Path Activities

Orchestration Retreat 2017

Published on in ORCHESTRATION (PATH ACTIVITIES)

The PhD students, postdocs and investigators of the Orchestration Path met from 3rd to 5th July for the annual retreat in Altenberg on the ridge of the Eastern Ore Mountains (Osterzgebirge). Most of the time was dedicated to intense working sessions about collaborative projects and future research ideas. Additionally, the invited talk by Diana Göhringer, the new professor for adaptive dynamic systems at the faculty of computer science, and the participation of Steffen Lange and Jens Karschau two postdocs from the Biological Systems Path have been used to discuss and initiate future collaborations. We also enjoyed the nice weather on a short hiking tour to Kahleberg.

"Wildly Heterogeneous Post-CMOS Technologies Meet Software" at Schloss Dagstuhl

Published on in ORCHESTRATION (PATH ACTIVITIES)

From February 5 to 10, 2017 the Dagstuhl seminar 17061 „Wildly Heterogeneous Post-CMOS Technologies Meet Software“ took place, organized by Jeronimo Castrillon (TU Dresden, leader of the Orchestration path), Tei-Wei Kuo (National Taiwan University), Heike E. Riel (IBM Research Zurich), and Sayeef Salahuddin (University of California, Berkeley). The 30 participants, among them 5 from cfaed, enjoyed one week of interesting talks and intensive, interdisciplinary discussions about the challenges and opportunities of advancing computing beyond current CMOS technology. The participants covered the whole spectrum from post-CMOS materials research up to software development for future heterogeneous computing systems. During the next months a report about the seminar will be compiled and made available at the seminar website.

Orchestration Retreat 2016

Published on in ORCHESTRATION (PATH ACTIVITIES)

In the first week of July, 2016 the Orchestration Path held its annual retreat in the monastery St. Marienthal (Ostritz). The PhD students, postdocs and investigators met and discussed recent developments and results achieved, worked on collaborative projects and exchanged ideas for future cooperations. Overall, it was a very intensive week with inspiring talks by the investitors, an invited Talk by Reinhard Berger of the Molecular Functional Materials Group on Synthetic Carbon Structures for Integration on the System Level, many fruitful discussions within and outside the working sessions completed by a boat&bike tour as team building event.

Orchestration path co-organizes Dagstuhl Seminar

Published on in ORCHESTRATION (PATH ACTIVITIES)

The path leader of the Orchestration path, Jeronimo Castrillon, is one of the four organizers of the Dagstuhl Seminar: "Wildly Heterogeneous Post-CMOS Technologies Meet Software", February 5–10, 2017. The aim of this seminar is to bring together computer scientists with experts in emerging technologies and researchers working interdisciplinarily across the fields, to foster a mutual understanding about the challenges and opportunities of advancing computing beyond current CMOS technology.

Invited Talk of Tal Ben-Nun in cfaed Seminar Series

Memory Access Patterns: The Missing Piece of the GPU Programming Puzzle

Published on in ORCHESTRATION (PATH ACTIVITIES)

In April, 2016 the Orchestration Path invited Tal Ben-Nun from Hebrew University of Jerusalem as a visior. In his talk he reported on his recent and ongoing work on memory access patterns [1].

Abstract: GPUs play an increasingly important role in high-performance computing, accelerating applications ranging from machine learning and computer vision to quantitative finance. In spite of their popularity, GPU programming remains a challenging task. Manual management of the multi-level memory hierarchy, inter-device synchronization, and various development constraints often generate lengthy, error-prone, and architecture-specific code. Therefore, it is imperative to develop new programming paradigms for efficient utilization of GPU and multi- GPU systems, without sacrificing code simplicity and portability. The talk presents Memory-Oriented Programming (MOP) - a data-centric programming model organized around memory access patterns. MOP categorizes parallel algorithms by their input and output parameters, where each parameter is characterized by a single pattern. Using these patterns, which cover most existing GPU applications, we show that it is possible to write short and intelligible code that attains high performance on a variety of GPU architectures and multi-GPU nodes. Furthermore, we demonstrate that the model can be used to increase the resilience of multi- GPU nodes by providing transparent failure protection mechanisms. We show that the resulting code is on par, and in some cases outperforms, existing manually optimized production-grade libraries, exhibiting near-linear speedup on a single node with multiple GPUs. Performance is measured on fundamental computational operations, as well as real-world applications in deep learning and non-negative matrix factorization.

