Guillaume Aupy earnt his PhD from Ecole Normale Supérieure de Lyon in 2014. After a post-doctoral position at Argonne National Lab, he has been a research assistant professor at Penn State and Vanderbilt universities. He now joins TADaaM as a permanent Inria research scientist.
Juan Luis García Zapata from the University of Extremadura will present his work on “Spectral graph partition: old and new results”
Abstract: A distributed application is modeled as a weighted graph where tasks or processes are the vertices and the edge between two task is weight by the total volume of communications between them. The problem of graph bi-partitioning is to find, among all possible partitions of the set of tasks into two equal parts, that partition that minimizes communication of one of the two parts to the other. This is of practical interest in mapping tasks to processors of a particular architecture, to take advantage of the
communication channels on the machine.
We’ll talk about the spectral solution to the problem of bipartition. It is based on certain eigenvector of the Laplacian matrix of the graph. The presentation will be based on visual and intuitive examples. We also discuss modifications in this classical solution, to incorporate into the graph model a computational load of tasks, perhaps uneven. Also to do the partition into two parts of predefined volume (for example 1/3 and 2/3 of the total load), for mapping to heterogeneous architectures.
Guillaume Aupy (Vanderbilt Univ.) will give a talk on Optimal Multi-stage algorithm for Adjoint Computation.
Abstract: We reexamine the work of Stumm and Walther on multistage algorithms for adjoint
computation. We provide an optimal algorithm for this problem whenthere are two levels of check- points, in memory and on disk. Previously, optimal algorithms for adjoint computations were known only for a single level of checkpoints with no writing and rea ding costs; a well-known example is the binomial
checkpointing algorithm of Griewank and Walth er. Stumm and Walther extended that binomial checkpointing algorithm to the case of two levels of checkpoints, but they did not provide any optimality results. We bridge the gap by designing the first optimal algorithm in this context. We experimentally
compare our optimal algorithm with that o f Stumm and Walther to assess the difference in performance.
Balasz Gerofi from RIKEN will give a talk.Title: An Overview of RIKEN’s System Software Research and Development Targeting the Next Generation Japanese Flagship Supercomputer
RIKEN Advanced Institute for Computation Science has been appointed by the Japanese government as the main organization for leading the development of Japan’s next generation flagship supercomputer, the successor of the K. Part of this effort is to design and develop a system software stack that suits the needs of future extreme scale computing. In this talk, we first provide a high-level overview of RIKEN’s system software stack effort covering various topics, including operating systems, I/O and networking. We then narrow the focus on OS research and describe IHK/McKernel, our hybrid operating system framework. IHK/McKernel runs Linux with a light-weight kernel side-by-side on compute nodes with the primary motivation of providing scalable, consistent performance for large scale HPC simulations, but at the same time, to retain a fully Linux compatible execution environment. We detail the organization of the stack and provide preliminary performance results.
Balazs Gerofi is a research scientist at RIKEN Advanced Institute for Computational Science, where he is primarily involved with system software development for high performance computing. He earned his M.Sc. degree and Ph.D. degree in computer science from the Vrije Universiteit Amsterdam and The University of Tokyo, respectively. His research is mainly focused on operating systems, high-performance computing, cloud-computing, and fault tolerant computing. Balazs is a member of the IEEE Computer Society and the Association for Computing Machinery (ACM).
Nicolas Denoyelle begins his PhD thesis co-supervised by the Atos/Bull company.
François Tessier successfully defended his PhD thesis in Bordeaux on January 26th 2015.