@article{WendeMarsmanKimetal.2018, author = {Wende, Florian and Marsman, Martijn and Kim, Jeongnim and Vasilev, Fedor and Zhao, Zhengji and Steinke, Thomas}, title = {OpenMP in VASP: Threading and SIMD}, journal = {International Journal of Quantum Chemistry}, number = {Emerging Architectures in Computational Chemistry}, publisher = {Wiley Online Library}, doi = {10.1002/qua.25851}, pages = {e25851}, year = {2018}, abstract = {The Vienna Ab initio Simulation Package (VASP) is a widely used electronic structure code that originally exploits process-level parallelism through the Message Passing Interface (MPI) for work distribution within and across nodes. Architectural changes of modern parallel processors urge programmers to address thread- and data-level parallelism as well to benefit most from the available compute resources within a node. We describe for VASP how to approach for an MPI + OpenMP parallelization including data-level parallelism through OpenMP SIMD constructs together with a generic high-level vector coding scheme. We can demonstrate an improved scalability of VASP and more than 20\% gain over the MPI-only version, as well as a 2x increased performance of collective operations using the multiple-endpoint MPI feature. The high-level vector coding scheme applied to VASP's general gradient approximation routine gives up 9x performance gain on AVX512 platforms with the Intel compiler.}, language = {en} } @article{SalemSchintkeSchuettetal.2018, author = {Salem, Farouk and Schintke, Florian and Sch{\"u}tt, Thorsten and Reinefeld, Alexander}, title = {Data-flow scheduling for a scalable FLESnet}, journal = {CBM Progress Report 2017}, isbn = {978-3-9815227-5-4}, doi = {10.15120/GSI-2018-00485}, pages = {130 -- 131}, year = {2018}, language = {en} } @inproceedings{GholamiSchintkeSchuett2018, author = {Gholami, Masoud and Schintke, Florian and Sch{\"u}tt, Thorsten}, title = {Checkpoint Scheduling for Shared Usage of Burst-Buffers in Supercomputers}, booktitle = {Proceedings of the 47th International Conference on Parallel Processing Companion; SRMPDS 2018: The 14th International Workshop on Scheduling and Resource Management for Parallel and Distributed Systems}, doi = {10.1145/3229710.3229755}, pages = {44:1 -- 44:10}, year = {2018}, abstract = {User-defined and system-level checkpointing have contrary properties. While user-defined checkpoints are smaller and simpler to recover, system-level checkpointing better knows the global system's state and parameters like the expected mean time to failure (MTTF) per node. Both approaches lead to non-optimal checkpoint time, intervals, sizes, or I/O bandwidth when concurrent checkpoints conflict and compete for it. We combine user-defined and system-level checkpointing to exploit the benefits and avoid the drawbacks of each other. Thus, applications frequently offer to create checkpoints. The system accepts such offers according to the current status and implied costs to recalculate from the last checkpoint or denies them, i.e., immediately lets continue the application without checkpoint creation. To support this approach, we develop economic models for multi-application checkpointing on shared I/O resources that are dedicated for checkpointing (e.g. burst-buffers) by defining an appropriate goal function and solving a global optimization problem. Using our models, the checkpoints of applications on a supercomputer are scheduled to effectively use the available I/O bandwidth and minimize the failure overhead (checkpoint creations plus recalculations). Our simulations show an overall reduction in failure overhead of all nodes of up to 30\% for a typical supercomputer workload (HLRN). We can also derive the most cost effective burst-buffer bandwidth for a given node's MTTF and application workload.}, language = {en} } @misc{GholamiSchintkeSchuettetal.2018, author = {Gholami, Masoud and Schintke, Florian and Sch{\"u}tt, Thorsten and Reinefeld, Alexander}, title = {Modeling Checkpoint Schedules for Concurrent HPC Applications}, journal = {CoSaS 2018 International Symposium on Computational Science at Scale}, year = {2018}, language = {en} } @inproceedings{SchmidtkeSchintkeSchuett2018, author = {Schmidtke, Robert and Schintke, Florian and Sch{\"u}tt, Thorsten}, title = {From Application to Disk: Tracing I/O Through the Big Data Stack}, booktitle = {High Performance Computing ISC High Performance 2018 International Workshops, Frankfurt/Main, Germany, June 24 - 28, 2018, Revised Selected Papers, Workshop on Performance and Scalability of Storage Systems (WOPSSS)}, doi = {10.1007/978-3-030-02465-9_6}, pages = {89 -- 102}, year = {2018}, abstract = {Typical applications in data science consume, process and produce large amounts of data, making disk I/O one of the dominating — and thus worthwhile optimizing — factors of their overall