Felix Wittwer

Felix Wittwer

E-mail

Phone

Fax

Visitor's Address

felix.wittwer1@tu-dresden.de
S/MIME certificate GPG Public Key

Phone: +49 (0)351 463 43711

Fax +49 (0)351 463 39995

Helmholtzstrasse 18
3rd, BAR III59
01069 Dresden
Germany

 

Curriculum Vitae

Felix Wittwer received his Bachelor's and Master's degree in Computer Science from the TU Dresden in September 2017 and August 2020, respectively. During his studies, he specialized in compilers, operating systems and natural language processing. Since his Bachelor's degree, he became a regular at the Chair for Compiler Construction, working with Sebastian Ertel on the Ohua compiler framework for implicit parallelism. In his master thesis he worked on language and compiler support in Ohua for the deterministic execution of parallel applications with shared state.

In November 2020 he joined the chair as research assistant, where he works on the project “EVEREST: Design environment for extreme-scale big data analytics on heterogeneous platforms”, funded by the European Union’s Horizon 2020 research and innovation program under grant agreement no 957269.

Research Interests/Open Topics

I am highly interested in the topic of "compilers with a societal impact". How can our compiler research be applied to large-scale real-world problems? As part of the EVEREST project, which is funded by the European Union’s Horizon 2020 research and innovation program under grant agreement no 957269, I am therefore looking into weather prediction and traffic simulation applications.

Weather Prediction — Tensor and Stencil Operations

The weather is undoubtedly one of the most impactful environment factors in our everyday life. It's correct prediction depends on the correct analysis and evaluation of large amounts of different data. These calculations are driven by various stencil and tensor operations which are part of a larger forecasting model. Specifically, we are looking into the WRF model, an established and widely used model for weather analysis and forecasting. We are looking to apply this model within the EVEREST project to air quality measurements for industrial plants producing air pollution, weather predictions to maximize the energy production of renewable power producers and lastly also to desaster prevention in a traffic routing use case.

The challenge here lies within the sheer size and complexity of the WRF model. Hence, we are looking into optimizing the aforementioned calculations by building a specifically tailored Domain Specific Language (DSL) which can optimize certain often-used tensor operations for instance.

Also, we want to work on providing a sort of high-level algorithmic description language to split the work up and place it on different nodes of our hardware, which are highly heterogeneous in their computating capacities:

Flow of Data in EVEREST

The idea is to preprocess as much data as possible as far away from the central server farm as possible to reduce the data and computing load.

Traffic Routing — Dataflow Optimizations

Cars are for many people, especially those living in suburban areas, a necessity. But how can navigation systems route the daily traffic efficiently to minimize traffic jams and thereby pollution of the environment? Traffic simulators pose a simple, yet elegant solution to this question:

Intended flow of the Traffic Simulator

But given the size of even a single modern city's road network, these algorithms need optimizations on the dataflow level to make them more efficient and modular. This would allow the incorporation of additional modules like live weather prediction to route traffic away from potentially dangerous regions (e.g. in case of an upcoming storm or an approaching flood).

Hence, I am interested in finding optimizations on the dataflow level for large-scale applications like this.

Are you interested in working in any of these areas?

If you are interested in participating in the research on any of these topics, either as SHK or as part of your curriculum, feel free to contact me via mail so we can have a chat.

Publications

  • 2020

  • Felix Wittwer, "Ohua as an STM Alternative for Shared State Applications" , Master's thesis, TU Dresden, Aug 2020. [Bibtex & Downloads]