Report on the Workshop on HPC Application Acceleration

Beijing, China
October 21-23 2010

Workshop conveners:

Gregory B. Newby (NSF Principal Investigator), Arctic Region Supercomputing Center, USA
Tarek El-Ghazawi (NSF Co-Principal Investigator), The George Washington University, USA
Depei Qian (China Coordinator), Beihang University, China
Xue-Bin Chi (China Co-Coordinator), Chinese Academy of Sciences
Guangwen Yang (China Co-Coordinator), Tsinghua University

Workshop on High Performance Computing Application Acceleration


Researchers in the field of high performance computing (HPC) worldwide have similar interests in application acceleration, yet there has been comparatively little direct communication between researchers in China and those in the US. This includes formal conferences and workshops, and ongoing research collaborations. In this report, we describe a workshop, held in China in October 2010, which took steps to open pathways to effective research collaborations on HPC application acceleration between leading institutions and scholars in both countries.

The workshop was sponsored by the US National Science Foundation (NSF), which funded travel for the US-based scholars to China. China participants were sponsored by their home institutions. China local events such as evening banquets and tourism for US participants, as well as a logistical coordinator, were funded by the three co-coordinating institutions. These were Beihang University, the Chinese Academy of Sciences, and Tsinghua University. Supplemental support came from Intel (China) and NVIDIA (US).

Workshop Overview

High performance computing (HPC) is at the leading edge of solving computationally based problems on powerful next-generation computer systems. These systems incorporate advances in hardware and software that create a need to consider how best to use them, in order to take full advantage of their unique capabilities and architectures. The workshop examined trends, issues, concepts, challenges, solutions, and best practices for effective utilization of today and tomorrow's HPC systems.

Workshop coordinators met in Beijing in April, and again in Illinois in July 2010, to plan the workshop. Based on mutual interests, two major application areas were chosen to structure the workshop: earth systems (including climate and weather) and molecular dynamics (including computational chemistry).

Fourteen attendees from the US were matched by a similar number from China. Participants were selected based on their knowledge of application acceleration for HPC. They included senior faculty members, junior faculty members, a doctoral student, and industry participants. For US participants, an effort was made to include members of under-represented groups.

A Web site includes links to most of the presentations, as well as biographical information about participants (who are named in an Appendix to this report). The site is:, with username "faster" and password "nsf".

Major Outcomes

Participants got to know each other, across national and disciplinary/academic boundaries, and found there were many common interests. They included fundamental challenges to HPC (parallelization, latency, bandwidth, job throughput), practical issues such as time to solution, use of community- and commercially-supported software, and other pragmatic concerns. As might be expected, these concerns crossed disciplinary and national boundaries, and are commonplace in computational science.

Scientific challenges were also described, from problems of maintaining accurate model physics, verification of model outcomes, and issues of coupling/interfacing among different constituent models. The workshop participants tended to have a formal background in a disciplinary science (i.e., doctoral degrees in areas such as chemistry and atmospheric science), but would be accurately characterized today as computational scientists. They utilize computational models as primary mechanisms for scientific discovery and innovation.

As the workshop's end, participants had identified a variety of dominant themes:

International collaboration is difficult, with insufficient formal mechanisms to seek funding for joint projects.

The workshop helped demonstrate many overlapping interests among participants, and led towards first steps on collaborative work.

Molecular dynamics, earth systems, and a relatively small number of other scientific application areas are very good candidates for application acceleration using current and forthcoming HPC technologies at the largest scale; yet, these advances will take significant effort.

Nearly every other area of science is aided by computation, including data-intensive computing, analysis, and visualization. Many of the findings at the largest scale of computing are applicable or helpful to a broad range of scientific domains.

US-China cooperation is viable and desirable for pursuing application acceleration for HPC.

Next Steps

Several participants planned to attend the AITP workshop on China HPC at the annual IEEE/ACM Supercomputing conference, and would discuss further collaboration. This would potentially introduce additional partners from China and the US.

