We should use in place `solves` and preallocated workspaces instead of allocating at each iteration, i think this is the reason why FH is so slow. Ultimately, all the NLPModels API functions for the FH problem should be allocation free.
We should use in place
solvesand preallocated workspaces instead of allocating at each iteration, i think this is the reason why FH is so slow.Ultimately, all the NLPModels API functions for the FH problem should be allocation free.