[1] Tal Ben-Nun, Ely Levy, Amnon Barak, and Eri Rubin. 2015. Memory access patterns: the missing piece of the multi-GPU puzzle. In Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis (SC '15). ACM, New York, NY, USA, , Article 19 , 12 pages. DOI=http://dx.doi.org/10.1145/2807591.2807611

Project kickoff: Compilers for Computational Fluid Dynamics

Published on in ORCHESTRATION (PATH ACTIVITIES)

The project "A compiler approach for optimizing matrix operations in spectral-element methods" was kicked off with members from the Chair for Compiler Construction and the Professur für Strömungsmechanik. The project connects two layers of the Orchestration Stack, namely, algorithms and compilers. This is a so-called seed grant to start exploring languages and optimizations for spectral methods. We expect this activity to lead to a continued collaboration between the chairs.

Workshop on Attribute Grammars

Published on in ORCHESTRATION (PATH ACTIVITIES)

A workshop on attribute grammars was organized by Prof. Uwe Aßmann -- Chair of Software Technology at the TU Dresden, and Prof. Wolf Zimmermann - Chair of Software Engineering and Programming Languages at the Martin Luther University Halle-Wittenberg and took place on February 2nd, 2016.

The talks included various topics, such as possible application areas of attribute grammars, engineering and optimizing domain specific languages and runtime models. The used attribute grammar formalisms ranged from Ordered Attribute Grammars to Rewritable Reference Attribute Grammars.

A lively discussion of similarities, differences and overlaps followed the nine talks. For instance, both groups use attribute grammars to generate sets of linear equations to solve various, domain-specific problems.

Another workshop to follow up on the lively discussion and to intensify the exchange of the groups is expected to take place in May 2016.

 

2015 Internship of Tomas Karnagel at IBM Almaden Research Center, San Jose, CA, USA

Published on in ORCHESTRATION (PATH ACTIVITIES)

At the beginning of 2015, our PhD student Tomas Karnagel decided to do an internship at IBM to experience world-class industry research in order to get new inspiration for his research within the orchestration path of cfAED. Tomas met some IBM researchers at a conference in the USA half a year earlier and they found that his research is so closely connected to their work that an internship would be beneficial for both sides. At TU Dresden, Tomas works with database systems using GPUs to simulate highly heterogeneous environments, which will arise from the material paths of cfaed. At IBM, he investigated one single database operator and explored all the execution effects he could observe using a high-end GPU. The outcome of his research during his internship resulted in a joint (research) paper of IBM and Tomas.

 


 
As stated below, Tomas is very grateful for this internship:   „This internship broadened my mind substantially, as I received a deep insight into the impact of research on industry. The results are relevant for cfAED in many ways. They show how to port database operators to heterogeneous environments, and the pitfalls that have to be considered when doing so. It increased my attentiveness when looking at performance issues, which I can use for my own research within cfAED. I also hope to be able to use the results of this internship in my PhD thesis. During my internship, I did several talks, including a presentation of my research within the cfAED, thus spreading the word of this excellence cluster. Moreover, I found new friends among my IBM colleagues which established a firm basis for a lasting research relationship “

Grand Professors Week 2015

Published on in ORCHESTRATION (PATH ACTIVITIES)

On Nov. 23rd, 2015 the Orchestration Path presented and discussed the recent path achievements with the Grant Professors James R. Cordy, Hélène Kirchner, Jan M. Rabaey, and Itamar Willner at the Grand Professor Week 2015. The demos included the operating system M3 for heterogeneous manycores with an integrated tracing infrastructure (Vampier) running on Tomahawk T2 together with our linear algebra library m3la and a specifically designed computational fluid dynamics application. With a database demo on Tomahawk T3, the path presented first results in building a processor for database systems. The third demo on Orchestration Style Sheets illustrated the progress and future perspectives regarding the middleware. The path team wants to thank the Grand Professors for their time and the fruitful and inspiring discussion.

CC Chair in California: PACT conference and DREAM seminar

Published on in ORCHESTRATION (PATH ACTIVITIES)

Prof. Castrillon co-organized and presented at a tutorial co-located with the prestigious International Conference on Parallel Architectures and Compilation Techniques (PACT), held on October 18th, 2015 in San Francisco, CA USA. Prof. Castrillon talked about dataflow programming models and how they can be used to program heterogeneous computing systems. In the same week, Prof. Castrillon was invited to give a talk at the “Design of Robotics and Embedded systems, Analysis, and Modeling Seminar” (DREAMS) at UC Berkeley. He talked about “Analysis and software synthesis of KPN applications” and had an hour slot devoted to meet and discuss research directions with PhD students and Postdocs of the same university.