Search

Emerging computer architectures and systems that combine multi-core CPUs and accelerator technologies, like many-core Graphic Processing Units ( GPUs) and Intel's Many Integrated Core ( MIC) coprocessors, would provide substantial computing power for many time-consuming spatial-temporal computation and applications. Although a distributed computing environment is suitable for large-scale geospatial computation, emerging advanced computing infrastructure remains unexplored in GIScience applications. This article introduces three categories of geospatial applications by effectively exploiting clusters of CPUs, GPUs and MICs for comparative analysis. Within these three benchmark tests, the GPU clusters exemplify advantages in the use case of embarrassingly parallelism. For spatial computation that has light communication between the computing nodes, GPU clusters present a similar performance to that of the MIC clusters when large data is applied. For applications that have intensive data communication between the computing nodes, MIC clusters could display better performance than GPU clusters. This conclusion will be beneficial to the future endeavors of the GIScience community to deploy the emerging heterogeneous computing infrastructure efficiently to achieve high or better performance spatial computation over big data. [ABSTRACT FROM AUTHOR]

The "memory wall" is an architectural property introducing high memory access latency that can manifest application performance, and this wall becomes even taller in the context of big data. Although the use of GPU-based systems could achieve high performance, it is difficult to improve the utilization of GPU systems due to the "memory wall". The intensive data exchange and computation remains a challenge when confronting applications with a massive memory footprint. Quantum-mechanics-based ab initio calculations, which leverage high-performance computing to investigate multi-electron systems, have been widely used in computational chemistry. However, ab initio calculations are labor-intensive and can easily consume more than hundreds of gigabytes of memory. Previous efforts on heterogeneous accelerators via GPU and CPU suffer from high-latency off-device memory access. In this paper, we introduce heterogeneous processing-in-memory (PIM) to mitigate the overhead of data movement between CPUs and GPUs, and deeply analyze two of the most memory-intensive parts of the quantum chemistry, for example, the FFT and time-consuming loops. Specifically, we exploit runtime systems and programming models to improve hardware utilization and simplify programming efforts by moving computation close to the data and eliminating hardware idling. We take a widely used software, the QUANTUM ESPRESSO (opEn-Source Package for Research in Electronic Structure, Simulation, and Optimization), to perform our experiments, and our results show that our design provides up to 4. 09 × and 2. 60 × of performance improvement and 71% and 88% energy reduction over CPU and GPU (NVIDIA P100), respectively. [ABSTRACT FROM AUTHOR]