Chair News

Published on

With three presentations and two best paper nominations, the chair for compiler construction was well represented at this year's virtual DATE conference. Asif Ali Khan presented his work on Generalized Data Placement Strategies for Racetrack Memories . In this paper, Asif presents an efficient data placement heuristic for racetrack memories (RTM) which leads to a significant reduction in RTM shifts and ultimately improves performance and energy consumption.  Christian Menard presented the paper Achieving Determinism in Adaptive AUTOSAR which was nominated for best paper award in the A-Track. In this paper, Christian describes how a novel synchronous reactive programming model can be used to make the execution of future automotive software deterministic and thus overall more reliable. Robert Khasanov presented his work on Energy-efficient Runtime Resource Management for Adaptable Multi-application Mapping which was best paper award candidate in the E-Track. In this paper, Robert presents a new job scheduling algorithm for firm real-time multi-threaded applications, which generates mapping segments due to enlarged scope of analysis, and increase the overall energy-efficiency of the schedules. 

Published on

We welcome a new research assistant at the Chair for Compiler Construction: Alex Brauckmann. Alex just received his Diploma degree on Computer Science from the TU Dresden. He has plenty of experience as software engineer and in large open source projects. Alex has worked at the CC chair as student research assistant and did his final thesis with us, resulting in a high-profile international publication at the  ACM SIGPLAN International Conference on Compiler Construction (CC). At the CC Chair, Alex will continue to work on machine learning for compilers, building representation models of code and proposing novel predictive compiler heuristics among others. We are happy to have Alex with us and look forward to working with him in the next years. 

Published on

Alex successfully presented his work on “Compiler-Based Graph Representations for Deep Learning Models of Code” at this year’s ACM SIGPLAN International Conference on Compiler Construction (CC). CC is a excellent venue for presenting compiler research that became an ACM SIGPLAN conference for the first time this year. The work shows a novel deep learning method that improves predictive compiler heuristics. Alex has been working at the CC chair on applications of machine learning to compilers in the context of his final thesis. We are glad to count with Alex to continue this exciting line of research with us in the context of his doctorate. 

Published on

We welcome Julian Robledo to Chair for Compiler Construction. Julian received his Bachelor degree in Electronic Engineering Computer Engineering (2015) from University of Antioquia (UdeA) in Medellín Colombia and his Master’s degree in Embedded Systems from the Polytechnic University of Turin in 2017 in Italy. After finishing his master’s program, Julian work in the automotive sector. At the CC Chair, Julian will work on a collaborative project on developing programming models and tools for future 5G communication standards and beyond. He will also contribute with his expertise in the automotive sector to running projects revolving around adaptive automotive applications (see here for instance). We are very glad to finally welcome Julian, after experiencing bureaucratic problems, and are looking forward to working with him.

Published on

The Chair for Compiler Construction was well represented at this year's HiPEAC Conference in Bologna, Italy. The HiPEAC conference is a great networking oportunity and the premier European forum for experts in computer architecture, programming models, compilers and operating systems for embedded and general-purpose systems. Asif Ali Khan presented his ACM TACO paper ShiftsReduce: Minimizing Shifts in Racetrack Memory 4.0 in the main track. Christian Menard presented his work on Achieving Determinism in Adaptive AUTOSAR in the poster session on Monday and in the co-located workshop on "Energy Efficiency for Reactive and Adaptive Automotive Applications". This workshop was organized in a joint effort by the Chair for Compiler Construction together with Silexica, ICE RWTH Aachen, and Videantis. Both researchers participated in the social event and many other conference sessions where they used the opportunity to exchange idead with like-minded researchers from around the world.

Published on

Pirah Noor Soomro is visiting the Chair for Compiler Construction from January through March 2020. Her visit is supported by a Eurolab4HPC Grant that aims at improving the efficiency of tensor applications on heterogeneous platforms. A specific area of focus is the processing of tensors during the inference phase of state-of-the-art machine learning models, including deep neural networks, on heterogeneous multi-core CPUs. Pirah’s work at the Chair for Compiler Construction will build on the XiTAO runtime, which Pirah has previously been using in her work at her home institution, Chalmers University of Technology.

Published on

The chair for compiler construction, together with Prof. Chadlia Jerad from the University of Manouba (Tunis), organized a lecture series and a workshop on "Embedded programming in practice: teaching and research", held at the National School of Computer science (ENSI). The event took place during the week of the 1st of December 2019. Prof. Castrillon, assisted by his Ph.D. Students Andres Goens and Hasna Bouraoui, gave an intensive course for Tunisian students, where he presented fundamentals of parallelism, polyhedral compilers, and data-flow based programming models. Besides, he gave a research talk entitled "Embedded manycore programming: From auto-parallelization to domain-specific languages" during the workshop session. The research talk ended with a round table discussion about Research landscape and opportunities for technology while debating two viewpoints from two different countries: Germany and Tunisia.  The event was sponsored by the  Daad Project Transformation line 2, under a collaboration between TU Dresden and the University of Manouba, Tunis. 

Published on

We are happy to host the first of hopefully many exchange PhD students from the Universidad de Antioquia (UdeA) in Medellin Colombia. John Byron works under the supervision of Prof. Fredy River and Prof. Ricardo Velasquez at the Department of Electronics where he is affiliated to the SISTEMIC lab. John Byron Buitrago received his BSc. diploma in Electronics Engineering in 2005 and his Master Degree on Engineering in 2014 from the UdeA. He started his PhD in the same institution in 2016, working on methodologies for the automatic synthesis of computer architecture simulators onto parallel platforms. John Byron will be collaborating with researches at the CC Chair for the next six months on methodologies to automatically generate efficient simulators with special focus on the uncore. We look forward for a collaborative winter term!

Published on

The CC Chair was present at the International Conference on Software Language Engineering (SLE) co-located with SPLASH and OOPSLA. Lars Schütze presented our work on Efficient Late Binding of Dynamic Function Composition. The conference, known for its applications of modeling, domain-specific languages (DSLs), and their implementation, was a good fit for the presented topic. The conference offered a nice atmosphere and space for vivid discussions and networking with researchers across the globe. The talk was attended by roughly 30 researchers. A video feed of the presentation is also available. 

Published on

Prof. Castrillon gave the first keynote in the 13th IEEE International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC’19), held in Singapore during  October 1 to 4. Prof. Castrillon gave similar a talk to the one in the CPSS summer school and talked about “Embedded manycore programming: From auto-parallelization to domain specific languages”. The presentation reviews work done by Prof. Castrillon and collaborators over the past 13 years for programming multi-core systems, going from auto-parallelization, over dataflow programming to domain specific languages. The presentation was attended by about 70 researchers working on different aspects of multi-core systems, from architectures to end-user applications.