Parallel Architecture Research Eindhoven - PARsE

Welcome to PARsE

Welcome to the website of the parallel architecture research team. The PARsE team is a subdivision of the Electronic Systems group, part of the Electrical Engineering departement at Eindhoven University of Technology (TU/e) in The Netherlands.

Contents of the website

This website contains information on previous, current and future research, education and events. We update the research log periodically with new posts concerning the latest updates in research, discussing articles published at journals and conferences from among others ACM and IEEE. You will find the following sections on the website:
Latest research log entries

IEEE TC July 2016 Spotlight Paper

IEEE Transactions on Computers (TC) has selected our paper Configurable XOR Hash Functions for Banked Scratchpad Memories in GPUs as the July 2016 Spotlight Paper. This month you can download the article for free from the IEEE TC website.

We also made a short video presentation which can be watched on YouTube (English). The video presentation is also available in Spanish and Chinese.

Microserver has landed

The first microserver from IBM/Astron has arrived at the TU/e. We have started benchmarking the microserver. Next, we plan to update our Bones source-to-source compiler and add the microserver as a target in the coming weeks.

More news on the microserver: IBM and ASTRON provide microserver prototypes to three Dutch partners, Nieuwe microserver van ASTRON kan Noord-Nederland honderden banen opleveren (in Dutch) and Data niet meer naar computer, maar computer naar de data (in Dutch).

test setup: microserver in a box

Bones v1.6.0

Today we released version 1.6.0 of our A-Darwin and Bones tools. In this release it is now possible to have multiple scops in a single source file. Also some bugs have been fixed, including processing of empty scops and a skeleton argument mismatch. The source code is available on Github, the documentation can be found online or in PDF.

If you have any questions, suggestions or bug reports, feel free to contact us.

About Bones
Bones is a source-to-source compiler based on algorithmic skeletons and a new algorithm classification. The compiler takes C-code annotated with class information as input and generates parallelized target code. Targets include NVIDIA GPUs (through CUDA), AMD GPUs (through OpenCL) and x86 CPUs (through OpenCL and OpenMP). More information on the Bones project page or in the paper Bones: An Automatic Skeleton-Based C-to-CUDA Compiler for GPUs.

Latest publications

Code generation for reconfigurable explicit datapath architectures with LLVM
Adriaansen, M.; Wijtvliet, M.; Jordans, R.; Waeijen, L. and Corporaal, H., DSD 2016 - 19th Euromicro Conference on Digital System Design, 2016
Link to publication
Tags: compiler, LLVM, CGRA