Chair News

Published on

We congratulate Christian Menard for having successfully defended his PhD on April 25th, 2024 on "Deterministic Reactive Programming for Cyber-physical Systems”. Christian spent several fruitful years with us, starting with his Diploma Thesis for which he received the Hermann-Willkomm in 2016. He worked on several programming models for parallel and distributed computing, leading to upwards of 20 international publications. In the past 4 years he led the development of the Lingua Franca project on the side of TU Dresden. A successful collaboration with the UC Berkeley and several other academic partners worldwide. Special thanks to Prof. Stephen Edwards from Columbia University for acting as external reviewer. Christian will move on to lead Xronos, a company that develops tools and services for building software-defined cyber-physical systems that span deeply embedded, edge, and cloud platforms. We wish Christian all the best and hope to keep in touch! 

Published on

As has become a tradition, the CC chair participated in the REWE Team Challenge, a 5K run through downtown Dresden. This year, our chair was represented by three mixed teams: Com(e)-pile-run, consisting Clément Fournier, Conny Okuma, Julian Robledo Mejia, and Jerónimo Castrillón; Byte me if you can, featuring Christian Menard, João Paulo Cardoso de Lima, Nesrine Khouzami, and Robert Khasanov; and Caffeine Circuits, a joint CC-PD team with Maryam Eslami, Siddharth Gupta, Steffen Märcker, and Tassilo Tanneberger. This year, we started our team training sessions earlier than in the previous years. Thanks to these sessions, many of us significantly improved our personal records! We are already looking forward to the next year's competition!

Published on

We welcome Guilherme (again) to the CCC team! Guilherme had spent 8 months with us in 2022-2023 prior to finishing his PhD in Computer Science at the Federal University of Rio Grande do Sul (UFRGS) in Brazil under the supervision of Prof. Dr. Antonio Carlos Beck. Guilherme’s research has focused on accelerating machine learning applications on FPGAs in edge computing. At the CC chair, Guilherme will continue to research methods to optimise applications in the computing continuum. We are extremely happy to have Guilherme back in the team! 

Published on

Felix Suchert represented the CC Chair at the 32nd International Symposium for Field-Programmable Custom Computing Machines, held from 5. to 8. May in Orlando, Florida. Together with Stephanie Soldavini from Politecnico di Milano, he presented their joint work on "Etna: MLIR-Based System-Level Design and Optimization for Transparent Application Execution on CPU-FPGA Nodes". It details a programming framework that allows a more easy and efficient integration of offloaded compute kernels into a dataflow-driven application. The work was presented both as poster and as a demo during demo night. Both presentations attracted lots of interested researchers and industry representatives and yielded productive and fruitful talks. Etna is part of a larger effort to ease the development of heterogeneous applications, which is driven by the EVEREST project. As such, it was also presented in this context in more detail during the EVEREST tutorial on day one of the conference, together with Prof. Christian Pilato from Politecnico di Milano. Felix also presented the efforts at CCC to expand the MLIR ecosystem as part of this tutorial session.

Published on

Prof. Castrillon and Asif Ali Khan represented the CC Chair at the prestigious International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) 2024, held in San Diego, USA from April 27th till May 1st, 2024. Asif presented our research article titled "C4CAM: A compilation framework for CAM-based accelerators". C4CAM is an MLIR-based high-level compilation framework for content-addressable memories (CAMs) that makes CAM devices accessible to non-experts, by bridging the abstraction gap between high-level programming frameworks and low-level device API calls supported by CAM devices. This framework is developed in close collaboration with CAM device experts at the University of Notre Dame, namely Prof. X. Sharon Hu and her research group.

The conference was attended by over 800 researchers and experts from both academia and industry. The beautiful weather in La Jolla and the vibrant ambiance at San Diego's USS Midway Museum provided a perfect environment for networking, exchanging thoughts and ideas, and engaging in discussions with fellow experts in the field.

