Computing Low-Rank Approximation of a Dense Matrix on Multicore CPUs with a GPU and Its Application to Solving a Hierarchically Semiseparable Linear System of Equations

Figure 5

Illustration of the hybrid QP3 implementation.

(a) Column-wise (left) and blocked (right bar) algorithm

(b) Ratio of exact and estimated conditioner number of leading triangular factors after the QPR factorization

(c) StruMF solution time using task-1 (left), task-2 (middle), and task-1 and task-2 (right bar) postprocessing

(d) StruMF solution time with QP3 (left) and QPR (right bar)

(e) Hybrid paradigm

(f) Algorithmic flow