skip to content

ORKA-HPC - OpenMP for Reconfigurable Heterogenous Architectures

Partner

Associated partners

  • Intel GmbH

Universität zu Köln, Regionales Rechenzentrum (RRZK)

- Project coordination - 

Prof. Dr. Ulrich Lang, Viktor Achter, Dr. Stefan Zellmann

At the RRZK, a wide variety of processor architectures are being investigated for their suitability for user programs for many years. The goal here is to be able to run user codes on clusters of CPUs or other architectures such as GPGPUs, Xeon Phi and FPGAs in a scalable and efficient manner and to better understand the knowledge about these technologies and parallelisation schemes. The RRZK led the BMBF project "NGSgoesHPC", with the aim of making various alignment and assembly algorithms from the field of Next Generation Sequencing scalable. A deep understanding of these algorithms and their suitability for modern processors was developed in the process. In the BMBF project "SMOOSE - Systemic Analysis of Modulators of Oncogenic Signalling", the team worked on accelerating a successful gene analysis pipeline based on FPGAs.

Friedrich Alexander Universität (FAU)

Prof. Dr. Michael Philippsen

Michael Philippsen's group can draw on many years of experience for a large part of the research areas relevant to the project. This includes translator construction, parallel programming languages for multicore systems, code generation and the implementation of middleware for heterogeneous platforms, source-to-source translators in the OpenMP environment for GPUs, the programming of multicore CPUs and accelerators in clusters and grids as well as the development of runtime systems for heterogeneous and parallel environments including HPC. In addition, there is experience in the efficient implementation of distributed shared memory, system software optimization for embedded systems and the analysis of source code archives for similar code segments.

Konrad-Zuse-Zentrum für Informationstechnik Berlin (ZIB)

Dr. Thomas Steinke

At the ZIB, competence in the use of heterogeneous platforms (FPGA, GPGPU, Intel MIC) in HPC and enterprise applications has been built up over the past decade. In the field of reconfigurable computing, a number of computing platforms (Cray XD1, SGI RASC100, Convey HC1) and FPGA boards (Nallatech HC101) with high-level language approaches (Mitrion-C, ROCCC) were evaluated. As part of the OpenFPGA initiative, the generic platform API "OpenFPGA GenAPI" and an API for the application class of molecular dynamics have been developed.

Fraunhofer SCAI

Horst Schwichtenberg

SCAI and ZIB were project partners in the BMBF project "ENHANCE - enabling heterogeneous hardware acceleration using novel programming and scheduling models" to enable better integration and simplified use of heterogeneous computing systems with multicore CPUs, GPUs and FPGAs. With the developed framework, partial solutions for easier programmability, performance estimation and scheduling on heterogeneous computing units could be developed and successfully tested by industrial partners, especially for multicore CPUs and GPUs. In addition, modules were developed in the area of image processing and data reduction on FPGAs, which are used for real-time processing of large amounts of data. The knowledge gained from this work is to be incorporated into this project.

Jesko Schwarzer

Freelancer (External contracting)

Jesko Schwarzer performed a proof of concept of a ray-tracing ASIC in his StartUp (CreateIt AG, 1999-2000). Within a 43-member team, he was vice-chairman of the company and, as CTO, responsible for controlling the pre-development which included simulations. Since then, he has worked on various software and hardware development projects for various global companies.