performance. Distributed processing frameworks, such as Hadoop, Flink and Spark, hide a lot of complexity from the programmer when they parallelize these applications across a compute cluster. This exacerbates reasoning about I/O of both the application and the framework, through the distributed file system, such as HDFS, down to the local file systems. We present SFS (Statistics File System), a modular framework to trace each I/O request issued by the application and any JVM-based big data framework involved, mapping these requests to actual disk I/O. This allows detection of inefficient I/O patterns, both by the applications and the underlying frameworks, and builds the basis for improving I/O scheduling in the big data software stack.}, language = {en} } @misc{Steininger2018, type = {Master Thesis}, author = {Steininger, Lennart}, title = {Atomarer Datenzugriff auf verteilten persistenten Arbeitsspeicher}, year = {2018}, abstract = {Nicht fl{\"u}chtiger Arbeitsspeicher bzw. Non-volatile Random Access Memory oder auch Persistent Memory befindet sich im fr{\"u}hen Markteintritt. Diese Art von Speicher nimmt eine Stellung zwischen bisherigem fl{\"u}chtigen Arbeitsspeicher (Memory) und bisherigem Permanentspeicher (Storage) ein. Im Zusammenspiel mit verteilten Algorithmen und Systemen sowie dem Zugriff auf entfernte Speicherbereiche durch die Netzwerkinterfaces per Remote Direct Memory Access sind die Konzepte zum Umgang mit Durability und allen einhergehenden Herausforderungen bisher nicht ausgereift. Diese Arbeit untersucht, wie sich in Rechnernetzen mit persistentem Speicher und m{\"o}glichst passivem Zugriff auf entfernten Speicher performante verteilte Systeme implementieren lassen. Es werden Anforderungen und Voraussetzungen eines solchen Systems bestimmt und daraufhin L{\"o}sungsm{\"o}glichkeiten aufgezeigt, wie sich diese Anforderungen mit Nutzung nicht fl{\"u}chtigen Arbeitsspeichers umsetzen lassen. Die notwendigen Eigenschaften der Hard- und Software sowie der eingesetzten Netzwerkprotokolle werden beschrieben.}, language = {de} } @article{SeibertPetersSchintke2018, author = {Seibert, Felix and Peters, Mathias and Schintke, Florian}, title = {Improving I/O Performance Through Colocating Interrelated Input Data and Near-Optimal Load Balancing}, volume = {2018}, journal = {Proceedings of the IPDPSW; Fourth IEEE International Workshop on High-Performance Big Data, Deep Learning, and Cloud Computing (HPBDC)}, doi = {10.1109/IPDPSW.2018.00081}, pages = {448 -- 457}, year = {2018}, abstract = {Most distributed file systems assign new files to storage servers randomly. While working well in some situations, this does not help to optimize the input performance for most MapReduce computations' data access patterns. In this work, we consider an access pattern where input files are partitioned into groups of heterogeneous size. Each group is accessed by exactly one process. We design and implement a data placement strategy that places these file groups together on the same storage server. This colocation approach is combined with near-optimal storage load balancing. To do so, we use a classical scheduling approximation algorithm to solve the NP hard group assignment problem. We argue that local processing is not only beneficial because of reduced network traffic, but especially because it imposes an even resource schedule. Our experiments, based on the parallel processing of remote sensing images, reveal an enormous reduction of network traffic and up to 39 \% faster input read times. Further, simulations show that our approximate assignments limit storage server imbalances to less than 5 \% above the theoretical minimum, in contrast to more than 85 \% with random assignment.}, language = {en} } @inproceedings{NoackReinefeldKrameretal.2018, author = {Noack, Matthias and Reinefeld, Alexander and Kramer, Tobias and Steinke, Thomas}, title = {DM-HEOM: A Portable and Scalable Solver-Framework for the Hierarchical Equations of Motion}, booktitle = {2018 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), 19th IEEE Int. Workshop on Parallel and Distributed Scientific and Engineering Computing (PDSEC 2018)}, isbn = {978-1-5386-5555-9}, doi = {10.1109/IPDPSW.2018.00149}, pages = {947 -- 956}, year = {2018}, abstract = {Computing the Hierarchical Equations of Motion (HEOM) is by itself a challenging problem, and so is writing portable production code that runs efficiently on a variety of architectures while scaling from PCs to supercomputers. We combined both challenges to push the boundaries of simulating quantum systems, and to evaluate and improve methodologies for scientific software engineering. Our contributions are threefold: We present the first distributed memory implementation of the HEOM method (DM-HEOM), we describe an interdisciplinary development workflow, and we provide guidelines and experiences for designing distributed, performance-portable HPC applications with MPI-3, OpenCL and other state-of-the-art programming models. We evaluated the resulting code on multi- and many-core CPUs as well as GPUs, and demonstrate scalability on a Cray XC40 supercomputer for the PS I molecular light harvesting complex.}, language = {en} } @article{KramerNoackReinefeldetal.2018, author = {Kramer, Tobias and Noack, Matthias and Reinefeld, Alexander and Rodr{\´i}guez, Mirta and Zelinskyi, Yaroslav}, title = {Efficient calculation of open quantum system dynamics and time-resolved spectroscopy with Distributed Memory HEOM (DM-HEOM)}, volume = {39}, journal = {Journal of Computational Chemistry}, number = {22}, publisher = {Wiley Periodicals, Inc.}, arxiv = {http://arxiv.org/abs/arXiv:1803.03498}, doi = {doi:10.1002/jcc.25354}, pages = {1779 -- 1794}, year = {2018}, abstract = {Time- and frequency resolved optical signals provide insights into the properties of light harvesting molecular complexes, including excitation energies, dipole strengths and orientations, as well as in the exciton energy flow through the complex. The hierarchical equations of motion (HEOM) provide a unifying theory, which allows one to study the combined effects of system-environment dissipation and non-Markovian memory without making restrictive assumptions about weak or strong couplings or separability of vibrational and electronic degrees of freedom. With increasing system size the exact solution of the open quantum system dynamics requires memory and compute resources beyond a single compute node. To overcome this barrier, we developed a scalable variant of HEOM. Our distributed memory HEOM, DM-HEOM, is a universal tool for open quantum system dynamics. It is used to accurately compute all experimentally accessible time- and frequency resolved processes in light harvesting molecular complexes with arbitrary system-environment couplings for a wide range of temperatures and complex sizes.}, language = {en} } @article{KramerNoackReimersetal.2018, author = {Kramer, Tobias and Noack, Matthias and Reimers, Jeffrey R. and Reinefeld, Alexander and Rodr{\´i}guez, Mirta and Yin, Shiwei}, title = {Energy flow in the Photosystem I supercomplex: comparison of approximative theories with DM-HEOM}, volume = {515}, journal = {Chemical Physics}, publisher = {Elsevier B.V.}, arxiv = {http://arxiv.org/abs/arXiv:1805.10484}, doi = {10.1016/j.chemphys.2018.05.028}, pages = {262 -- 271}, year = {2018}, abstract = {We analyze the exciton dynamics in PhotosystemI from Thermosynechococcus elongatus using the distributed memory implementation of the hierarchical equation of motion (DM-HEOM) for the 96 Chlorophylls in the monomeric unit. The exciton-system parameters are taken from a first principles calculation. A comparison of the exact results with Foerster rates and Markovian approximations allows one to validate the exciton transfer times within the complex and to identify deviations from approximative theories. We show the optical absorption, linear, and circular dichroism spectra obtained with DM-HEOM and compare them to experimental results.}, language = {en} } @misc{Kammeyer2018, type = {Master Thesis}, author = {Kammeyer, Alexander}, title = {Radix sort for massively parallel hardware: a multi-GPU implementation}, pages = {67}, year = {2018}, abstract = {Computers, servers and supercomputers with more than one GPU become more and more common. To fully utilize such systems the application needs to be aware of the additional GPUs. E.g., computational workload needs to be distributed across the GPUs. Along with that goes data movement as well as communication. Recently, NVIDIA introduced new features, Unified Memory and Cooperative Groups, that enable programmers to extend their applications with multi-GPU capabilities. In this thesis the aforementioned features are presented in detail and evaluated in the context of the GPU molecular dynamics simulation HAL's MD package on the latest NVIDIA GPU architecture Pascal. A hotspot analysis with the NVIDIA Visual Profiler identified the radix sort as a major hotspot within HAL's MD package. This thesis presents an approach towards a multi-GPU implementation of radix sort. The individual kernels are analysed and an implementation is discussed. Then, the radix sort is transformed into a multi-GPU implementation kernel per kernel. For each step the individual runtime measurements and bottlenecks are presented to analyse the approach. A general sequence diagram of the implementation shows how the individual parts interact with each other. The outlook discusses the further development of HAL's MD package, especially the force computation for multi-GPU which has not been covered in this thesis.}, language = {en} } @article{KramerNoackBaumetal.2018, author = {Kramer, Tobias and Noack, Matthias and Baum, Daniel and Hege, Hans-Christian and Heller, Eric J.}, title = {Dust and gas emission from cometary nuclei: the case of comet 67P/Churyumov-Gerasimenko}, volume = {3}, journal = {Advances in Physics: X}, number = {1}, doi = {10.1080/23746149.2017.1404436}, pages = {1404436}, year = {2018}, abstract = {Comets display with decreasing solar distance an increased emission of gas and dust particles, leading to the formation of the coma and tail. Spacecraft missions provide insight in the temporal and spatial variations of the dust and gas sources located on the cometary nucleus. For the case of comet 67P/Churyumov-Gerasimenko (67P/C-G), the long-term obser- vations from the Rosetta mission point to a homogeneous dust emission across the entire illuminated surface. Despite the homogeneous initial dis- tribution, a collimation in jet-like structures becomes visible. We propose that this observation is linked directly to the complex shape of the nucleus and projects concave topographical features into the dust coma. To test this hypothesis, we put forward a gas-dust description of 67P/C-G, where gravitational and gas forces are accurately determined from the surface mesh and the rotation of the nucleus is fully incorporated. The emerging jet-like structures persist for a wide range of gas-dust interactions and show a dust velocity dependent bending.}, language = {en} } @article{SchmidtkeErleben2018, author = {Schmidtke, Robert and Erleben, Kenny}, title = {Chunked Bounding Volume Hierarchies for Fast Digital Prototyping using Volumetric Meshes}, volume = {24}, journal = {IEEE Transactions on Visualization and Computer Graphics}, number = {12}, issn = {1941-0506}, doi = {10.1109/TVCG.2017.2784441}, pages = {3044 -- 3057}, year = {2018}, abstract = {We present a novel approach to using Bounding Volume Hierarchies (BVHs) for collision detection of volumetric meshes for digital prototyping based on accurate simulation. In general, volumetric meshes contain more primitives than surface meshes, which in turn means larger BVHs. To manage these larger BVHs, we propose an algorithm for splitting meshes into smaller chunks with a limited-size BVH each. Limited-height BVHs make guided, all-pairs testing of two chunked meshes well-suited for GPU implementation. This is because the dynamically generated work during BVH traversal becomes bounded. Chunking is simple to implement compared to dynamic load balancing methods and can result in an overall two orders of magnitude speedup on GPUs. This indicates that dynamic load balancing may not be a well suited scheme for the GPU. The overall application timings showed that data transfers were not the bottleneck. Instead, the conversion to and from OpenCL friendly data structures was causing serious performance impediments. Still, a simple OpenMP acceleration of the conversion allowed the GPU solution to beat the CPU solution by 20\%. We demonstrate our results using rigid and deformable body scenes of varying complexities on a variety of GPUs.}, language = {en} } @inproceedings{LaeuterKramerRubinetal.2018, author = {L{\"a}uter, Matthias and Kramer, Tobias and Rubin, Martin and Altwegg, Kathrin}, title = {Gas production of comet 67P/Churyumov-Gerasimenko reconstructed from DFMS/COPS data}, volume = {12}, booktitle = {Europlanet Science Congress}, publisher = {Europlanet Science Congress}, pages = {EPSC2018-515-1}, year = {2018}, abstract = {We reconstruct the temporal evolution of surface emissions for the four major gas species H2O, CO2, CO, and O2 emitted during the 2015 apparition of comet 67P/Churyumov-Gerasimenko (67P/C-G). Measurements from the Double Focusing Mass Spectrometer (DFMS) of the Rosetta Orbiter Spectrometer for Ion and Neutral Analysis (ROSINA) and the COmet Pressure Sensor (COPS) are used to determine the gas sources on the surface with an inverse gas model for the entire coma. For all species, peak production rates and integrated production rates per orbit are evaluated separately for the northern and the southern hemisphere. Complemented with the total mass production, this allows us to estimate the dust-to-gas ratio of the emitted material.}, language = {en} } @inproceedings{KramerLaeuter2018, author = {Kramer, Tobias and L{\"a}uter, Matthias}, title = {Modelling the inner coma of comet 67P/Churyumov-Gerasimenko}, volume = {12}, booktitle = {Europlanet Science Congress}, publisher = {Europlanet Science Congress}, pages = {EPSC2018-115-1}, year = {2018}, abstract = {Based on about 1 million of pressure measurements around comet 67P/Churyumov-Gerasimenko we reconstruct the gas emission across the entire nucleus. Dust particles are seeded in the gas model and the resulting dust distribution follows a daily pattern which agrees with observations if a uniform dust release across the entire sunlit surface is assumed.}, language = {en} }