Themes of interest for future workshops, collaborations or other joint work were identified. Attendees were particularly interested in discussing tools for HPC, including compilers, libraries, algorithms and technologies. They would also like to focus on related technologies including runtime systems, multi/manycore processors, coprocessors (GPU, FPGA), and operating systems.

Also of interest were other disciplines with a computational science focus. Areas mentioned include bioinformatics and life sciences. Participants were interested in forming multi-institutional transnational teams for these and other disciplinary areas, to work on collaborations, joint funding, and workshops or other information sharing.

A practical concern of many participants has to do with HPC center operation and planning. There was interest in a multinational consortium for this, perhaps a US-China consortium. This is in recognition that the world's largest supercomputers are in the US and China, and it is these two countries that, more than any others, are at the leading edge of next-generation HPC centers and systems.

For joint fundraising, the NSF-China deadline is in March annually, but down-selection of potential projects happens earlier. Near-term planning is needed to identify some projects of mutual interest that China participants could submit to NSFC for funding, anticipating involvement and pursuit of US-based funding (from the NSF and other agencies as appropriate).

Finally, there is interest in position papers or other statements of the concepts in this workshop report, perhaps with additional detail or for particular audiences. The workshop co-coordinators are encouraging of such statements, and will foster their creation.


The workshop was judged by attendees to be highly successful. It was intended as a first step in ongoing collaborations, and the remaining steps will be pursued during upcoming months and years. These steps are anticipated to occur at several levels. At the national and transnational levels, workshop co-coordinators and participants will interact with national funding agencies (particular NSF and NSFC) to foster joint funding opportunities and other mechanisms for multi-nation funding.

At the institution level, participants will develop memoranda of understanding, site visits, campus-based events, short- and longer-term visits, and other ways to make it easier to build deeper cross-border research relationships.

At the disciplinary level, additional workshops or meetings, including meetings that are subsidiary to larger meetings (such as the Supercomputing conference), will help bring in new people, and to insure discussions and outcomes are relevant to specifically identified areas of computational science.

At the individual level, the workshop served to introduce participants from the US and China, and the combination of formal presentations, informal events, and group sessions helped participants to identify areas of common interests. There were firm expectations of follow-up conversations, meetings, information sharing, and grant seeking.

Information sharing, and the development of new interpersonal relationships, was a major outcome of the workshop. The co-coordinators are committed to monitoring, mentoring, and otherwise supporting the development of new research collaborations, and to broaden participation among US and China researchers. A future workshop, anticipated for 2011, will look back at what worked and what needs further work. We hope to be able to report further progress on all aspects of collaborations and national-level coordination opportunities at that time.

Appendix: Workshop Agenda

Underlined text is hyperlinked to presentation files at the conference Web site.

Wednesday October 20

Pre-Workshop Planning


Coordination among some US and China delegates and co-coordinators

Thursday October 21

Pre-Workshop Events


Meet in hotel hotel lobby for Great Wall tour (optional)


Meet at CAS for welcoming reception & dinner (South Beauty Restaurant)


Return to hotel

Note: Some US and China delegates will miss Thursday's events

Friday October 22


Note: Breakfast is in on your own, in the hotel dining room

Workshop Opening


Workshop Welcome (D. Qian)


Workshop Background, Goals. Charge to Attendees (G. Newby)

China-US Joint Funding Opportunities
Session Chair: F. Williams, Arctic Region Supercomputing Center


NSF Programs and Initiatives for International Cooperation (Alexander P. DeAngelis, Director. China Office of the U.S. National Science Foundation, Embassy of the United States of America)


National Natural Science Foundation (Huai Chen, Division Director of the International Cooperation Department)


Discussion of additional Programs and Initiatives for International Cooperation


Coffee Break


Self-Introductions of US and China Delegates

Session 1: Climate, Weather and Earth Science
Session Chair: J. Li, Institute of Atmospheric Physics, Chinese Academy of Sciences


Recent Work in Earth System Modeling (X. Wang, Tsinghua University)


Climate Modeling Efforts in the Institute of Atmospheric Physics (Y. Yu)


