I’m using JuMP to solve lots of (fairly) small problems sequentially. Is there a major reason for/against reusing solver objects? I.e. better (A) or (B):
A
for i in 1:50
m = Model(solver = CbcSolver())
...
solve(m)
...
end
B
s = CbcSolver()
for i in 1:50
m = Model(solver = s)
...
solve(m)
...
end