A Performance Study of a Dual Xeon-Phi Cluster for the Forward Modelling of Gravitational Fields

Figure 5

Partitioning by prisms. Each thread requires a separate memory location.