Published on

We welcome Anderson to the CCC team! Anderson is a professor at the Department of Informatics of the State University of Maringá - Paraná - Brazil. Prior to his faculty position, Anderson received his bachelors, masters and PhD degrees from from Federal University of Rio de Janeiro. His research background and interests are on parallel programming, programming languages, computer architecture, code generation and code optimization. Recently he has worked on leveraging machine learning for compiler analysis and optimization. It is in this area that Anderson will enrich the CCC team in the coming years. We look forward to joint research work with Anderson!

Published on

The CC Chair attended the Design, Automation, and Test in Europe Conference (DATE'24) in Valencia, Spain from March 25-27, 2024. Prof. Castrillon delivered a talk on "Automatic optimization for heterogeneous in-memory computing" in the Focus Session "Smoothing Disruption Across The Stack: Tales Of Memory, Heterogeneity, And Compilers", where he presented the latest results of our work on compilers for emerging near and in-memory computing. This is the result of the fruit of collaborations with experts in emerging memory technologies and system architecture, including X. Sharon Hu and Mehdi Tahoori, among others. João Paulo de Lima presented the paper titled "Full-Stack Optimization for CAM-Only DNN Inference", a novel hardware/software co-optimization for ternary-weight neural networks. In this paper, we introduce an in-memory architecture based on racetrack memory and a compilation flow that capitalizes on redundant additions/subtractions to improve convolution performance. Apart from this, Prof. Castrillon participated in the multi-partner project session where the latest results of the EVEREST project were presented (see paper for details).

Valencia was absolutely fantastic for enabling lively and fruitful discussions with the DATE community!

Jeronimo at DATE'24 Joao at DATE'24

Published on

We would like to welcome Jiahong Bi to the CC Chair! Jiahong obtained his master’s degree on computer science from the TU Dresden in 2023 and his bachelor’s degree on engineering from eh Xidian University in China in 2019. Jiahong worked at the CC Chair as a student researcher, contributing for instance to the base2 MLIR dialect for custom representations, and in the context of his master thesis. In his thesis, entitled “A Lowering for High-Level Data Flows to Reconfigurable Hardware”, Jiahong worked on a novel dataflow MLIR dialect and a lowering path via high-level synthesis using the CIRCT project for FPGAs. He was supervised by Karl A. Friebel and Felix Suchert. With his expertise in high-level representations in compilers and tool-flows for heterogeneous systems, Jiahong will help our team in several projects that build around MLIR, most prominently on the MYRTUS EU Project. We are very happy to have Jiahong with us and are looking forward to working with him as PhD student. 

Published on

The CC chair was present at the 29th Asia and South Pacific Design Automation Conference (ASP-DAC 2024), held in Incheon, South Korea during January 22 to 25, 2024. The conference targets emerging topics related to system design for embedded computing and Electronic Design Automation (EDA). At ASP-DAC 2024, Robert Khasanov presented his work on "Flexible Spatio-Temporal Energy-Efficient Runtime Management". This paper researches the impact of mapping decision models on performance and energy efficiency in (firm) real-time systems equipped with heterogeneous CPUs. It proposes offline and online algorithms that fully exploit the spatio-temporal scheduling solution space, a key factor for adaptive energy-efficient computing.

Published on

This year’s HiPEAC conference took place in January 17-19. As usual, it was great for the CC Chair to visit the conference, reinforce collaborations and kick-start new ones. The CC chair co-organized the workshop EVEREST + DAPHNE workshop, on design and programming high-performance, distributed, reconfigurable and heterogeneous platforms for extreme-scale analytics. The workshop included excellent talks from Stephen Neuendorffer (AMD) and Torsten Hoefler (ETH Zurich). Karl gave a talk on “MLIR stack for heterogeneous systems: Experiences and Results”, detailing the compiler work done by the CC Chair in the EVEREST project. Ciao Vieira, a visiting PhD student from Universidade Federal do Rio Grande do Sul, presented his work on “Hyperdimensional Computing Quantization with Thermometer Codes" in the co-located workshop Accelerated Machine Learning (AccML).