The 27th International Heterogeneity in Computing Workshop and the 16th International Workshop on Algorithms, Models and Tools for Parallel Computing on Heterogeneous Platforms

Heterogeneity is now a pervasive characteristic of computing. From the macrolevel, where networks of distributed computers composed of diverse node architectures are interconnected with potentially heterogeneous networks, to the microlevel, where deeper memory hierarchies and various accelerator architectures are increasingly common, the impact of heterogeneity on all computing tasks is profound. High-performance computing (HPC) clusters, clouds, and data centers today exhibit heterogeneity at various levels of their software stacks and hardware topologies. These platforms commonly feature tight integration of multicore CPU processors and accelerators such as graphical processing units (GPUs), field programmable gate arrays (FPGAs), Intel Xeon Phis, and so on, empowering them to provide not just unprecedented computational power but also to address the critical concern of energy efficiency. The TOP500 list1 showcases the snapshot of dominant hardware trends in HPC. It currently features around 150 systems that contain multicore CPUs integrated with accelerators/coprocessors. While multicore CPU space is dominated by three processors, Intel Xeon E5 (Broadwell), Intel Xeon Gold, and Intel Platinum, the accelerators are quite diverse: GPUs that include Tesla V100, Tesla P100, Tesla K80, from NVIDIA, Vega 20 from AMD and Many-cores such as Intel Xeon Phi 5120D, Intel Xeon Phi 7120P, Intel Xeon Phi 5110P, Matrix-2000, PEZY-SC2, and so on. Many heterogeneous platforms (of similar flavor as those in this list) are fueling rapid advances not just in scientific application areas but also in the data science fields of big data analytics, deep learning, and so on. The perennial programming challenges on such platforms have been to maximize their efficiency and resource utilization since the platforms are ever-changing with novel and multifarious parallel architectures. New ideas, innovative algorithms, and specialized programming environments and tools are constantly needed to address the challenges. The Heterogeneity in Computing Workshop (HCW) and International Workshop on Algorithms, Models and Tools for Parallel Computing on Heterogeneous Platforms (HeteroPar) have been the flagship forums bringing together researchers to discuss these challenges and the solutions. The wide range of topics deliberated includes, to name a few, heterogeneous parallel programming paradigms, algorithms, models and tools for performance and energy optimization on heterogeneous platforms, and fault tolerance of parallel computations on heterogeneous platforms. The accepted articles in the workshops this year covered topics, techniques, and applications, exhibiting lucidly the depth, breadth, and growth of the heterogeneous computing field. Two promising developments are, however, apparent. The first is the slow but steady adoption of FPGAs as yet another acceleration technology competing with GPUs and Intel Xeon Phis in HPC and data science. This is evidenced from the increasing number of publications submitted to the two workshops featuring FPGAs for accelerating software algorithms in the fields of cryptography, anomaly detection, and so on. The second is the growing awareness of energy of computing as a serious environmental concern and a grand technological challenge. Energy is now considered a fundamental design constraint along with performance in all computing settings. The number of submissions focusing on single-objective optimization for performance with energy constraints and biobjective optimization for both performance and energy on heterogeneous platforms is steadily increasing. This special issue contains five selected articles from the HCW'2018 and HeteroPar'2018 workshops. We hope you find the articles, whose summaries are below, informative, interesting, and thought-provoking.