Optimising a Parallel Conjugate Gradient Solver

Taking the result that 100x100x100 calculation will take approximately 45 day to run 30 nanoseconds (30,000,000 steps at stability limit) I (Frank) propose improvement to the CG solver
Following the paper "Optimising a Parallel Conjugate Gradient Solver" (https://doi.org/10.1137/S1064827596302205) we should implement a version of Yale format for matrices and optimized routines then proceed to parallel routines.