HiPC Academic Birds of a Feather (BoF)

HiPC introduced a new session format in 2014, referred to as Academic BoF. The purpose is to help foster greater participation by the academic community at HiPC, including both domestic and international representation.

Academic BoFs focus on emerging areas of interest, where discussion among HiPC participants will help stimulate new research ideas, address specific problems, and build a research community. The BoF session format is flexible and left to the decision of the organizers, though sessions that actively engage with the participants are encouraged.

Academic BoFs will complement the main conference technical program and workshops that have peer-reviewed papers. It is expected that successful academic BoFs will advance to full workshops in subsequent years.

As of this posting, two Academic Birds-of-a-Feather (BoF) sessions have been organized for HiPC 2015. All HiPC 2015 registered attendees are invited and encouraged to attend any and all of the BoF sessions.

Academic Birds-of-a-Feather 1
Thursday, December 17, 2015
4:00 PM - 6:00 PM

Compilation Research using LLVM

This is an academic "birds of a feather" event on compilation research using the LLVM infrastructure. The Low Level Virtual Machine (LLVM) compiler is increasingly being used as the vehicle to develop compiler analysis and optimization infrastructure both in the industry and the academia. Several research groups in the India academia have also started using LLVM, and some are actively contributing to it. The objective is this BoF is to disseminate information on existing efforts using LLVM in India, make the community aware of existing opportunities for research as well as software development, share experience and expertise in using LLVM for academic research, and instigate collaboration between groups. One of this BoF's goals is also to evolve into a full-fledged workshop in the next three years. For more details and program schedule, please visit the ABoF website.

Organizers and Program Chairs

Uday Reddy B, Indian Institute of Science ([email protected])
Dibyendu Das, AMD

Academic Birds-of-a-Feather 2
Saturday, December 19, 2015
10:00 AM - 12:00 PM

Advanced Numerical Schemes for Massively Parallel Computing: Challenges and Opportunities

Many physical phenomena and industrial processes are described by a set of partial differential equations (PDEs), and almost all of these PDE models are coupled and nonlinear in nature. Most popular methods for solving PDEs are the finite difference method, finite volume method and finite element method. Applying one of these methods to model problems results in a large sparse system of (mostly linear) algebraic equations. In general, the size of the system will be few hundred millions, and solving these huge systems is impossible without supercomputers. Apart from other challenges associated with the parallel implementations, parallel computations require not only efficient parallel algorithms, but also highly scalable numerical methods. Moreover, the supercomputing architecture is moving toward massive parallelism with millions of cores and use of hardware accelerators. This requires a fundamental change in the methods, algorithms and implementations to exploit full compute power from these architectures. The development of robust and efficient high order algorithms and their use in HPC systems is an active research area.
It is the purpose of this session to discuss the challenges and opportunities in developing parallel, efficient and scalable numerical schemes for massively parallel supercomputers. The topics of the discussion are mentioned below. In addition, research areas that are related to these topics will also be considered.

  • Scalable and robust high order schemes
  • High performance numerical linear algebra
  • Efficient implementations
  • Scalable time integration strategies
  • Algorithms for GPUs, heterogeneous systems
The schedule and program for this ABoF will be posted soon.

Organizers and Program Chairs

1. Sashikumaar Ganesan, IISc, Bangalore ([email protected])
2. Praveen C, TIFR-CAM , Bangalore
3. Biswajit Mishra, Shell Technology Center, Bangalore