You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
ParOpt has seen some recent updates that have changed it's API and added support for general sparse constraints that pyOptSparse could take advantage of. See here
Potential solution
We need to update the ParOpt interface to return sparse and dense constraints, and their jacobians, separately.
As a first attempt, we could simply treat linear constraints as sparse and nonlinear constraints as dense, I think this would already result in a large performance improvement. After that we could look into something like determining which constraints are sparse using the density of each row in the full constraint Jacobian.
The text was updated successfully, but these errors were encountered:
Description of feature
ParOpt has seen some recent updates that have changed it's API and added support for general sparse constraints that pyOptSparse could take advantage of. See here
Potential solution
We need to update the ParOpt interface to return sparse and dense constraints, and their jacobians, separately.
As a first attempt, we could simply treat linear constraints as sparse and nonlinear constraints as dense, I think this would already result in a large performance improvement. After that we could look into something like determining which constraints are sparse using the density of each row in the full constraint Jacobian.
The text was updated successfully, but these errors were encountered: