Nachrichten - PC2 (Uni Paderborn) http://pc2.uni-paderborn.de/ Aktuelle Informationen des PC2 (Universität Paderborn) de_DE Uni Paderborn Wed, 24 Apr 2019 11:42:51 +0200 Wed, 24 Apr 2019 11:42:51 +0200 Uni Paderborn news-89710 Fri, 12 Apr 2019 16:15:14 +0200 Materials from DATE2019 OpenCL FPGA Tutorial available about-pc2/announcements/news-events/article/news/materials-from-date2019-opencl-fpga-tutorial-available/ On March 25 2019, Tobias Kenter conducted a tutorial at the DATE 2019 conference in Florence under the title OpenCL design flows for Intel and Xilinx FPGAs - common optimization strategies, design patterns and vendor-specific differences. The tutorial material is now available for download. On March 25 2019, Tobias Kenter conducted a tutorial at the DATE 2019 conference in Florence under the title OpenCL design flows for Intel and Xilinx FPGAs - common optimization strategies, design patterns and vendor-specific differences. Featuring tools for both Intel and Xilinx FPGAs, this tutorial provides the community with practical guidance based on the experience in OpenCL based FPGA acceleration at PC².

While the DATE website only allowed on-side access to the tutorial material for registered participants, we have now made the slide deck available on our website along with the repository of sample codes and generated reports. For further details refer to this page.

]]>
news-89654 Mon, 08 Apr 2019 12:00:00 +0200 Computing Time Allocated for Call 2019-1 about-pc2/announcements/news-events/article/news/computing-time-allocated-for-call-2019-1/ The resource allocation board of the Paderborn Center for Parallel Computing (PC2) has approved the allocation of 156 million core-hours to 59 computing projects. Overall 13 large projects, requesting more than 2 million core-hours and 46 small projects (below 2 million core-hours) have been approved thereof 19 external projects of principal investigators not affiliated with Paderborn University. Paderborn/Germany, April 8 2019 – The resource allocation board of the Paderborn Center for Parallel Computing (PC2) has approved the allocation of 156 million core-hours to 59 computing projects. Overall 13 large projects, requesting more than 2 million core-hours and 46 small projects (below 2 million core-hours) have been approved thereof 19 external projects of principal investigators not affiliated with Paderborn University. 76% of the allocated computing time has been granted to users from Paderborn University, 10% to users at other scientific institutions in North Rhine-Westfalia and 14% to researchers from other states in Germany.  The supported projects are predominantly from the domains of condensed matter physics, physical chemistry, electrical- and mechanical engineering, and computer science. The 2019-1 call was the first call, where users could apply for resources on the new Noctua system at PC2. 21 projects have been allocated 80 million core-hours on this system alone. Three projects have been assigned 10 million or more core-hours on Noctua:

  • Condensed matter physics: "Photonic materials from ab-initio theory" under the principal investigator Prof. Dr. Wolf Gero Schmidt, Paderborn University (17 million core-hours on Noctua)
  • Physical chemistry: "DFTB calculations on water catalysis" under the principal investigator Dr. A Henao Aristizábal, Paderborn University (14.4 million core-hourse on Noctua)
  • Computer science: "On-the-fly machine learning" under principal investigator Prof. Dr. Eyke Hüllermeier, Paderborn University (9.9 million core-hours on Noctua and 9.9 million core-hours on Oculus)

"We are pleased to have received so many high-quality proposals from a large variety of different scientific disciplines and institutions. We are particularly encouraged that 7 projects using FPGA resources will be funded, indicating the emerging interest for this accelerator technology, which is a key competence of PC2. We have also seen steeply increasing interest in GPUs. This confirms the trend to use accelerators, which we will consider for future HPC system procurements", says Prof. Thomas D. Kühne, chairmen of the resource allocation board.

The newly approved projects will start on April 8.

Computing time on the HPC systems operated by PC2 is allocated for a period of 12 month on a competitive basis after a technical and scientific evaluation in a peer review process. For large projects there are two calls per year. The last call for proposals (2019-1) is available here. The next call for proposals will be opened in Juli 2019 and the next allocation period will start on October, 1 2019.

Further Information

Prof. Dr. Thomas D. Kühne
Chairman PC2 Resource Allocation Board

]]>
news-89112 Wed, 06 Feb 2019 10:35:36 +0100 Frühjahrstagung des ZKI AK-SC 2019 will be held at PC2 about-pc2/announcements/news-events/article/news/fruehjahrstagung-des-zki-ak-sc-2019-will-be-held-at-pc2/ On 14th and 15th of March 2019 the spring meeting of the ZKI work group on supercomputing will be held at the Paderborn Center for Parallel Computing (PC2). Participants can expect interesting talks and workshops on the current state and progress of high-performance computing in Germany. On 14th and 15th of March 2019 the spring meeting of the ZKI work group on supercomputing (Frühjahrstagung des ZKI Arbeitskreises Supercomputing 2019) will be held at the Paderborn Center for Parallel Computing (PC2). Participants can expect interesting talks and workshops on the current state and progress of high-performance computing in Germany.

For more information and registration, refer to https://pc2.uni-paderborn.de/go/zki2019 (event and website in German).

]]>
news-88957 Thu, 17 Jan 2019 16:39:14 +0100 Call for Proposals 2019-1 - Access to HPC Resources at PC² about-pc2/announcements/news-events/article/news/call-for-proposals-2019-1-access-to-hpc-resources-at-pc2/ The first Call for Proposals for HPC resources at PC² is now open. Scientists qualified in their respective field of research can submit proposals for access to HPC resources operated by Paderborn Center for Parallel Computing (PC²). We offer multi-million core-hour projects on the most recent generation of Intel Xeon processors with a low-latency interconnect. We intend to grant up to 30 million core-hours to an outstanding "flag-ship project".

PC² provides a portfolio of preinstalled applications and offers support for software installation/configuration of project-specific applications. We advise our users on using the resources in an efficient way. In particular, we provide consulting services for users from computational chemistry and solid-state physics. Furthermore, PC² encourages approximate computing and the acceleration of applications with Field-Programmable Gate Arrays (FPGA).

Important Dates

Opening date     17 January 2019    
Closing date 28 February 2019
Scientific reviews of proposals March 2019
Announcement of allocation decision End of March 2019
Allocation period for awarded large proposals 1 April 2019 until 31 March 2020


Detailed information on the available resources and how to apply can be found at https://pc2.uni-paderborn.de/go/hpc-proposal.

The Noctua system is a first phase of a multi-year funding of HPC resources. A significant extension will follow in a second phase in 2020/2021.
Refer to: https://pc2.uni-paderborn.de/go/hpc-noctua-news

We recommend current users of PC2 resources to apply for a small project in the call for proposals. Proposals for small projects can be submitted at any time and are continuously evaluated.
 

]]>
news-88453 Mon, 26 Nov 2018 16:22:25 +0100 Chemical bond formation in real time about-pc2/announcements/news-events/article/news/chemical-bond-formation-in-real-time/ A major breakthrough in the measurement and understanding of the electronic structure changes during chemical reactions has been achieved by a team of physicists led by Martin Wolf (Fritz-Haber-Institut Berlin) and Wolf Gero Schmidt (Universität Paderborn). The researchers were able to monitor for the first time the redistribution of electrons during a chemical reaction in real time. The corresponding driving forces and mechanisms could be unveiled in large-scale numerical simulations. The results of the joint research from the Paderborn and Berlin scientists have been published in the renowned multidisciplinary Science Magazine (https://doi.org/10.1126/science.aar4183). Numerical simulations that go beyond the molecular movie

A major breakthrough in the measurement and understanding of the electronic structure changes during chemical reactions has been achieved by a team of physicists led by Martin Wolf (Fritz-Haber-Institut Berlin) and Wolf Gero Schmidt (Universität Paderborn). The researchers were able to monitor for the first time the redistribution of electrons during a chemical reaction in real time. The corresponding driving forces and mechanisms could be unveiled in large-scale numerical simulations. The results of the joint research from the Paderborn and Berlin scientists have been published in the renowned multidisciplinary Science Magazine (https://doi.org/10.1126/science.aar4183).     

Schmidt explains: “Chemical reactions are associated with the formation or breaking of bonds between atoms. The accompanying atomic movements are extremely fast and occur on a femtosecond to picosecond time scale.  At the same time, the electronic structure changes, i.e., the positions and the energies of the electrons involved in the chemical bond are altered. This electron dynamics is decisive for chemical bond formation. Its observation, however, has been an elusive goal for a long time.”

In order to unveil the electron dynamics, atomic scale indium nanowires on top of a silicon surface were prepared. The indium valence electrons occupy localized orbitals in their ground state. Optical excitation by a laser pulse redistributes the electrons and results in the formation of a metallic bond along the wire direction. The modification of the electronic momentum and energy distribution during the bond formation – which lasts only a few femtoseconds – was measured with time- and angle-resolved photoemission spectroscopy. Numerical simulations were used to relate the measured quantities to the time-dependent spatial distribution of the electrons. Thus, the full reaction pathway – including the complete electronic structure dynamics – could be determined, thereby going beyond the molecular movie concept. 

The quantum mechanical modelling of several hundred excited electrons and atoms requires substantial high-performance computer resources. They were provided by the Paderborn Center for Parallel Computing and the High-Performance Computing Center Stuttgart. The simulations bridge the fundamental physical concept of energy and momentum distribution to describe electronic properties with the concept of chemical bond formation. “The understanding of the mutual interactions between the atomic and electronic degrees of freedom during chemical reactions can be considered the ‘Holy Grail’ of chemistry” Schmidt explains. “The numerical simulations resolve in unprecedented detail the relation between electronic excitation and reaction pathway. This is extremely relevant, as it will allow to tailor electronic configurations such that specific chemical reactions are favored.”

]]>
news-88236 Thu, 08 Nov 2018 21:53:16 +0100 PC2 User Meeting about-pc2/announcements/news-events/article/news/pc2-user-meeting/ We invite all current and prospective users to the first PC2 User Meeting. On 10 December 2018, we will hold the first annual PC2 User Meeting. The purpose of the meeeting is to inform all current and prospective users of the PC2 about the development of our HPC infrastructure services and to foster connections between users and members of PC2. A focus of this year's meeting will also be to explain the new process for computing time application which will be implemented in Spring 2019.

Apart from the technical program, we will also have an interesting scientific program, with invited talks by Prof. Jadran Vrabec and Prof. Jens Förstner, and a keynote presentation by Prof. Dominik Marx.

For more information about the event and and registration please refer to the PC2 User Meeting program.

]]>
news-87904 Tue, 02 Oct 2018 16:51:20 +0200 Tutorial on Working with OpenCL for Intel FPGAs about-pc2/announcements/news-events/article/news/tutorial-on-working-with-opencl-for-intel-fpgas/ On September 24 ­­­­­­– 25 2018, a tutorial on Working with OpenCL for Intel FPGAs was held at the Paderborn Center for Parallel Computing. With the new supercomputer Noctua that was inaugurated on September 21, the PC² has installed 32 state of the art FPGAs (Field Programmable Gate Arrays) as particularly energy efficient accelerators. On September 24 ­­­­­­– 25 2018, a tutorial on Working with OpenCL for Intel FPGAs was held at the Paderborn Center for Parallel Computing. With the new supercomputer Noctua that was inaugurated on September 21, the PC² has installed 32 state of the art FPGAs (Field Programmable Gate Arrays) as particularly energy efficient accelerators. With a focus on current users of the Paderborn HPC (high-performance computing) systems with little or no experience in FPGA programming, the training extends the base of developers that can write customized application code for the FPGAs.

The tutorial was conducted by Wolfgang Loewer of the FPGA Consulting company El Camino. Founded by former Altera employees in 1999, El Camino specialized from the beginning on engineering, training and consulting for FPGAs and CPLDs manufactured by Altera, now Intel FPGA. While most industrial customers still focus on FPGA designs specified with hardware description languages (HDL), new high-level synthesis design flows, in particular based on OpenCL have increasingly raised interest over the last couple of years.

On the first day of the tutorial, general concepts of OpenCL were introduced, along with a special emphasis on pipeline parallelism as the key to efficient execution on FPGAs. The second day covered advanced optimization techniques including making best use of the flexible layout of local memory resources within the FPGA. The presentation of concepts and examples was complemented with hands-on exercises that allowed the 16 participants to apply the new knowledge in practice.

]]>
news-87834 Mon, 24 Sep 2018 21:42:07 +0200 Supercomputer Noctua inaugurated about-pc2/announcements/news-events/article/news/supercomputer-noctua-inaugurated/ On September 21 2018, the new supercomputer “Noctua” (Latin: Little Owl) was inaugurated at Paderborn University. The high-performance computing system is expected to be one of the ten most powerful university installations within Germany. This is the first step on the path into a new era for Paderborn University and PC2. On September 21, the new supercomputer “Noctua” (Latin: Little Owl) was inaugurated at Paderborn University. The high-performance computing system is expected to be one of the ten most powerful university installations within Germany. This is the first step on the path into a new era: The construction of a new data center and research facility will commence in 2019, with the federal state of Germany, the state North Rhine-Westphalia and Paderborn University contributing a total of €25.4M until 2022. Out of this, a total of €10M will be invested towards the installation of the Noctua supercomputer. Based on the recommendation of the German Council for Science and Humanities (Wissenschaftsrat), the Joint Science Conference (Gemeinsame Wissenschaftskonferenz, GWK) approved funding for the project in June 2017.

“This is a special and identity-establishing day for the University,” said Prof. Dr. Johannes Blömer, Vice President for Research and Junior Academics. On the one hand, the computing system would be able to perform highly complex simulations, and on the other hand, it would provide the necessary computing power for researchers throughout Ostwestfalen-Lippe, said Blömer.

With the commissioning of the new system, the university’s Paderborn Center for Parallel Computing (PC²), replaces their former flagship system by a significantly more powerful and more energy-efficient HPC (high-performance computing) system. Prof. Dr. Christian Plessl, Chairman of the PC² board and computer scientist: “The new Cray CS 500 HPC cluster is the most powerful installation in Paderborn so far and delivers a performance of 535 TFLOPS in the LINPACK benchmark. The unique property: In addition to the 256 compute nodes with Intel Xeon Gold 2.4 GHz processors there are 16 additional nodes that contain two so-called FPGA cards each. These customizable hardware blocks play an increasingly important role in modern computer architecture.” Plessl continues: “FPGAs (Field Programmable Gate Arrays) already attract high attention as particularly energy-efficient accelerators in commercial data centers. With a unique HPC installation of 32 state of the art FPGAs that are especially promising for scientific applications, the PC² can become a pioneer in establishing this technology. Noctua is currently expected to be the biggest installation of its kind in the world and provides the best conditions for prolific research.”

Prof. Dr. Thomas D. Kühne, Vice Chairman of the PC² board and theoretical chemist, is convinced: “With Noctua, all computational science and engineering users get the much-needed computing power to be internationally competitive. In particular, highly complex simulations that otherwise could be done only at national highest-performance computing centers—or not at all, can now be performed”.

Pascal Barbolosi, Vice President EMEA sales at Cray, confirms: “Cray has always been devoted to pushing the limits of scientific computing, by combining the most advanced technologies to achieve the desired performance. Integrating a high ratio of state-of-the-art FPGAs in an advanced supercomputer cluster system will provide a remarkable heterogeneous computing platform that enables researchers to perform a new dimension of experiments. We are happy to support Paderborn University researchers in this exciting phase with our long-standing experience in reconfigurable systems.” 

Frank Klapper, Head of Information Management and University Development at Bielefeld University and spokesman for “Digitale Hochschule NRW”, was also visibly excited: “Here [in Paderborn], you have the highest expertise. Paderborn is simply passionate about this topic.” 

The Noctua HPC system is expected to be generally available for users by the end of October 2018. The complete specifications of the system are available here.

A movie documenting the construction of the Noctua system can be found on our youtube channel

]]>
news-87562 Thu, 09 Aug 2018 14:11:43 +0200 Exact results with inexact computations: Prof. Plessl and Kühne receive Paderborn University Research Award about-pc2/announcements/news-events/article/news/exact-results-with-inexact-computations-prof-plessl-and-kuehne-receive-paderborn-university-researc/ According to recent estimations, IT systems consume about 15% of the global electrical power. Due to the continuing exponential growth of computing systems, there is a strong economic and ecologic interest to increase their energy-efficiency. How this goal can be achieved is demonstrated by the computer scientist Prof. Dr. Christian Plessl and the theoretical chemist Prof. Dr. Thomas D. Kühne with their interdisciplinary research project, which is supported by the executive board of Paderborn University with an award of 150,000 Euro. Exact results with inexact computations: Prof. Dr. Christian Plessl and Prof. Dr. Thomas D. Kühne are winning the Paderborn University Research Award for their GreenIT project

According to recent estimations, IT systems consume about 15% of the global electrical power. Due to the continuing exponential growth of computing systems, there is a strong economic and ecologic interest to increase their energy-efficiency. How this goal can be achieved is demonstrated by the computer scientist Prof. Dr. Christian Plessl and the theoretical chemist Prof. Dr. Thomas D. Kühne with their interdisciplinary research project, which is supported by the executive board of Paderborn University with an award of 150,000 Euro. 

Change of paradigm in information processing

Within just a few decades, information technologies have developed to a key technology and exert a strong and continuously increasing influence on economy, science, and society. The exponential growth of IT does however not only concern the available computational capabilities but also the number of IT systems. This fast-paced development urgently demands energy-efficient computational methods. However, the pursuit of ever higher performance and efficiency exposes challenges. “If we insist on reproducible computation and storage of information, the further increase of efficiency is limited by several practical limitations. Ultimately, the efficiency is limited by a hard theoretical limit, the so-called Landauer-limit,” explains Christian Plessl, professor of computer science and director of the Paderborn Center for Parallel Computing (PC²).

These roadblocks can be addressed with the help of “Approximate Computing.”  Thomas D. Kühne, professor of theoretical chemistry, adds, “The concept of ‘Approximate Computing’ which we will be studying in this project is fundamentally changing the paradigm of computing by deliberately giving up exact reproducibility and allowing inexact computation and storage.” The benefit of this approach is that it enables extremely efficient computation. The approach should even allow for undercutting the Landauer-limit in the long run.  

Exact results and improved energy efficiency despite inexact computations

For turning this aggressive approach into practice, new computational methods must be developed that allow devising exact results from inexact computations. On the computer architecture side, Plessl and Kühne will use application-specific hardware devices (FPGAs) which can be customized and are substantially more energy-efficient than conventional processors (CPUs) which are very flexible but provision lots of unused chip area for operations and data. The novel fault-tolerant numerical methods will be integrated in the CP2K molecular dynamics simulation software, which has been developed with substantial contributions of Kühne’s group for many years.

„If we succeed, our project would provide an impressive demonstration of the feasibility and potential of Approximate Computing even beyond scientific computing. It is conceivable that the approach could also be successfully applied to numerous other domains, such as global optimization problems, numerics, computer graphics, quantum computing, and machine learning,” according to Kühne. 

Even across scientific disciplines, the research at Paderborn University will also profit from this project according to Plessl, “The adoption of the Approximate Computing-concept would not only allow researchers from the departments of chemistry, physics, and computer science to simulate larger problems but would also reduce the electrical energy demand in the University’s high-performance computing center PC².”

Plessl and Kühne will receive the research award in a ceremony at the next New Year’s Reception of Paderborn University in January 2019. The Paderborn University research award aims at researchers with visionary ideas, daring hypotheses, creative and unconventional technologies, or innovative and bold methods to fund projects beyond the mainstream.

 

]]>
news-87354 Wed, 18 Jul 2018 13:27:15 +0200 University celebrates 20th anniversary of training program for IT specialists in system integration at Paderborn Center for Parallel Computing about-pc2/announcements/news-events/article/news/university-celebrates-20th-anniversary-of-training-program-for-it-specialists-in-system-integration/ Since August 1998, the University of Paderborn has been training IT specialists in system integration. The Paderborn Center for Parallel Computing (PC2) celebrated this anniversary by organizing a reunion of the program’s graduates, and the invitation was accepted by 13 of the 16 former and current students. At 38 percent, the share of women who have completed the PC² program is well above the 2016 national German average of 7.9 percent. Since August 1998, the University of Paderborn has been training IT specialists in system integration. The Paderborn Center for Parallel Computing (PC2) celebrated this anniversary by organizing a reunion of the program’s graduates, and the invitation was accepted by 13 of the 16 former and current students. At 38 percent, the share of women who have completed the PC² program is well above the 2016 national German average of 7.9 percent.

During their training and later in their professional lives, system integrators plan, install and administer systems and networks. Another important aspect of the program is IT security, including firewalls and encryption. “When they finish our three-year program, our trainees have excellent career opportunities,” said trainer Andreas Krawinkel.

Krawinkel opened the reunion with a welcome, after which Professor Dr. Christian Plessl, the Managing Director and Chairman of the Board of the PC², provided a brief overview of the center’s current activities and projects. Technology played a role, too: Following the presentation, the group toured the O Building at the university from the basement to the roof. In addition to the high-performance data center, the tour included areas that are normally off limits to the general public, such as the basement that houses the emergency power system, the fire-extinguishing system and the roof with its emergency cooling systems. The program concluded with a dinner where everyone had lots of opportunities to talk and catch up. The former students all agreed that they should not have to wait 20 years for the next reunion.

More information about training programs offered by the university is available here: http://www.uni-paderborn.de/en/zv/4-1/ausbildung/uebersicht-der-ausbildungsberufe/ 

]]>
news-87172 Tue, 26 Jun 2018 13:20:23 +0200 New Research Spotlight: Investigation of Photonic Structures about-pc2/announcements/news-events/article/news/new-research-spotlight-investigation-of-photonic-structures/ Within the last decade the research field of Optoelectronics and Photonics has emerged as an active and successful focal area at the Paderborn University. Some highlights investigated are novel light diodes & lasers, ultrafast digital communication via fibers, integrated optical circuits, biological photonic structures, concepts for optical quantum computers, optical sensors, antennae for light, metamaterials, and holograms from ultrathin layers. We have added a new research spotlight to our website: Prof. Dr. Jens Förstner and Dr. André Hildebrandt provide an insight into the Investigation of Photonic Structures.

Within the last decade the research field of Optoelectronics and Photonics has emerged as an active and successful focal area at the Paderborn University. Some highlights investigated are novel light diodes & lasers, ultrafast digital communication via fibers, integrated optical circuits, biological photonic structures, concepts for optical quantum computers, optical sensors, antennae for light, metamaterials, and holograms from ultrathin layers.

Read more...

]]>
news-87148 Thu, 21 Jun 2018 20:30:48 +0200 PC2 at ISC'18 about-pc2/announcements/news-events/article/news/pc2-at-isc18/ We are participating at ISC 2018 in Frankfurt, which is the largest European conference and trade show on High-Performance Computing. You can meet us at the Gauss-Allianz booth (M-230) on Tuesday from 10:00-11:00. We are participating at ISC 2018 in Frankfurt, which is the largest European conference and trade show on High-Performance Computing. You can meet us at the Gauss-Allianz booth (M-230) on Tuesday from 10:00-11:00.

Dr. Tobias Kenter will present a talk on "Accelerating Modern Scientific Simulations with FPGAs"

Abstract

The Paderborn Center for Parallel Computing is investing into Field-Programmable Gate Arrays (FPGAs) as accelerator technology for HPC. Complementary to a recently signed procurement of a system with 32 latest generation Stratix 10 FPGAs, development efforts are put into FPGA acceleration libraries, infrastructure, and applications. First results of a collaboration project for the acceleration of modern scientific simulation codes have recently been presented at one of the major scientific conferences of the FPGA community, the 26th IEEE International Symposium on Field-Programmable Custom Computing Machines (FCCM). At ISC, we want to make the HPC community aware of these results and their context.

The exploration of FPGAs as accelerators for scientific simulations has so far mostly been focused on small kernels of methods working on regular data structures, for example in the form of stencil computations for finite difference methods. In computational sciences, often more advanced methods are employed that promise better stability, convergence, locality, and scaling. Unstructured meshes are shown to be more effective and more accurate, compared to regular grids, in representing computation domains of various shapes. Using unstructured meshes, the discontinuous Galerkin method preserves the ability to perform explicit local update operations for simulations in the time domain. 

In our current work, we investigate FPGAs as the target platform for implementation of the nodal discontinuous Galerkin method to find time-domain solutions of Maxwell’s equations in an unstructured mesh. In a first step, off-chip memory bandwidth demands are limited by maximizing data reuse and fitting constant coefficients into registers and on-chip RAM blocks. The data layout and parallel access patterns are suitably partitioned to allow this local memory to feed wide data paths with hundreds of parallel floating point operations reliably with data in every cycle. 

Despite the high arithmetic intensity obtained after customizing the local memory layout, some phases and variants of the application require high sustained off-chip bandwidth in the order of tens of GB/s and thus close to the target platforms peak bandwidth. Achiving this bandwidth is particularly challenging due to indirect addressing into the unstructured mesh, which leads to deterministic, yet irregular memory access patterns. We show that by decoupling off-chip memory accesses from the computations on the FPGA, high memory bandwidth can be sustained even for this indirect addressing pattern.

Using the Intel OpenCL SDK for FPGAs, we present implementations for different polynomial orders of the discontinuous Galerkin method. In different phases of the algorithm, either computational or bandwidth limits of the Arria 10 platform are almost reached, achieving more than 100 GFLOPS/s in several kernels and thus overall outperforming a highly multithreaded CPU implementation by around 2x. With ongoing efforts towards the maintainability of these codes for computational scientists, towards scaling the designs for the bigger next generation Stratix 10 FPGAs and over multiple compute nodes, we see this application domain as a highly promising field for FPGA acceleration in HPC.

 

]]>
news-86897 Mon, 14 May 2018 13:36:00 +0200 CP2K User Tutorial on Computational Spectroscopy about-pc2/announcements/news-events/article/news/cp2k-user-tutorial-on-computational-spectroscopy/ The Dynamics of Condensed Matter Research Group (Prof. Thomas D. Kühne) and the Paderborn Center for Parallel Computing are co-organizing a users tutorial on Computational Spectroscopy with CP2K. The Dynamics of Condensed Matter Research Group (Prof. Thomas D. Kühne) and the Paderborn Center for Parallel Computing are co-organizing a users tutorial on Computational Spectroscopy with CP2K.

The tutorial addresses beginners and practitioners and will cover the following topics:

  • Gaussian Plane Wave (GPW) and Gaussian Augmented Plane Wave (GAPW) methods
  • Ab-initio Molecular Dynamics & Nuclear Quantum Effects
  • NEXAFS, NMR & Vibrational Spectroscopy (IR, Raman, VCD, ROA)
  • Absolutely Localized Molecular Orbitals (ALMO)
  • Implicit Solvent Methods & DFT+U

The tutorial will be held on August 27-29th at Paderborn University, Germany. Attendance is free of charge but participants must cover their own travel and accommodation costs. For more detailed information and to register go to:

  https://pc2.uni-paderborn.de/teaching/trainings/cp2k-tutorial/

]]>
news-86641 Fri, 13 Apr 2018 13:55:00 +0200 Massively Parallel Computation of Approximate Inverse Roots for Large Sparse Matrices about-pc2/announcements/news-events/article/news/massively-parallel-computation-of-approximate-inverse-roots-for-large-sparse-matrices/ Our Submission “A Massively Parallel Algorithm for the Approximate Calculation of Inverse p-th Roots of Large Sparse Matrices” has been accepted for publication at the Platform for Advanced Scientific Computing (PASC) conference, taking place in July 2018 in Basel, Switzerland. Our Submission “A Massively Parallel Algorithm for the Approximate Calculation of Inverse p-th Roots of Large Sparse Matrices” has been accepted for publication at the Platform for Advanced Scientific Computing (PASC) conference, taking place in July 2018 in Basel, Switzerland. The paper is the result of combined effort by members of the Departments of Computer Science and Chemistry at Paderborn University, the Paderborn Center for Parallel Computing and the Barcelona Supercomputing Center.

In our paper we describe a massively parallel algorithm for the computation of approximate inverse p-th roots of large sparse matrices. By deliberately breaking up dependencies between elements of the output matrix, we allow parallel computation of the elements on a large compute cluster with only little communication overhead. For an N x N matrix, our algorithm allows distributing the load over N processing nodes, where each of the node only has to operate on a significantly smaller but dense matrix. The result matrix has the same sparsity pattern as the input matrix, allowing efficient reuse of allocated memory structures. Following the idea of Approximate Computing, we allow more efficient and highly parallel computation by accepting the result to deviate from a precisely calculated solution.

In our work, we demonstrate two applications for such approximately calculated matrices, namely orthonormalization of basis functions in electronic structure codes and preconditioning of ill-conditioned linear systems for subsequent application of iterative solvers. We discuss the performance of the proposed algorithm both from a theoretical and from a practical standpoint. For the latter, we demonstrate its scalability using a distributed implementation on part of our OCuLUS cluster, utilizing up to 1024 CPU cores.

A non-final preprint of our work can be found on arXiv: http://arxiv.org/abs/1710.10899

Additionally, we have published our prototype implementation as well as scripts used in our evaluation under the MIT license on github: https://github.com/pc2/SubmatrixMethod/

Reference:

M. Lass, S. Mohr, H. Wiebeler, T.D. Kühne, and C. Plessl
A Massively Parallel Algorithm for the Approximate Calculation of Inverse p-th Roots of Large Sparse Matrices

In Proc. Platform for Advanced Scientific Computing (PASC). 2018. Accepted for publication.

Abstract:

We present the submatrix method, a highly parallelizable method for the approximate calculation of inverse p-th roots of large sparse symmetric matrices which are required in different scientific applications. We follow the idea of Approximate Computing, allowing imprecision in the final result in order to be able to utilize the sparsity of the input matrix and to allow massively parallel execution. For an n x n matrix, the proposed algorithm allows to distribute the calculations over n nodes with only little communication overhead. The approximate result matrix exhibits the same sparsity pattern as the input matrix, allowing for efficient reuse of allocated data structures.

We evaluate the algorithm with respect to the error that it introduces into calculated results, as well as its performance and scalability. We demonstrate that the error is relatively limited for well-conditioned matrices and that results are still valuable for error-resilient applications like preconditioning even for ill-conditioned matrices. We discuss the execution time and scaling of the algorithm on a theoretical level and present a distributed implementation of the algorithm using MPI and OpenMP. We demonstrate the scalability of this implementation by running it on a high-performance compute cluster comprised of 1024 CPU cores, showing a speedup of 665x compared to single-threaded execution.

]]>
news-86547 Tue, 03 Apr 2018 12:00:00 +0200 Cray Commissioned to Deliver FPGA-Accelerated Supercomputer to Paderborn University about-pc2/announcements/news-events/article/news/cray-commissioned-to-deliver-fpga-accelerated-supercomputer-to-paderborn-university/ Paderborn University announced today that it has selected a Cray® CS500™ cluster system as its next-generation supercomputer. This procurement is the first phase of the Noctua project in which a multi-petaflop-system with a total budget of 10M euros will be commissioned until 2022. The initial high-performance computing (HPC) system provides academic researchers from Paderborn University and nationwide with computing resources primarily for computational material science, optoelectronics and photonics, and computer system research. The system is expected to go into production in 2018. PADERBORN, April 3, 2018 — Paderborn University announced today that it has selected a Cray® CS500™ cluster system as its next-generation supercomputer. This procurement is the first phase of the Noctua project in which a multi-petaflop-system with a total budget of 10M euros will be commissioned until 2022. The initial high-performance computing (HPC) system provides academic researchers from Paderborn University and nationwide with computing resources primarily for computational material science, optoelectronics and photonics, and computer system research. The system is expected to go into production in 2018.

Translating the continuous innovation in HPC technologies into meaningful benefits for researchers requires a close collaboration of technology providers, method and code developers, and users. "Cray has a reputation both as a technology pioneer and for pushing the boundaries of supercomputing unlike any other IT company," said Prof. Dr. Christian Plessl, director and head of the board of Paderborn Center for Parallel Computing (PC²). "Cray is an ideal partner for delivering the computing power required by our users and for collaborating with PC² in forward-looking HPC systems research."

A distinctive feature of the cluster is the deployment of 32 FPGA accelerators featuring the latest generation Intel® Stratix® 10 FPGAs, placing it at the very top of academic HPC production systems with state-of-the-art FPGAs. The selected FPGAs, with 5,760 variable-precision DSP blocks each, are well suited to floating-point heavy scientific computations. They reached general availability just in time to be installed in the Noctua cluster. A first set of applications that benefit from FPGA acceleration is currently ported and was reengineered in close cooperation with computational scientists. This infrastructure will be used to study the potential of FPGAs for energy-efficient scientific computing, allowing PC² to maintain its leading role in establishing this technology in HPC. “Cray is dedicated to supporting academic researchers around the world in achieving their initiatives and we are pleased that Cray’s system provides the flexibility in supporting state-of-the-art FPGAs for scientific computing,” said Pascal Barbolosi, vice president of EMEA sales at Cray. “This latest commission will allow Cray to support Paderborn University researchers with this exciting phase of their computer system research.”

The Cray® CS500™ system will be installed at the Paderborn Center for Parallel Computing. It is expected that the new system will deliver triple the performance of the previous flagship system, OCuLUS. The new supercomputer will be deployed with 272 dual-socket compute nodes powered by Intel's latest generation Xeon® Skylake processors. The used 20-core CPUs are the highest-performing of Intel’s Xeon Gold portfolio. Overall, the system will provide a total of 11,000 cores and 51 TB RAM. The system is connected by a 100 Gbps Intel® OmniPath interconnect supporting highly parallel MPI applications.

Additionally, a Cray® ClusterStor™ L300N storage appliance will be installed. The combination of Lustre®, GridRAID and the NXD flash accelerator technology will offer 720 TB of highly reliable storage for I/O-intensive compute jobs. "With the updated HPC system our users will again have the latest HPC technology with the most powerful computing nodes available," said Dr. Jens Simon, manager of HPC systems and services at the Paderborn Center for Parallel Computing.

When signing the purchase agreement, operations vice president Dr. Simone Probst emphasized the importance of the supercomputer for the strategy of Paderborn University. "It is essential that we enable our researchers to perform world-class research by providing them with powerful computing resources. The university as a whole also profits from this infrastructure in the competition to attract the best talent." 

]]>
news-86319 Wed, 28 Feb 2018 11:55:00 +0100 Efficient Computation of Inverse Matrix Roots about-pc2/announcements/news-events/article/news/efficient-computation-of-inverse-matrix-roots/ Our article "A General Algorithm to Calculate the Inverse Principal p-th Root of Symmetric Positive Definite Matrices has been accepted for publication in the Communications in Computational Physics (CICP). This article is a joint work by the research groups of three PC² board members from the Departments of Chemistry, Computer Science and Mathematics. Our article A General Algorithm to Calculate the Inverse Principal p-th Root of Symmetric Positive Definite Matrices has been accepted for publication in the Communications in Computational Physics (CICP). This article is a joint work by the research groups of three PC² board members from the Departments of Chemistry, Computer Science and Mathematics. In the article, we describe an iteration scheme for the calculation of inverse roots of matrices and prove and analyze its convergence under specific conditions. The proposed method allows trading of the number of iterations required to obtain a result against the computational effort within each iteration. The approach is a generalization of previously published methods such as the hyperpower method by Altman (1960) and the iteration scheme proposed by Bini, Higham and Meini (2005).

Inverse matrices and inverse roots are required in different applications, such as preconditioning and solving generalized eigenvalue problems, in particular to solve Schrodinger and Maxwell equations. In earlier work [1,2,3] we have shown the great potential of Approximate Computing techniques for these application areas, as results are not always required to be exact. These techniques aim at increasing the energy-efficiency of algorithms, opposing the ever-growing energy demands of HPC systems. Iterative methods, as the one described in our article, are especially well suited for the use of approximative methods, as introduced errors can be corrected in subsequent iterations.

An early, non-final preprint of our article is available from https://arxiv.org/abs/1703.02456

 

Abstract

We address the general mathematical problem of computing the inverse p-th root of a given matrix in an efficient way. A new method to construct iteration functions that allow calculating arbitrary p-th roots and their inverses of symmetric positive definite matrices is presented. We show that the order of convergence is at least quadratic and that adjusting a parameter q leads to an even faster convergence. In this way, a better performance than with previously known iteration schemes is achieved. The efficiency of the iterative functions is demonstrated for various matrices with different densities, condition numbers and spectral radii.

 

[1] Lass, Kühne, Plessl: Using Approximate Computing in Scientific Codes, Workshop on Approximate Computing. Oct. 2016.

[2] Lass, Kühne, Plessl: Using Approximate Computing for the Calculation of Inverse Matrix p-th Roots, IEEE Embedded Systems Letters. Accepted for publication.

[3] Lass, Kühne, Plessl: A Massively Parallel Algorithm for the Approximate Calculation of Inverse p-th Roots of Large Sparse Matrices, Preprint.

]]>
news-85985 Tue, 19 Dec 2017 16:10:45 +0100 Paderborn University becomes a full member of the Gauß-Allianz about-pc2/announcements/news-events/article/news/paderborn-university-becomes-a-full-member-of-the-gauss-allianz/ The Gauß-Allianz member status of the Paderborn University is raised from associated member to full member. With this, PC² enters the top 10 of German HPC computer centres. The Gauß-Allianz consists of the leading German supercomputer centres. The association was founded to promote research and development in the field of High Performance Computing (HPC) and to coordinate the HPC-related activities in Germany. Another important goal is to increase the visibility on an international level by bundling the expertise of the participating supercomputing centres. Paderborn University, represented by the Paderborn Center for Parallel Computing (PC²) has been a founding member of the Gauß-Allianz.
Based on the recent decision of the Joint Science Conference (Gemeinsame Wissenschaftskonferenz, GWK) to approve the funding of the Paderborn University for a new national level high-performance computer system, the Gauß Alliance member status of the Paderborn University is raised from associated member to full member. With this, PC² enters the top 10 of German HPC computer centres. This opens further opportunities for scientists of the Paderborn University to do their research more efficient and effective.

The Gauß-Allianz is a non-profit association for the promotion of science and research. To this end, it supports the scientific community in Germany by creating the conditions for sustainable and efficient use of supercomputing resources of the top performance classes, in particular through the coordination and pooling of complementary skills and diversified computer architectures and the associated access structure.
The mission of the Gauß-Allianz is to coordinate the HPC related activities of the members. With the provision of versatile computing architectures and by combining the expertise of the participating centres, this creates the ecosystem necessary for computational science. Strengthening the research and increasing the visibility to compete on an international level are further goals of the Gauß-Allianz. It is especially an important milestone to maintain the leading role of Germany in the European supercomputing activities.

]]>
news-85909 Fri, 08 Dec 2017 12:33:38 +0100 Research and service report 2014/16 about-pc2/announcements/news-events/article/news/research-and-service-report-201416/ The latest edition of the PC² research and service report is available for download now. It is our pleasure to present you the latest research and service report of PC². The format and contents of the report have been redesigned from the ground up. Our mission was to present you the exciting research that is enabled by our high-performance computing systems and services in a fresh and easily accessible way. The report also offers you the opportunity to look behind the scenes and learn more about our computing infrastructure, our innovative computing systems research, and the motivated team that makes almost anything possible.

The report is available for download. We will also be glad to send you a paper copy while supply lasts.

]]>
news-85370 Fri, 22 Sep 2017 10:24:19 +0200 Intel awards Paderborn University a cluster fueled by Xeon processors and Intel Arria 10 FPGAs about-pc2/announcements/news-events/article/news/intel-awards-paderborn-university-a-cluster-fueled-by-xeon-processors-and-intel-arria-10-fpgas/ The Paderborn Center for Parallel Computing (PC²) has been selected by Intel to host a computer cluster that uses Intel’s Xeon processor with its Arria 10 FPGA software development platform. The Paderborn Center for Parallel Computing (PC²) has been selected by Intel® to host a computer cluster that uses Intel’s Xeon® processor with its Arria 10 FPGA software development platform. This server cluster connects Intel® Xeon® processor with an in-package field-programmable gate array (FPGA) via the platform’s high-speed QuickPath interconnect improving system bandwidth. The Intel® FPGA can be programmed to serve as a workload-optimized accelerator offering substantial performance, agility, and energy-efficiency advantages.  This solution is suitable for a number of application domains, such as machine learning, data encryption, compression, image processing and video-stream processing. The platform also an ideal experimentation platform for innovative operating system or computing systems research, that focuses on novel approaches of integrating CPUs with accelerators at the software and hardware level. 

 “We are very happy to have been selected by Intel as one of only two academic sites worldwide to host a cluster based on Intel Xeon processors and Intel Arria 10 FPGAs . Our computing center has a strong research background in accelerating demanding applications with FPGAs. The availability of these systems allows us to further expand our leadership in this area and – as a next step – bring Intel FPGA accelerators from the lab to HPC production systems,” says Prof. Dr. Christian Plessl, director of the Paderborn Center for Parallel Computing, who is been active in this research area for almost two decades.

 Researchers worldwide can get access to the cluster by applying to Intel’s Hardware Accelerator Research Program. “We are looking forward to collaborate with Intel and other members of the Hardware Accelerator Research Program on using FPGA acceleration for emerging HPC and data center workloads. By provisioning access to the system to a large number of researchers, we are also gathering experience in how to manage systems with FPGA accelerators in a multi-user setting and for handling parallel applications that use multiple servers with FPGAs. This experience is crucial for deploying systems with FPGAs at scale,” explains Dr. Tobias Kenter, senior researcher and FPGA expert at the Paderborn Center for Parallel Computing.

 Currently, the Paderborn Center is working on accelerating applications including theoretical physics, material sciences and machine learning with FPGAs.  This work is in collaboration with scientists from the application areas. In addition, novel domain-specific programming approaches for FPGAs are being developed to simplify the use of FPGAs for developers without a hardware design background.

]]>
news-85238 Fri, 01 Sep 2017 07:56:15 +0200 Launch of the HighPerMeshes Project about-pc2/announcements/news-events/article/news/launch-of-the-highpermeshes-project/ Recently, a network of several German institutions from science and industry started the HPC project HighPerMeshes to tackle challenges associated with the efficient implementation of algorithms working on unstructured grids on heterogeneous processor architectures. At the kick-off meeting in April, organized by the Paderborn University, the status-quo of the relevant algorithms, the technological aspects and the requirements from the industrial and academic perspective were presented by the project partners. The goal of the project is the development of a state-of-the-art domain-specific software framework to help scientists in the implementation of simulation codes working on unstructured grids. Innovations in the IT infrastructure comprising many-core and reconfigurable processing devices pose challenges to the software developers in both academia and industry to take advantage of new hardware features within their software designs - the hardware-software gap. Recently, a network of several German institutions from science and industry started the HPC project HighPerMeshes to tackle challenges associated with the efficient implementation of algorithms working on unstructured grids on heterogeneous processor architectures. At the kick-off meeting in April, organized by the Paderborn University, the status-quo of the relevant algorithms, the technological aspects and the requirements from the industrial and academic perspective were presented by the project partners.

