temporary implementation to improve efficiency while we work towards edge element integration of tractions.
What?
Temporary change to how tractions are assembled into the global force vector. Currently, tractions have to be pre-integrated in the config file and attributed to their nodes, so we can just directly apply the integrated point loads into the force vector.
Why?
Previously used a for-loop that performed unnecessary iterations. Caused significant computational expense in fine meshes.
How?
Previously used a for-loop that looped through all elements and nodes to see if the traction node was part of the element, now we just directly assemble the forces.
Testing?
RunSim MasterConfig at high element counts, RunTests Change reduces Fext computational time from 38s to 1.4s at 3e5 dofs.
Anything Else?
Temporary implementation to improve efficiency while we work towards edge element integration of tractions.