Scheduling with Precedence Constraints in Heterogeneous Parallel Computing
Published:
Modern parallel computers are increasingly likely to be heterogeneous, comprising, for example, multiple powerful accelerators, such as GPUs, in addition to a host of multicore CPUs. But to exploit the awesome potential of this new landscape, we must ensure that computational jobs are executed by the processing resources which are best suited to handle them. My research was focused on developing cheap and effective heuristics for such scheduling problems.