TECHNICAL WORKING GROUP
Objectives and Guidelines
The primary goal of the ICCS technical working group (TWG) is to lead in defining a roadmap for leveraging state-of-the-art emerging technologies and methodologies in providing early adoption strategies to the high-performance scientific computing communities.
The TWG research will foster mutually advantageous growth of academic and vendor group collaborations.
The TWG research and outreach efforts will enable broad range of scientific research; generate high-performance scientific computing awareness with emerging technologies in next generation and the broader community.
To achieve the leadership role and meet the mission objectives, the TWG will comprise of four groups listed below along with their designated leads and participants.
TWG research products in the form of codes, reports, publications etc will be made available to the community via the ICCS portal.
The detailed annual report highlighting the overall progress will be released during the ICCS workshops.
The TWG leads will define their working procedures, milestones, deliverables and timelines. TWG leads will also define their meeting requirements.
The TWG progress meetings with the leadership groups will be held every 2 months.
The ICCS resources will be accessible to TWG.
The Applications and Libraries (AL) group will explore the evolution of HPC applications in science and engineering with emerging architectures in conjunction with exascale roadmap. The wide range of applications will be classified based upon the underlying techniques and/or algorithms. The group in close collaboration with other TWG will develop and recommend strategies to achieve optimal performance from new technologies.
The Programming Models (PM) group will narrow down technologies that can enhance programming performance, productivity and, portability on heterogeneous systems. These attributes will be addressed by exploring directives methodologies, auto-tuning, tools for profiling and debugging and existing languages and APIs.
The Hardware Resources (HR) group will explore the efficacy and the underlying techniques for developing large-scale heterogeneous systems leading to exascale roadmap. To measure the overall performance metrics of different machines and the group will also define figure of merit based upon common and essential architectural features such as memory hierarchy, bandwidth latencies, and most importantly power efficiency.
The Performance and Analysis (PA) will develop and recommend methodologies and tools for quantitatively measuring the performance of applications on existing and future systems. The tools will analyze and model optimal scaling of applications on various systems.