POAT2026

Special Session:
Performance Optimization and Auto-Tuning of Software on Multicore/Manycore Systems (POAT)

In conjunction with IEEE MCSoC-2026 (Dec 14-17, 2026, Shanghai Jiao Tong University (Xuhui Campus), Shanghai, China)

Multicore and manycore processors (CPUs, GPUs, AI accelerators, etc.) have become commonplace today and are widely used in systems ranging from smartphones to supercomputers. Their architectures and configurations are becoming increasingly diverse and complex, incorporating heterogeneous components and specialized accelerators. As a result, performance optimization of programs on these state-of-the-art processors has become an increasingly significant challenge.
This special session in MCSoC covers the general topic of performance optimization on multicore and manycore systems. “Performance” here includes not only computation speed, but also power and energy efficiency, resource utilization, and the quality of computation (e.g., numerical accuracy). The session welcomes contributions on techniques for efficiently exploiting parallelism and architectural features of modern processors across a wide range of applications, including scientific computing, data-intensive workloads, and emerging AI/ML workloads such as large-scale model training and inference.
Furthermore, this session aims to discuss auto-tuning and automated optimization techniques as important approaches for dealing with the complexity and diversity of multicore and manycore systems.
POAT is an extended successor to the special session formerly known as Auto-Tuning for Multicore and GPU (ATMG) (2012-2022).

Topics of Interest

Topics of interest include, but are not limited to:

  • Performance optimization of software on manycores: for computational speed, energy efficiency, accuracy, etc.
  • Quantum computing using manycores (quantum circuits simulators, quantum inspired annealing, etc.)
  • Mixed-precision computation and precision-tuning on manycores systems
  • Thread-parallelization / SIMD-vectorization on manycores systems
  • Auto-tuning and automatic code generation for manycores systems
  • Application of AI-accelerator (or AI-oriented extensions on processors) to general-purpose computations
  • Code porting and benchmarking between different manycore systems
  • Tools / programming models / languages for performance on manycore systems
  • Challenges for heterogeneity on manycore systems
  • Manycores on FPGAs
  • High-performance computing for AI4Science and data-driven scientific discovery
  • Efficient training and inference of large-scale AI/LLM models on manycore architectures
  • Performance optimization and parallelization of AI workloads on HPC systems
  • Integration of AI methods with traditional HPC simulations
  • Large-scale scientific data analysis using AI on manycore platforms

Submission

All contributions should be original, not published elsewhere, or intended to be published during the review period. Electronic Paper submission should not exceed 8 double-column IEEE format pages, including figures and references. Papers should be formatted as close as possible to the final format double-column, single-spaced, and Times or equivalent font of minimum, 10pt size. The acceptable format is PDF only. Any other file format and manuscripts received in hard-copy form will not be processed. Please use IEEE Formats for paper formatting.

Download the IEEE manuscript Templates (.doc OR Latex) for Conference Proceedings https://www.ieee.org/conferences/publishing/templates.html

Every submitted paper will be evaluated by at least three reviewers (Single-blind peer review) in technical quality, originality, significance to the symposium, and organization. We encouraged authors to proofread and spell-check their papers before submission. The 17th IEEE MCSoC 2026 proceedings will be published by the IEEE Computer Society, which will be included in the Computer Society Digital Library CSDL and IEEE Xplore. All CPS conference publications are also submitted for indexing to EI’s Engineering Information Index, Compendex, ISI Thomson’s Scientific and Technical Proceedings, ISTP/ISI Proceedings, and ISI Thomson. Note: MCSoC does not accept abstract-only submissions.

Avoiding Plagiarism and Self-Plagiarism:

Papers should not have plagiarized material and should not have been submitted to any other conference simultaneously (double submission).

Please check these links to guidelines on ethical writing:

Preprent Papers (e.g. arXiv):

We conform to the IEEE Preprint Policy. Because preprint papers are not peer-reviewed, we don't consider them as previous works. As long as you follow the policy, we don't prohibit you from submitting your papers that have preprints to POAT2026.

Submission System

Submit your paper online here: https://edas.info/N34632

Important Dates (AoE)

  • Full Paper Submission: Apr, 30, 2026
  • Acceptance notification: July 20, 2026
  • Camera-ready paper: July 31, 2026
  • Conference date: December 14-17, 2026

Program

TBD

Previous Events

  • POAT2025, Singapore, December 15-18, 2025.
  • POAT2024, Malaysia, December 16-19, 2024.
  • POAT2023, Singapore, December 18-21, 2023.
  • ATMG2022, Penang, Malaysia, December 19-22, 2022.
  • ATMG2021, Singapore, December 19-22, 2021.
  • ATMG2020 (Canceled due to COVID-19 travel restrictions)
  • ATMG2019, Singapore, October 1-4, 2019.
  • ATMG2018, Hanoi, Vietnam, September 12-14, 2018.
  • ATMG2017, Seoul, Korea, September 18-20, 2017.
  • ATMG2016, Lyon, France, September 21-23, 2016.
  • ATMG2015, Turin, Italy, September 23-25, 2015.
  • ATMG2014, Aizu-Wakamatsu, Japan, September 23-25, 2014.
  • ATMG2013, Tokyo, Japan, September 26-28, 2013.
  • ATMG2012, Aizu-Wakamatsu, Japan, September 20-22, 2012.

Committee

Program Chair:
Masatoshi Kawai (Tohoku University, Japan)

Program Vice-Chair:
Bing-Ze Lu (National Chung Cheng University, Taiwan)

Program Committee: (Tentative)
Beau Johnston (Oak Ridge National Laboratory, USA)
David Defour (Université de Perpignan, France)
Kazuhiko Komatsu (Tohoku University, Japan)
Masahiro Nakao (RIKEN R-CCS, Japan)
Satoshi Ohshima (Kyushu University, Japan)
Takeshi Fukaya (Hokkaido University, Japan)

Contact Person:
Masatoshi Kawai (Cyberscience Center, Tohoku University, Japan)
E-mail: m.kawai (at) tohoku.ac.jp (Please replace "at" by @.)