Xevolver CREST: An evolutionary approach to construction of a software development environment for massively-parallel heterogeneous systems


Nowadays, HPC system architectures are becoming larger and more heterogeneous. Thus future HPC systems will be more complex. Under this situation, there is a strong demand for abstraction techniques that can hide the complexity of system architectures. Moreover, it is also necessary for application developers to migrate quite a huge amount of software resource to the new abstracted environment even if the system complexity can be hidden by the abstraction techniques. Therefore, it is important to establish a smooth migration path to the environment as well.

In this project, we will explore a programming model and its development environment to support the migration. There are two approaches to design a programming model. One is an evolutionary approach that incrementally improves an existing model, and the other is a revolutionary approach that creates a completely new model. As the xevolver project focuses to support incremental migration of existing software resources, we adopt the evolutionary approach to establish new algorithms, implementation schemes, and development environments for massively-parallel heterogeneous systems.

Topics

Presentations at 4th International Workshop on Legacy HPC Application Migration
We will have these presentations at 4th International Workshop on Legacy HPC Application Migration(LHAM’16). Yasuharu Hayashi, Hiroyuki Takizawa and Hiroaki Kobayashi, “A User-Defined Code Transformation Approach to Overlapping MPI Communication with Computation,” The Fourth International Symposium on Computing and Networking, Hiroshima, Japan, November 22-25, 2016. (accepted) Reiji Suda and Hiroyuki Takizawa, “A software system supporting […] (11/15/2016)
A poster presentation at SC16
We will have a poster presentation at SC16. Hiroyuki Takizawa, Daichi Sato, Shoichi Hirasawa, and Hiroaki Kobayashi, “Making a Legacy Code Auto-tunable without Messing It Up,” ACM/IEEE Supercomputing Conference 2016 (SC16), 2016. (poster) (11/10/2016)
A Presentation at IEEE 10th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC-16)
We will have a presentation at IEEE 10th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC-16). Hang Cui, Shoichi Hirasawa, Hiroyuki Takizawa, Hiroaki Kobayashi, “A Code Selection Mechanism Using Deep Learning,” IEEE 10th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC-16), Lyon, France, September 21-23, 2016. (09/15/2016)
Journal Articles at International Journal of Networking and Computing
These journal articles have been published at International Journal of Networking and Computing. Kazuhiko Komatsu, Ryusuke Egawa, Shoichi Hirasawa, Hiroyuki Takizawa, Ken’ichi Itakura, Hiroaki Kobayashi, “Translation of Large-Scale Simulation Codes for an OpenACC Platform Using the Xevolver Framework,” International Journal on Networking and Computing (special issue on CANDAR’16), Vol. 6, No. 2, pp. 167-180 , […] (08/30/2016)
ICCSA 2016 conference papers have been published.
The following conference papers have been published in the proceedings of ICCSA 2016. Daisuke Takahashi, “Implementation of Multiple-Precision Floating-Point Arithmetic on Intel Xeon Phi Coprocessors,” Proc. 16th International Conference on Computational Science and Its Applications (ICCSA 2016), Part II, Lecture Notes in Computer Science, Vol. 9787, pp. 60–70, Springer International Publishing (2016). Hiroshi Maeda and […] (07/25/2016)