Description
Currently, the textures used to compute the angle towards north and the Coriolis parameter contains inaccuracies close to the boarders.
This is because these textures are coarsely sampled from the computational domain. Since texture data is defined in "cell centers" with $\Delta x $ intervals, the linear interpolation is not defined in the first and last
To fix this (for these two textures), we can normalize the texture fetches within #defines, so that we know the sizes of $\Delta x $ and $\Delta y $ , and also that we fill the texture so that the new interval covers the full computational domain. Note that this $\Delta x $ is not the same as the $\Delta x $ in the simulator.
This is not a serious bug, but it is nonethereless a bug which would be nice to get out of the way.