Report on the 26th IEEE International Conference on High Performance Computing, Data, and Analytics
(HiPC, December 16—18, 2020, Virtual)
by
Kishore Kothapalli (IIIT Hyderabad) and
Chiranjib Sur (Shell, Bangalore)
The 26th IEEE International Conference on High Performance Computing, Data, and Analytics (HiPC) was held in a fully virtual manner during December 16 to December 19, 2020. The conference committee decided in June 2020 to move the program of 2020 to a fully virtual year to the ongoing Covid-19 pandemic. This early decision allowed the conference committee to plan the virtual format carefully. Given this decision, and with a view to promote the conference across India in particular and the rest of the world in general, the conference did not charge any registration fee for attending the virtual sessions. Authors of contributed papers had to pay a nominal $100 fee that covered the cost of publication in IEEE Xplore. The conference attracted over 500 registrations. The entire program was run using Zoom webinars.
The conference had three excellent keynote talks. The talk by Prof. Yelick (UC Berkeley) titled Computing and Data Challenges in Climate Change discussed the various computing challenges that are to be solved in the context of problems arising from climate change. The talk by Prof. Anandkumar (California Institute of Technology, and Nvidia Research) titled Role of HPC in next-generation AI spoke on how machine learning workflows are evolving on emerging platforms and the role of tensors in these frameworks. The talk by Dr. Petrini (Intel) titled Breaking the Scalability Wall discussed the scalability issue and the need for low latency communications and some of the recent work at Intel in this direction.
The program committee accepted 34 papers out of the 170 submissions received and these are arranged across five sessions in addition to a session on best papers. A session of 60 minutes is set aside for presentation of the two papers that were voted to be the best papers. These papers are SimGQ: Simultaneously Evaluating Iterative Graph Queries, authored by Chengshuo Xu, Abbas Mazloumi, Xiaolin Jiang and Rajiv Gupta, and WarpCore: A Library for fast Hash Tables on GPUs, authored by Daniel Jünger, Robin Kobus, André Müller, Kai Xu, Weiguo Liu, Christian Hundt and Bertil Schmidt. Each other session was for a duration of 60 minutes which included a five-minute live presentation by authors followed by five minutes of discussion. Authors of each paper also uploaded a 20-minute presentation to the IEEE website.
The virtual content, which included talks, slides, and Q&A discussion is still available on Facebook at https://www.facebook.com/hipc.conference
The pandemic situation that significantly cut down global travel, among other things, meant that the entire conference was run in a shorter duration than usual. As registration for attending the live conference sessions was free, the conference did attract increased registration numbers. The year 2020 is likely to introduce several changes in how conferences will be run in the coming years.