Exploring the frontiers of computing science and technology: efficiently utilizing multicore and many-core processors
Multicore (e.g., Intel Westmere and IBM Power7) and many-core (e.g., NVIDIA Tesla and AMD FireStream graphics processing units (GPUs)) microprocessors are enabling more computeintensive and data-intensive computation in desktop computers, clusters, and leadership supercomputers. However, efficient utilization of these microprocessors is still a very challenging issue.
Their differing architectures require significantly different programming paradigms when adapting real-world applications. The actual porting costs are actively debated, as well as the relative performance between GPUs and CPUs. The goal of this special issue is to address such issues by assembling some of the latest researches on efficiently utilizing multicore and many-core processors in real-world applications, and their strategies for coping with those challenges.
The invited papers in this special issue represent amplified works originally presented at the Frontiers of Multicore Computing Conference 2010, held at the University of Maryland, Baltimore County in August 2010. The selected papers cover representative research addressing the issues above.
There are four papers addressing the issues related to CPUs and two on GPUs. Seelam et al. report their experiences in building and scaling enterprise business analytics benchmark, report generation, and rendering, on an IBM Power7 multicore system with eight Power7 cores and 32 hardware threads [1]. The paper by Tracy and Brown presents a multithreaded physics software design to eliminate overhead associated with bodies at rest and consequently accelerate physics simulation in large, continuous virtual environments on Intel multicore processors [2]. The paper by Hammond et al. presents multilevel performance analysis for the computational chemistry software, NW Chem, in Blue Gene/P and on two large-scale clusters [3]. The paper by Simon et al. provides a performance evaluation and investigation of the astrophysics code, FLASH, for a variety of Intel multicore processors [4]. The paper by Blattner and Yang presents the key steps in porting one data assimilation algorithm to GPU [5]. The paper by Malik et al. examines the programming paradigms of Compute Unified Device Architecture (CUDA), OpenCL, The Portland Group Accelerator Compiler (PGI), and MATLAB through developing kernels from the Numerical Aerodynamics Simulation (NAS) parallel benchmarking suite [6].