Trip Report: RIKEN R-CCS

By Weile Wei

Weile Wei, a graduate student in STE||AR group, was recently invited to participate RIKEN Center for Computational Science (R-CCS) 3rd Youth Workshop and 1st International Symposium in Kobe, Japan from February 15th to the 20th.

The 2-day Youth Workshop comprised around 30 students (approximately 20 Japanese students and 10 foreign students) all with strong High-performance Computing backgrounds. The students were divided into 4 groups and each group was paired with 2 mentors from RIKEN R-CCS. During the workshop, each student was required to individually present his/her research topic and, together with the team, collaboratively design an applicable research project that applies his/her research topics to the project solution.

Youth Workshop Group Photo
All Participants in RIKEN Center for Computational Science (R-CCS) 3rd Youth Workshop (Photo Credit to RIKEN Computational Science Promotion Office)

Weile Wei was placed in Group D and worked on a research project titled as “Particle-in-Cell System for Snowball Movement Simulation.” This project proposed an innovative PIC system which integrated Markov Chain techniques, high level parallelization, and a power-efficient strategy to handle dynamic particle movements which are verified by separation logic. Weile’s tiling strategy played an important role in the proposed system. His implemented a feature which dynamically assigned more CPUs to heavy computation areas of the particle simulation. In this way, the PIC system was able to more efficiently distribute tasks and accelerate overall computation time.

The 2-day symposium was focused on introducing the Post-K supercomputer – an ARM ecosystem which has more than 150,000 nodes. The symposium also included talks from world famous HPC specialists and panel discussions about the usage of the Post-K supercomputer.

Additionally, the symposium hosted a poster session that included around 70 HPC-related publications. Weile presented his research in a poster titled “Performance Analysis of Machine Learning Algorithms for Phylanx: An Asynchronous Array Processing Toolkit”. The poster describes Phylanx as an asynchronous array processing toolkit which transforms Python and Numpy operations into code which can be executed in parallel on HPC resources by mapping Python and NumPy functions and variables into a dependency tree executed by HPX, a general purpose, parallel, task-based runtime system written in C++. During the presentation, he shared early performance results of Phylanx on four machine learning algorithms, including Alternating Least Square, Logistic Regression algorithm, neural networks, and K-means.

Weile Wei and his poster
Weile Wei presenting his poster.
GD Star Rating
loading...
Be Sociable, Share!

Leave a Reply

Your email address will not be published. Required fields are marked *