I am starting to implement a DGFEM discretization of the Euler equations for 3D
Since I want to use a third-order Runge-Kutta time discretization, I have a
local mass matrix that I need to invert just once. My first question is how do I
5 block matrices in each element? The access to these matrices should be fast!
Another problem arising in turbomachine computations is the periodic boundaries,
since I want to compute just one blade. Supposing that I have 36 blades in a
turbomachine, for one of the blades I have two periodic surfaces that differ by
a rotation of 2*pi/36. Does anyone have good ideias how to implement it?
Thanks in advance.