Modelling the shell
The inversion of the data dates back to the 1930’s.
Variations dominate radially and make the model 1D (depth is the main factor).
Historically it is common to analyse the data by 2D projections, academic peers now want more.
Notes:
From this little number juggle we not only learn that bad programs scale better.
Knowing that processing data in big chunks gives better speedup we conclude that the more coarse grained our parallelization is, the better performance we expect.