GPU Acceleration Experience with WRF (J. Michalakes, NREL)



Lunch (hotel restaurant)

Session 2: Molecular Dynamics
Session Chair: T. El-Ghazawi, The George Washington University


Computational Chemistry Applications at NCSA/UIUC (V. Kindratenko)


Realistic Protein Folding with Next Generation Methodology and Supercomputer(G. Li, Dalian Institute of Chemical Physics, CAS)


Molecule/cluster statistical thermodynamics (MST/CST) (H. Wang, Institute of Mechanics, CAS)


(H. Guo)


Drug Discovery Based on Computational Protein Science (K. Yu, Shanghai Institute of Materia Medica, CAS)

Session 3: Systems
Session Chair: G. Yang, Tsinghua University


Accelerating Applications to Exascale (T. Sterling, Louisana State University)


Coffee Break and Group Photo


Accelerating Science with High Throughput Computing (HTC)(M. Livny, University of Wisconsin-Madison)

Session 4: Methodologies for Application Acceleration
Session chair: Zhong Jin, CNIC, CAS


Compilers and Tools at U. Houston (B. Chapman)


(T. El-Ghazawi & A. Kayi)



Methodologies for Application Acceleration (J. Dongarra, University of Tennessee and Oak Ridge National Laboratory)


Summary of First Day. Planning for Second Day (G. Newby)


Adjourn for the Day


Walk to Banquet Dinner (Quanjude Roast Duck Restaurant, off Wangfujing street), meet in hotel lobby.

Saturday October 23


Note: Breakfast is in on your own, in the hotel dining room


Session 5: Vendor Roles in Application Acceleration
Session Chair: X. Zhang, Xi'an Jiaotong University

Application Acceleration at IBM (M. Elnozahy)

Application Acceleration at Intel (D. Scott)

Application Acceleration at NVIDIA (S. Posey)


Session 6: Breakout Sessions on Applications Focus Areas and Acceleration Methodologies. Participants should join one or more sessions. Chairs will take notes.

Earth Systems, Weather and Climate

Molecular Dynamics

Libraries, Compilers and Languages

Architectures and Middleware


Coffee Break


Reports from Breakout Session #6 by Breakout Scribes



Lunch (hotel restaurant)


Session 7: Collaboration Planning
Session Co-Chair: X. Chi, Computer Network Information Center, Chinese Academy of Sciences
Session Co-Chair: T. El-Ghazawi, George Washington University


Closing Remarks. Summarizing Next Steps (G. Newby)


Workshop Adjourns


Optional Touring and Post-Workshop Events

Sunday October 24, Monday October 25

Post-Workshop Collaborations


Coordination among some US and China delegates to pursue research interests

Appendix: Workshop Participants

Note that there were also some drop-in participants from China.




Donnita Branch


Barbara Chapman

University of Houston

Debbie Chen


Jack Dongarra

University of Tennessee

Tarek El-Ghazawi

The George Washington University

Mootaz ElNozahy


Hoda ElSayed

Bowie State University

Hongua Guo

Institute of Chemistry, CAS

Warren He


Miaoquang Huang

University of Arkansas

Abdullah Kayi

The George Washington University

Vlad Kindratenko

National Center for Supercomputing Applications

Miron Livney

University of Wisconsin

Zhoyzhi Luan

Beihang University

John Michalakes

National Renewable Energy Laboratory

Greg Newby

Arctic Region Supercomputing Center

Stan Posey


Depei Qian

Beihang University

David Scott


Thomas Sterling

Louisana State University

H.Y. Wang

Institute of Mechanics, CAS

Zhe Wang


Frank Williams

Arctic Region Supercomputing Center

Xianghua Xie

Institute of Computing Technology, CAS

Guangwen Yang

Tsinghua University

Kunqian Yu

Shanghai Institute of Materia Medica

Yongqiag Yu

Institute of Atmospheric Physics

Xiaojun Zhang


Xingjun Zhang

Xi'an Jiaotong University

Back to Top