The goal of the project is the development of a state-of-the-art domain-specific software framework to help scientists in the implementation of simulation codes working on unstructured grids. Innovations in the IT infrastructure comprising many-core and reconfigurable processing devices pose challenges to the software developers in both academia and industry to take advantage of new hardware features within their software designs - the hardware-software gap.

 HighPerMeshes focuses on supporting efficient, parallel and scalable implementations of iterative algorithms on unstructured grids for heterogeneous computing platforms. Such algorithms are increasingly applied in academia and in the commercial sector, e.g., for Maxwell's equations, scattering problems, acoustics, or biomedical simulations, while extending or replacing traditional methods on regular grids.

By providing domain-specific constructs and library elements through C++ high-level abstraction, and complemented by a state-of-the-art compiler infrastructure, the HPC software developer will be enabled to generate performance optimized, parallel and scalable code for all major HPC processor architectures including multi-/many-core CPU, GPU and FPGA. The software framework developed within HighPerMeshes targets its applicability to complex real-world simulation packages like the Kaskade7 Finite Element Toolbox (ZIB, Berlin) or the implementation of the Discontinuous Galerkin Time Domain method of the research group of Prof. Dr. Jens Förstner (Paderborn University). We expect that our approach can provide an effective and sustainable solution for a variety of HPC software developers who are required to target high-end computing platfor

Background

The HighPerMeshes project started in April 2017 and will run until March 2020.

The primary project partners include

  • the High-Performance IT Systems group (Prof. Dr. Christian Plessl) and Theoretical Electrical Engineering group (Prof. Dr. Jens Förstner) of the Paderborn University,
  • the Architecture and Compiler Design group (Dr. Frank Hannig) at the Chair of Hardware/Software Co-Design (Prof. Dr. Jürgen Teich) of the Friedrich-Alexander University Erlangen-Nürnberg,
  • the CC-HPC at Fraunhofer ITWM in Kaiserslautern (Dr. Daniel Grünewald), and
  • the Numerical Mathematics Dept. (Dr. Martin Weiser) and the Algorithms for Innovative Architectures group (Dr. Thomas Steinke) at the Zuse Institute Berlin.

The associated partners

  • CST - Computer Simulation Technology AG (now Dassault Systèmes) and
  • Intel Deutschland GmbH

will contribute to the project with software requirements from the industrial perspective and expertise in the technological roadmap.

The project is funded by the German Federal Ministry of Education and Research with 1.6 Mio. €. 

For more information see www.highpermeshes.info or contact Prof. Dr. Christian Plessl, Paderborn University.

]]>
news-84843 Tue, 27 Jun 2017 14:05:38 +0200 Decision of the Joint Science Conference (Gemeinsame Wissenschaftskonferenz, GWK) for a high-performance computer at Paderborn University about-pc2/announcements/news-events/article/news/decision-of-the-joint-science-conference-gemeinsame-wissenschaftskonferenz-gwk-for-a-high-perform/ Following the recommendation by the German Council for Science and Humanities (Wissenschaftsrat), the Joint Science Conference (Gemeinsame Wissenschaftskonferenz, GWK) has decided on June 23 to approve funding for a new high-performance computer at Paderborn University. Following the recommendation by the German Council for Science and Humanities (Wissenschaftsrat), the Joint Science Conference (Gemeinsame Wissenschaftskonferenz, GWK) has decided on June 23 to approve funding for a new high-performance computer at Paderborn University. The Council for Science and Humanities had issued a recommendation with best ratings for the proposal of Paderborn University for a new high-performance computer in May.

From 2018 to 2022, the federal government, the state government of North Rhine-Westfalia and Paderborn University will provide a total of 25.4 million €. In two phases, a total of 10 million € will be invested in a new high-performance computer, which – presumably – will rank among the ten most powerful HPC systems in Germany when taken into operation. For the construction of the modern and particularly energy-efficient computing center, which will presumably be finished in 2020, an amount of 15.4 million € is provided.

Prof. Dr. Christine Silberhorn, vice president for research and junior academics: “With Noctua we establish the best conditions for interdisciplinary research at the cross section of applications, methods and computing systems research and we are well prepared for the requirements coming in future. For Paderborn University this is a further important step towards increasing the international visibility of its top level research. We are very proud to become one of the most powerful university HPC systems in Germany with our new research infrastructure.”

The new compute center will be the first pure research building being approved at Paderborn University.

Further information can be found in our press release from May 5, 2017.

]]>
news-84781 Wed, 21 Jun 2017 09:00:00 +0200 Meet us at ISC 2017 about-pc2/announcements/news-events/article/news/meet-us-at-isc-2017/ We are participating at ISC 2017 in Frankfurt, which is the largest European conference and trade show on High-Performance Computing. You can meet us on Wednesday in the exhibition hall at the HiPEAC booth (A-1430) from 10:00-13:30 and at the Gauss-Allianz boot from 16:00-18:00. We are participating at ISC 2017 in Frankfurt, which is the largest European conference and trade show on High-Performance Computing. You can meet us on Wednesday in the exhibition hall at the HiPEAC booth (A-1430) from 10:00-13:30 and at the Gauss-Allianz boot from 16:00-18:00.

Our presentations this year, center around the topic of Heterogeneous and FPGA-Accelerated Computing:

Approximate Computing – A Suitable Paradigm for Scientific Computing?

Approximate Computing (AC) is an emerging computing paradigm that compromises the exactness of the computation for improved performance or energy efficiency. While AC has already attracted attention for domains that are inherently resilient to approximations (e.g. processing of audio and video or machine learning) the potential in scientific computing is so far not well understood. In our current research, we study, which numerical methods can profit from AC techniques if accuracy bounds are desirable or needed.

Can OpenCL Compilers make FPGAs Accessible for HPC?

Numerous case studies have demonstrated that FPGA can provide substantial performance and efficiency benefits to computationally challenging applications.  Recently, FPGAs technology and tools have made significant strides towards improved floating-point capabilities and high-level programming toolflows, which makes the technology accessible for a wider range of developers. In this presentation, we report on our ongoing work of accelerating a finite difference solver for Maxwell’s equations using an OpenCL-based FPGA toolflow.

Automatic and Transparent Acceleration of Applications

Because of the stagnation of single-thread performance and the move towards accelerators, legacy applications do no longer see a performance increase when executed on the latest computing systems. If codes are expected to be used for a long time and have a large user base, code modernization, i.e. porting the code to modern programming languages and libraries with parallelism support, is likely worth the effort. For legacy codes, manual modernization may not be economical. In our work, we study just-in-time compilation approaches that automatically adapt these codes to use GPU and FPGA accelerators without any manual intervention.

Accelerating and Parallelizing Tree-Search for Crypto-Attacks with FPGAs

FPGAs are most efficient for regular, throughput computing on data streams. Hence, the acceleration of tree search problems with FPGAs is typically not regarded as particularly promising. In this work on accelerating so-called “cold boot” crypto attacks, we have proven this common knowledge wrong. We have demonstrated that FPGAs are beneficial for solving certain hard tree search problems and that we can also successfully apply a work stealing strategy to parallelize the hardware implementation. Finally, we show that it even makes sense to fully customized FPGA accelerators that tailored to the problem instance not only the problem class.

Speakers

Prof. Dr. Christian Plessl

Dr. Tobias Kenter

Dr. Jens Simon

]]>
news-84594 Tue, 06 Jun 2017 17:00:00 +0200 Hardware Accelerated Cold-Boot Attacks (Open Source Hardware and Software Release) about-pc2/announcements/news-events/article/news/hardware-accelerated-cold-boot-attacks-open-source-hardware-and-software-release/ The protection of computer systems and the information they store have gained massive importance in our modern society. Over the last years, we have worked in the area of hardware accelerated cryptographic attacks. In particular, we have studied how so-called Cold-Boot Attacks can be accelerated with the help of field programmable gate arrays (FPGAs). The full source code of our work is now available online. Computers are nowadays involved in nearly all aspects of our everyday life with the growth of new technology and algorithms. They have brought enormous benefits in various fields, including health care, travel, business or artificial intelligence. However, there is also a downside of the medal: Computer systems are constantly under threats of unintentional errors and criminal activities. Thus, their protection and especially the protection of the information they store gains more and more importance in our modern society. Computer security is not only a highly relevant topic in industry and science, but also attracts the attention of everyday users.

Over the last years, we have worked in the area of hardware accelerated cryptographic attacks. In particular, we have studied so-called cold-boot attacks and how they can be accelerated with the help of field programmable gate arrays (FPGAs).

