Hey all,
I am fairly new to julia and have been using gridap for PDE constrained optimization and the high level interface makes it really user friendly. But as I am transitioning to quasilinear PDEs where I am forced to update the fem matrices inside my optimization loop (specifically when solving the forward and adjoint PDEs for calculating my reduced gradients and objective functional), the allocations are really slowing down my optimizer. Even when using some lower-level syntax like pre-allocating the matrices and the assembler and then using assemble_matrix!. Is there a better way to do this or should I maybe consider switching to ferrite for these more complex problems?