Tampere University

Design Methodologies for Efficient Stream Processing


The Design Methodologies for Efficient Stream Processing research team belongs to the Tampere University Unit of Computing Sciences.

Research Themes

The team studies methodologies and tools for realizing efficient processing of applications on various computing platforms including multicore processors and GPUs in both embedded and desktop environments. The aim of our research is to promote parallel processing and code portability while maintaining high performance in terms of throughput and/or energy-efficiency. In order to achieve these aims, our research efforts rely on graph-based concurrent programming abstractions such as Kahn process networks and dataflow.

The application areas related to our research consist mostly of real-time applications from machine learning, wireless communications and image processing domains.

Active Research Projects

Compact and Efficient Deep Neural Networks for Ubiquitous Computer Vision (CoEfNet) [9/2017-8/2021] (Academy of Finland) Computer vision has progressed in many of its fields with the adoption of deep neural networks. Automatic image classification, segmentation and captioning, as well as object recognition and human pose estimation have been especially suitable application areas for deep neural networks. Besides academia, also the industry has expressed considerable interest towards neural network-based solutions, which can be seen in the form of company acquisitions and research investments by car manufacturers. The large-scale deployment of deep neural network -based solutions has however been hindered by computational challenges, which limit their portability to mobile platforms and vehicles. To this extent, the aim of this project is to develop new methodologies for constructing more efficient deep neural networks. The challenge is addressed from two aspects: first, by optimizing existing deep neural networks, and second, by developing efficient neural network architectures from scratch. The partners of the collaborative effort are Aalto University and Tampere University of Technology, as well as international collaborators from Grenoble, Oxford and University of Maryland.

Cost-Efficient Smart System Software Synthesis (COMPACT) [9/2017-8/2020] (ITEA / Business Finland) Due to the very limited resources provided by Internet-of-Things (IoT) nodes, today’s commonly used design approach to trade off development time with software efficiency is not competitive any longer. Therefore, an industry-wide effort is needed to provide measures for fast and efficient IoT software development. The main goal of the COMPACT project is to provide novel solutions for the application-specific and customer-oriented realisation of ultra-small IoT nodes with a focus on software generation for IoT nodes with ultra-small memory footprints and ultralow power consumption.

Past Research Projects

Unified platform-compiler design for embedded streaming applications (UNICODE) [1/2015-4/2018] This project is funded by the Academy of Finland and aims for significant advances in dataflow-based programming and processing. In particular, the project seeks to find efficient ways for programming GPUs and multicore TTA processors starting from dataflow-based program abstractions. Secondarily, the project aims to study dataflow-flavored multiprocessor interconnects.

Select Publications

Some representative publications co-authored by the team members:

  • The PRUNE dataflow environment for Linux-based multicore+GPU platforms can be found in GitLab. Also, a brief introduction is available here.
  • Binarized convolutional neural network implementation for vehicle classification in CUDA


Feb 1, 2019
Paper accepted to ICASSP.

April 15, 2018
Jiahao Wu from University of Maryland visits the team.

March 2, 2018
Alexandre Meirhaeghe from INSA Rennes starts his research visit.