The key idea of these cold-boot attacks is to break encrypted systems by obtaining remains of the secret key from main memory and to reconstruct the secret key afterwards. If successful, full disk encryption or encrypted communication (e.g. WLAN, HTTPS, SSH or VOIP) can be easily circumvented since the encryption key is revealed. In software, the reconstruction can be performed using a recursive, branch-and-bound tree-search algorithm that exploits redundancies for constraining the search space. However, the runtime of these algorithms grows rapidly with increasing complexity of the problem instance, which limits the practicability of the approach on conventional hardware. Therefore, we investigate how this branch-and bound algorithm can be accelerated with FPGAs.

Full source code now available online

The source code of our software and hardware implementations as well as our evaluation data and a demo application are now available online at github.com/pc2/coldboot. As special features, we

  • explore hardware (FPGA) workers that autonomously cooperate using work stealing to allow parallel execution and full utilization of the target FPGA,
  • show the advantages of instance-specific designs that target a specific problem instance to improve performance, and finally
  • demonstrate how instance-specific designs can be generated just-in-time such that the provided speedups outweigh the additional time required for design synthesis.

All hardware designs are compared against highly optimized parallel software implementations using Intel Cilk Plus. Our evaluation shows that our work stealing approach is scalable with the available FPGA resources and provides speedups proportional to the number of workers. 

Further information 

More information about our work on hardware accelerated cryptographic attacks can be found on the project’s webpage. Details on our work stealing and instance-specific designs are described in more depth in our upcoming publication in the ACM Transactions on Reconfigurable Technology and Systems (TRETS).

]]>
news-84401 Fri, 05 May 2017 11:45:29 +0200 The German Council for Science and Humanities (Wissenschaftsrat) recommends funding for a high-performance computer at Paderborn University: 25m€ for Noctua and a new computing center – ensuring sustained development of computational sciences about-pc2/announcements/news-events/article/news/the-german-council-for-science-and-humanities-wissenschaftsrat-recommends-funding-for-a-high-perfo/ The German Council for Science and Humanities has issued a funding recommendation with the highest ranking for the proposal of Paderborn University for a new high-performance computer. One of the ten most powerful university-hosted HPC (high-performance computing) centers in Germany

The German Council for Science and Humanities has issued a funding recommendation with the highest ranking for the proposal of Paderborn University for a new high-performance computer. Between 2018 and 2022, the federal government, the state government of North Rhine-Westfalia and Paderborn University will provide a total of 25.4 million €. In two phases, a total of 10 million € will be invested in a new high-performance computer, which – presumably – will rank among the ten most powerful HPC systems in Germany when taken into operation. The proposal is subject to final approval by the Joint Science Conference (Gemeinsame Wissenschaftskonferenz, GWK) in its session on June 23, 2017. The Joint Science Conference decides on all issues in funding of science that target the federal government and the state governments of Germany and decides on science and research policies.

For the construction of the modern and particularly energy-efficient computing center, an amount of 15.4 million € is provided. The computing center will be built on the site that was formerly used by the RailCab train test track. The building will provide office space for joint interdisciplinary research at the cross section of applications, methods and computing systems research.

Professor Dr. Christian Plessl, computer scientist and managing director and head of the board of the Paderborn Center for Parallel Computing (PC²), commenting the decision of the Council for Science and Humanities: “For the continuously growing group of researchers performing interdisciplinary research at Paderborn University, having a powerful high-performance computer infrastructure is necessary to continue to perform top-level research that is nationally and internationally recognized.” As a competence center for parallel computing, the PC² studies the efficient use of supercomputers with innovative technologies and operates HPC-systems to provide academic researchers in the region Ostwestfalen-Lippe with computing services.

Due to the rapidly growing demand for computing power and the spread of computational methods into almost all areas of science, engineering and increasingly also humanities, the currently installed HPC system OCuLUS and the existing computing center can no longer satisfy the future requirements. Hence, Paderborn University has submitted a proposal for a new, larger scale HPC system and a future-oriented computing center building to the research infrastructure funding program of the Council of Research and Humanities.

The funding also includes the construction of a computing center that is specifically tailored to the needs of high-performance computing. Christian Plessl: “This investment lays the foundation for a sustained development of computational sciences in Paderborn.” Because of its national relevance, the new computer system will also be used by researchers across Germany.

Dr. Thomas D. Kühne, professor for technical chemistry and vice head of the board of PC², commenting the decision of the Council of Science and Humanities: “The funding of the supercomputer Noctua is a milestone for the research groups working with numeric methods at Paderborn University, in particular for material sciences and optoelectronics/photonics. We will be able to routinely perform simulations that would otherwise only be possible occasionally at national Tier-1 high-performance computing centers, or would not have been possible at all.” According to Prof. Kühne, the approval of the proposal should also be considered as a proof of excellence for the whole area of computational sciences at Paderborn University: “This honor is invaluable to strengthen the proposal ‘Cluster for Analysis and Optimal Design of Functional Systems (CoreFun)’ that has recently been submitted as a Cluster of Excellence proposal to the German Research Foundation (DFG).”

The first installation phase of the Noctua HPC system is planned for 2018 and will initially be taken into operation in the existing computing center. The installation of the second phase is planned for 2020 in the new computing center after its completion.

Further details on the project and the evaluation can be found in the report of the Council of Science and Humanities.

]]>
news-84211 Mon, 03 Apr 2017 09:07:05 +0200 Simulations at PC² explain ultrafast phase transition - published in Nature about-pc2/announcements/news-events/article/news/simulations-at-pc2-explain-ultrafast-phase-transition-published-in-nature/ Wires just four atoms wide that change from being insulators to electrical conductors when struck by a laser pulse have been studied by researchers from Paderborn and Duisburg. The team has shown that the phase transition can occur as fast as quantum mechanics allows, something that was not previously thought possible. The technique could prove useful, e.g., for ultrafast detectors. Wires just four atoms wide that change from being insulators to electrical conductors when struck by a laser pulse have been studied by researchers from Paderborn and Duisburg. The team has shown that the phase transition can occur as fast as quantum mechanics allows, something that was not previously thought possible. The technique could prove useful, e.g., for ultrafast detectors. 

Phase transitions are ubiquitous in all forms of matter. Previous research has shown that phase changes may be completed within several picoseconds in bulk materials, but typically take far longer at surfaces. Now the Paderborn and Duisburg researchers demonstrated that surface phase transitions can occur in less than a picosecond. They prepared indium wires adsorbed on silicon surfaces. At room temperature, the wires are metallic, whereas at temperatures below 120 K the indium wire atoms rearrange into insulating hexagons. The researchers first cooled the system to 30 K and measured the electron diffraction pattern of the insulating surface. They then hit the surface with laser pulses and used electron pulses to see how the diffraction pattern had changed. For delay times longer than 350 femtoseconds, the diffraction pattern of the insulating state was replaced by that of the metallic wires, i.e. the indium atoms can rearrange themselves within one 350 quadrillionths of a second.

Computational modelling by theoreticians at Paderborn University shows that indium electrons are photo-excited to higher energy levels by the laser pulse. "The laser pulse populates electronic states that form previously non-existing In-In bonds. Other In-In bonds are broken by the laser pulse due to the depopulation of some electronic states," explains Wolf Gero Schmidt. This forces the atoms to move and eventually causes the phase transition. As the atoms move, the electronic properties of the wire change and it becomes metallic. Remarkably, this happens while the atoms are still cold. "The heating of the surfaces requires order of magnitude longer times than the phase change,” Schmidt explains. 

The theoreticians calculated the energy of the indium wire system for various electronic excitations and numerous structural configurations. Molecular dynamics calculations were performed on the thus obtained potential energy surfaces. “The quantum mechanical modelling of several hundred excited electrons and atoms requires substantial high-performance computer resources. They were provided by the Paderborn Center for Parallel Computing PC² and the Höchstleistungsrechenzentrum Stuttgart,” Schmidt acknowledges. The time calculated numerically for the phase change closely matches the measured value. The simulations show that the laser pulse energy soon dissipates into the silicon lattice. After the insulator-metal transition, therefore, the indium wire lacks the activation energy needed to return to the its original state. It is trapped in the metastable metallic state. The system studied by the researchers from Paderborn und Duisburg is probably the fastest electronic switch ever observed. 

The research is published in the 13 April 2017 issue of Nature, see http://dx.doi.org/10.1038/nature21432

]]>
news-84067 Tue, 07 Mar 2017 15:50:24 +0100 Appointment of new PC² advisory board about-pc2/announcements/news-events/article/news/appointment-of-new-pc2-advisory-board/ On recommendation by the PC² management board, the executive board of Paderborn University has appointed new advisory board members. On recommendation by the PC² management board, the executive board of Paderborn University has appointed the following advisory members to serve on the PC² advisory board:

  • Dr. Christoph Hagleitner, IBM Zurich Research Lab, Switzerland
  • Volker Hamer, NRW Ministry for Innovation, Science and Research (MWIF)
  • Prof. Dr. Dominik Marx, Ruhr-Universität Bochum
  • Prof. Dr. Alexander Reinefeld, Zuse Institute Berlin
  • Marie-Christine Sawley, PhD, Intel Exascale Reserach Lab Paris, France
  • Dr. Felix Wolfheimer, Computer Simulation Technology AG Darmstadt

We welcome our new advisory board members and are looking forward to a fruitful collaboration.

