Differentiable Fortran Applications with LFortran and Enzyme

It is now possible to backpropagate through Fortran, C, or C++ simulation code, integrating it into JAX and PyTorch as a high-performance differentiable physics engine. Enzyme enables automatic differentiation at the LLVM IR level, allowing gradients to be derived from existing source code without rewriting legacy physics simulations. This method provides a critical solution for optimizing nonlinear systems without the need for extensive rewrites.