Special Session: Auto-Tuning for Multicore and GPU (ATMG) In conjunction with the IEEE MCSoC-15 at Politecnico di Torino, Turin, Italy.

Computer architectures and computer systems are becoming increasingly complex. Multicore processors and heterogeneous computing systems equipped with accelerators pose demanding challenges to programmers when optimizing the performance of their codes. Auto-tuning is a promising technology for tackling these challenges. Several methodologies and auto-tuning approaches have been proposed, especially for numerical computations. More recently, the applicability of auto-tuning has been extended to other areas of computer science and engineering, from embedded systems to extreme scale computing systems.

This special session on "Auto-Tuning for Multicore and GPU (ATMG)" creates an opportunity for the discussion of the state-the-art in auto-tuning research, not only for experts in the field but also a broader audience of users of the technology.

Topics of Interest

papers are solicited in following areas of auto-tuning technologies for multicore and GPU/MIC computing, but not limited to:

  • Optimized Algorithms for Numerical Libraries
  • Automatic Code Generation and Empirical Compilation Hybrid Programming for Threads and Processes
  • Communication Optimization
  • Mixed Precision and Accuracy Assurance for Numerical Computing
  • Power Consumption Optimization
  • Fault Tolerance


Submission information including paper format and submission site is listed here.

  • Each submitted paper will be reviewed by at least three reviewers and will be evaluated based on originality, relevance to the session's theme, organization, and technical quality.
  • Submission of a paper implies that at least one of the authors will register for the symposium at the full rate and an author will present the paper.
  • The papers should be prepared according to the IEEE CS format, and no longer than 8 pages.
  • Proceedings of the symposium will be published by IEEE Computer Society Press.
  • Online submission site:

Session Program

Session date was fixed, 15:20-17:20 on September 24 at Room 9T. The venue and other session programs are listed here.

“CLTune: A Generic Auto-Tuner for OpenCL Kernels” Cedric Nugteren and Valeriu Codreanu (SURFsara HPC Centre, The Netherlands)
“Enhancement of Incremental Performance Parameter Estimation on ppOpen-AT” Riku Murata, Jun Irie, Akihiro Fujii and Teruo Tanaka (Kogakuin University, Japan); Takahiro Katagiri (University of Tokyo & Information Technology Center, Japan)
“Improving Auto-Tuning Convergence Times with Dynamically Generated Predictive Performance Models” James Price and Simon McIntosh-Smith (University of Bristol, United Kingdom)
“The Approximate Discrete Radon Transform: A Case Study in Auto-Tuning of OpenCL Implementations” Tobias Beier (Carl Zeiss Jena GmbH, Germany); Ralf Seidler, David Neuhäuser and Hanns Martin Bücker (Friedrich Schiller University Jena, Germany)

Important Date

Paper submission due: April 7, 2015
April 30, 2015
May 15, 2015
Author notification: May 31, 2015
June 22, 2015
June 25, 2015
Camera ready manuscript:June 30, 2015
Paper submission due:April 7, 2015
Session Date: September 23, 2015


Program Chair:
Akihiro Fujii (Kogakuin University, Japan)

Program Vice-Chair:
Tony Drummond (Lawrence Berkeley National Laboratory, USA)

Program Committee:
Christophe Calvin (CEA, France)
Christos Kartsaklis(Oak Ridge National Laboratory, USA)
Franz Franchetti (Carnegie Mellon University, USA)
Domingo Gimenez (University of Murcia, Spain)
Toshiyuki Imamura (RIKEN, Japan)
Takahiro Katagiri (The University of Tokyo, Japan)
Hisayasu Kuroda (Ehime University, Japan)
Osni Marques (Lawrence Berkeley National Laboratory, USA)
Naoya Maruyama (RIKEN, Japan)
Satoshi Ohshima (The University of Tokyo, Japan)
Serge Petiton (University of Lille 1, Sciences and Technologies, France)
Reiji Suda (The University of Tokyo, Japan)
Daisuke Takahashi (University of Tsukuba, Japan)
Hiroyuki Takizawa (Tohoku University, Japan)
Weichung Wang (National Taiwan University, Taiwan)
Yusaku Yamamoto (The University of Electro-Communications, Japan)

return to top