Emerging Trends in HPC Systems and Application Modernization
Authors: Mr. Hans-Christian Hoppe (Intel Corporation)
Abstract: This BOF discusses the application evolution required by clearly apparent HPC system trends: complex memory hierarchies with fast limited-size memory, addition of large capacity non-volatile memory, substantial increase in cores, and integration of high-performance interconnects. Applications need to reshape to achieve good performance, taking advantage of these architecture trends, while limiting complexity to allow efficient code development and maintenance.
Developers of significant HPC applications will discuss their approaches and experiences in code evolution, preparing a discussion with the audience. This BOF aims to foster a global community of practitioners across funding organizations and schemes and combining academic and industrial participants.
Long Description: This BoF takes up the topic of co-design between HPC systems and applications, focusing on the evolution of application codes required by the now clearly apparent lines of development in general-purpose HPC systems: the emergence of more complex memory hierarchies incorporating fast yet limited memory per core plus the addition of large capacity non-volatile memory, the substantial increase in cores per shared-memory "island", and the closer and closer integration of high-performance interconnects with the CPU and memory subsystem. Application codes that want to take advantage of future general-purpose HPC systems will need to reshape to achieve good levels of performance (both in terms of runtime and energy). At the same time, it is important to ensure that codes ca be maintained and developed further without undue complexity imposed by the execution systems.
Developers of significant HPC applications from scientific and industrial contexts with plans to actually use Exascale-class systems will discuss their approaches and experiences so far in evolving their codes, triggering a hopefully lively discussion with the audience. An important outcome of this BoF should be to foster a global community of practitioners that transcends the boundaries of funding organizations and schemes and combines both academic and industrial participants.
At the time of submission, speakers have been identified from the technical University of Munich (Prof. Bader on behalf of the SEISSOL code and the EXAHYPE colaborative R&D project) and T-Systems (Christian Simmendinger on behalf of next-generation CFD codes and the EXA2CT collaborative R&D project), plus a representative of Intel (on behalf of lessons learnt in the Intel Parallel Computing Centers). We will have two additional speakers from the US side, and one presenter from Japan.
Conference Presentation: pdf
Birds of a Feather Index