The simulation of oil or gas reservoirs requires compute-intensive
simulation methods. That part of such a simulation that consumes most
of the compute time is the repeated solution of very large linear
systems. The matrix of such a linear system has a special structure:
it is a sparse matrix.
In this project selected compute intensive operations on sparse
matrices are optized for future parallel computer architectures. In
particular accelerators like graphic processors and the Xeon Phi
processor family are of interest. One of the operations of interest
ist the multiplication of a sparse matrix with a (dense) vector.