On behalf of the management board, we would also like to thank our previous advisory members who retired from their position with the end of their term:

  • Dr. Karsten Beins
  • Dr. Horst Joepen
  • Prof. Dr.-Ing. Michael Resch
  • Prof. Dr.-Ing. Ulrich Rückert
  • Dipl. Phys. Thomas Rüter

We are grateful for their longstanding support of our institute and we are looking forward to continue our collaborations in the future.

 

]]>
news-83910 Wed, 08 Feb 2017 14:48:26 +0100 Presentation of upcoming project group CustoNN about-pc2/announcements/news-events/article/news/presentation-of-upcoming-project-group-custonn/ On Monday we presented our plans for the upcoming project group CustoNN, open for Master students in Computer Science and Computer Engineering. On Monday we presented our plans for the upcoming project group CustoNN, open for Master students in Computer Science and Computer Engineering. The slides and additional material are now available on the course webpage and registration will take place during the first registration phase in PAUL.

]]>
news-83526 Wed, 30 Nov 2016 22:24:51 +0100 Website relaunched and new logo about-pc2/announcements/news-events/article/news/website-relaunched-and-new-logo/ PC² has launched a redesigned web presence and logo. Today, PC² has launched a new web presence. Our new website follows the corporate design of Paderborn University and offers a cleaner visual and structural design to communicate our activities research, teaching and HPC services. The website now also supports a responsive design, which significantly improves the usability of the website on devices with smaller screens.

Additionally, we are also launching the redesigned logo for our institute. Taking up visual cues from the predecessor version, the new logo features a more contemporary and lighter look and matches with the revised corporate design language of the university.

]]>
news-83393 Mon, 07 Nov 2016 09:55:00 +0100 Prof. Christian Plessl is the new chairman of the PC2 about-pc2/announcements/news-events/article/news/prof-christian-plessl-is-the-new-chairman-of-the-pc2/ Prof. Christian Plessl appointed as chairmen of the PC² management board On June 24, the PC2 management board appointed Prof. Dr. Christian Plessl as the new chairman of the management board effective October 1, 2016. Prof. Plessl is the successor to Prof. Dr. Holger Karl, who we would like to thank for his work in recent years. Prof. Plessl will also continue to serve as managing director PC2.

]]>
news-83391 Tue, 06 Sep 2016 13:55:00 +0200 European researchers simplify optimal software execution on complex electronics about-pc2/announcements/news-events/article/news/european-researchers-simplify-optimal-software-execution-on-complex-electronics/ Prof. Plessl's research group has participated in the European FP7 research project SAVE, which has been completed successfully this week. Prof. Plessl's research group has participated in the European FP7 research project SAVE, which has been completed successfully this week.

Please find below the official press release.

­

European researchers simplify optimal software execution on complex electronics

Milan, Italy, 31 August 2016


Today marks the completion of the SAVE project, which challenged a team of engineers and researchers to explore how complex hardware systems can more efficiently execute data intensive applications. Funded by the European Union, SAVE has led to a number of innovations in hardware, software and operating system (OS) components. When integrated together, they can reduce application deployment costs and maximize usage of heterogeneous system computing units, resulting in energy efficiency being improved by up to twenty per cent. 

A range of complex electronic systems stands to benefit from these innovations, including computer data centers, consumer electronics, automotive products and complex industrial electronics. 

The computing units can be on chip, for example central processing units (CPUs) ranging from small and low-power to high-end and efficient, graphics processing units (GPUs), and dedicated accelerators. Alternatively, the units can be off chip, such as racks of dedicated accelerators or field-programmable gate arrays (FPGAs).The prototyped technologies will enable performance and energy-efficiency gains in high-performance computing (HPC) and embedded heterogeneous systems.

Key achievements include:

 

  • Platform behavior monitoring and task dispatching hardware and software: the first toolset closely tracks the performance and use rate of the various computing units available in the heterogeneous systems. The second toolset decides which computing units are best suited for the job.
  • Just-in-Time compilation technology: using SAVE technologies, at runtime, a single application-code representation is optimized to the many possible hardware targets of the platform: CPUs, GPUs, accelerators, FPGAs.
  • Hardware and software virtualization technologies: these technologies efficiently expose the dedicated processing engines to the many virtual machines (VM) running on these systems.

 

The teams successfully prototyped virtualized GPUs, virtualized FPGA-based data-flow engines (DFEs), and virtualized application-specific accelerators.

About SAVE

These innovations are the culmination of three years of collaborative research by a team of three academic and four industrial partners. The international team worked together to achieve dynamic optimization of workload assignments across multicore system computation units, operated simultaneously from several virtualized operating systems. These developments lay the foundations for industrial partners to further optimize ever more complex systems, including HPC systems for finance applications and automotive embedded systems.Funded by the European Commission’s Seventh Framework Program (FP7), the project was launched on 01 September 2013, under the project name SAVE: ‘Self-Adaptive Virtualization-Aware High-Performance/Low-Energy Heterogeneous System Architectures.’

SAVE project website: http://www.fp7-save.eu/

Further information: Cristiana Bolchini, SAVE project coordinator, cristiana.bolchini(at)polimi.it.


SAVE innovations enable the dynamic optimization of task assignments across processing units operated from several OSes, all from single code representation of each application.

 

Partners’ views on the SAVE project

Politecnico di Milano (Italy, SAVE project coordinator)

“Dynamic trade-offs in performance and energy are becoming increasingly synonymous with heterogeneous system management. Within the SAVE project, researchers designed self-adaptive software components and associated optimization policies to manage at runtime the resources offered by the heterogeneous computing systems developed by our hardware vendor partners.”

Prof. Cristiana Bolchini, Professor, Dipartimento di Elettronica Informazione e Bioingegneria, Politecnico di Milano,

ARM (United Kingdom)

“The next generation of graphics technology and smart connected products will require the optimized execution of simultaneous compute-intensive tasks through several operating systems. The ARM team’s focus on the SAVE project was to prototype virtualization technologies and identify pathways to maximize system efficiency in emerging GPU use cases in markets such as automotive and HPC.”

Eric Hennenhoefer, Vice President, Research, ARM

Maxeler Technologies (United Kingdom)

“Self-adaptive behavior and dynamic resource allocation is key to improving performance and efficiency in heterogeneous HPC systems with dedicated DFEs. Adding orchestration and virtualization to Maxeler’s dataflow computers helps us to efficiently run large dynamic workloads in a multitenant environment, a technology that is highly promising for a range of users in finance, science and engineering.”

Georgi Gaydadjiev, Vice President, Dataflow Software Engineering

STMicroelectronics (France)

“In complex heterogeneous multicore systems, managing data communication efficiently and securely between host processors, on-chip computing units (such as the GPU and hardware accelerators) and off-chip islands of computation is key to providing power-efficient and fast dynamic kernel offloading.  SAVE has brought the task of optimizing heterogeneous System-on-Chip solutions much closer to full industrialization.”

Philippe Quinio, Group Vice President, IP Sourcing and Strategy, STMicroelectronics

Technological Educational Institute of Crete (Greece)

“Two key components for self-adaptive heterogeneous system architecture are the hardware resource usage monitoring mechanisms and the workload dispatch mechanisms. By working closely with hardware vendors, we designed low-level communication solutions between the general purpose CPUs and the specialized computation units like GPUs and DFEs.”

Dr George Kornaros, Professor, Technological Educational Institute of Crete

Paderborn University (Germany)

“A key obstacle for best use of heterogeneous computing is the manual effort to optimize code for each hardware resource type. Our autonomous runtime and compilation system removes this burden: critical parts of applications are automatically identified and they are seamlessly translated and optimized for execution on GPU and FPGAs accelerators. With SAVE, Paderborn University paved the way for wider deployment of the underlying technologies, namely, static and dynamic code analysis and just-in-time code generation techniques.”

Dr Christian Plessl, Professor, Paderborn University

Virtual Open Systems

“The SAVE project has been for Virtual Open Systems the testing ground to demonstrate that hardware assisted virtualization and innovative hypervisor extensions are able to provide near native performance, thus paving the way for extensive usage of virtualization of programmable accelerators in many market segments. The SAVE project has strengthened the company's position in the open-source community, while creating new business opportunities in both product and service lines.”

Daniel Raho, Director, Virtual Open Systems

]]>