@mlange05 , @navjotk
I updated the visco acoustic version. The boundary conditions are still bad, the Born function is now correct, I just need to fix the gradient (probably errors due to recording data inside the boudary).
MAIN THING
The main thing about this update is two cells I added before the stencil. he stencil is very complicated if directly derived from the symbolic expression. However in practice, the density (rho) is known and is not dependent of time. It is then a lot more efficient to pre-compute rho.grad(rho) (you need to store the d/dx and d/dy separately) and just use it as a constant vector in the propagation. It saves, for nt time step, nt computations of a gradient.
@mlange05 , @navjotk I updated the visco acoustic version. The boundary conditions are still bad, the Born function is now correct, I just need to fix the gradient (probably errors due to recording data inside the boudary).
MAIN THING The main thing about this update is two cells I added before the stencil. he stencil is very complicated if directly derived from the symbolic expression. However in practice, the density (rho) is known and is not dependent of time. It is then a lot more efficient to pre-compute rho.grad(rho) (you need to store the d/dx and d/dy separately) and just use it as a constant vector in the propagation. It saves, for nt time step, nt computations of a gradient.