TY - JOUR
AB - The rise of exascale supercomputers has fueled competition among GPU vendors, driving lattice QCD developers to write code that supports multiple APIs. Moreover, new developments in algorithms and physics research require frequent updates to existing software. These challenges have to be balanced against constantly changing personnel. At the same time, there is a wide range of applications for HISQ fermions in QCD studies. This situation encourages the development of software featuring a HISQ action that is flexible, high-performing, open source, easy to use, and easy to adapt. In this technical paper, we explain the design strategy, provide implementation details, list available algorithms and modules, and show key performance indicators for SIMULATeQCD, a simple multi-GPU lattice code for large-scale QCD calculations, mainly developed and used by the HotQCD collaboration. The code is publicly available on GitHub.
AU - Mazur, Lukas
AU - Bollweg, Dennis
AU - Clarke, David A.
AU - Altenkort, Luis
AU - Kaczmarek, Olaf
AU - Larsen, Rasmus
AU - Shu, Hai-Tao
AU - Goswami, Jishnu
AU - Scior, Philipp
AU - Sandmeyer, Hauke
AU - Neumann, Marius
AU - Dick, Henrik
AU - Ali, Sajid
AU - Kim, Jangho
AU - Schmidt, Christian
AU - Petreczky, Peter
AU - Mukherjee, Swagato
ID - 46120
JF - Computer Physics Communications
TI - SIMULATeQCD: A simple multi-GPU lattice code for QCD calculations
ER -
TY - JOUR
AU - Altenkort, Luis
AU - Eller, Alexander M.
AU - Francis, Anthony
AU - Kaczmarek, Olaf
AU - Mazur, Lukas
AU - Moore, Guy D.
AU - Shu, Hai-Tao
ID - 46119
IS - 1
JF - Physical Review D
SN - 2470-0010
TI - Viscosity of pure-glue QCD from the lattice
VL - 108
ER -
TY - JOUR
AB - While FPGA accelerator boards and their respective high-level design tools are maturing, there is still a lack of multi-FPGA applications, libraries, and not least, benchmarks and reference implementations towards sustained HPC usage of these devices. As in the early days of GPUs in HPC, for workloads that can reasonably be decoupled into loosely coupled working sets, multi-accelerator support can be achieved by using standard communication interfaces like MPI on the host side. However, for performance and productivity, some applications can profit from a tighter coupling of the accelerators. FPGAs offer unique opportunities here when extending the dataflow characteristics to their communication interfaces.
In this work, we extend the HPCC FPGA benchmark suite by multi-FPGA support and three missing benchmarks that particularly characterize or stress inter-device communication: b_eff, PTRANS, and LINPACK. With all benchmarks implemented for current boards with Intel and Xilinx FPGAs, we established a baseline for multi-FPGA performance. Additionally, for the communication-centric benchmarks, we explored the potential of direct FPGA-to-FPGA communication with a circuit-switched inter-FPGA network that is currently only available for one of the boards. The evaluation with parallel execution on up to 26 FPGA boards makes use of one of the largest academic FPGA installations.
AU - Meyer, Marius
AU - Kenter, Tobias
AU - Plessl, Christian
ID - 38041
JF - ACM Transactions on Reconfigurable Technology and Systems
KW - General Computer Science
SN - 1936-7406
TI - Multi-FPGA Designs and Scaling of HPC Challenge Benchmarks via MPI and Circuit-Switched Inter-FPGA Networks
ER -
TY - CHAP
AU - Hansmeier, Tim
AU - Kenter, Tobias
AU - Meyer, Marius
AU - Riebler, Heinrich
AU - Platzner, Marco
AU - Plessl, Christian
ED - Haake, Claus-Jochen
ED - Meyer auf der Heide, Friedhelm
ED - Platzner, Marco
ED - Wachsmuth, Henning
ED - Wehrheim, Heike
ID - 45893
T2 - On-The-Fly Computing -- Individualized IT-services in dynamic markets
TI - Compute Centers I: Heterogeneous Execution Environments
VL - 412
ER -
TY - CONF
AU - Opdenhövel, Jan-Oliver
AU - Plessl, Christian
AU - Kenter, Tobias
ID - 46190
T2 - Proceedings of the 13th International Symposium on Highly Efficient Accelerators and Reconfigurable Technologies
TI - Mutation Tree Reconstruction of Tumor Cells on FPGAs Using a Bit-Level Matrix Representation
ER -
TY - CONF
AU - Faj, Jennifer
AU - Kenter, Tobias
AU - Faghih-Naini, Sara
AU - Plessl, Christian
AU - Aizinger, Vadym
ID - 46188
T2 - Proceedings of the Platform for Advanced Scientific Computing Conference
TI - Scalable Multi-FPGA Design of a Discontinuous Galerkin Shallow-Water Model on Unstructured Meshes
ER -
TY - CONF
AU - Prouveur, Charles
AU - Haefele, Matthieu
AU - Kenter, Tobias
AU - Voss, Nils
ID - 46189
T2 - Proceedings of the Platform for Advanced Scientific Computing Conference
TI - FPGA Acceleration for HPC Supercapacitor Simulations
ER -
TY - CONF
AB - The computation of electron repulsion integrals (ERIs) over Gaussian-type orbitals (GTOs) is a challenging problem in quantum-mechanics-based atomistic simulations. In practical simulations, several trillions of ERIs may have to be
computed for every time step.
In this work, we investigate FPGAs as accelerators for the ERI computation. We use template parameters, here within the Intel oneAPI tool flow, to create customized designs for 256 different ERI quartet classes, based on their orbitals. To maximize data reuse, all intermediates are buffered in FPGA on-chip memory with customized layout. The pre-calculation of intermediates also helps to overcome data dependencies caused by multi-dimensional recurrence
relations. The involved loop structures are partially or even fully unrolled for high throughput of FPGA kernels. Furthermore, a lossy compression algorithm utilizing arbitrary bitwidth integers is integrated in the FPGA kernels. To our
best knowledge, this is the first work on ERI computation on FPGAs that supports more than just the single most basic quartet class. Also, the integration of ERI computation and compression it a novelty that is not even covered by CPU or GPU libraries so far.
Our evaluation shows that using 16-bit integer for the ERI compression, the fastest FPGA kernels exceed the performance of 10 GERIS ($10 \times 10^9$ ERIs per second) on one Intel Stratix 10 GX 2800 FPGA, with maximum absolute errors around $10^{-7}$ - $10^{-5}$ Hartree. The measured throughput can be accurately explained by a performance model. The FPGA kernels deployed on 2 FPGAs outperform similar computations using the widely used libint reference on a two-socket server with 40 Xeon Gold 6148 CPU cores of the same process technology by factors up to 6.0x and on a new two-socket server with 128 EPYC 7713 CPU cores by up to 1.9x.
AU - Wu, Xin
AU - Kenter, Tobias
AU - Schade, Robert
AU - Kühne, Thomas
AU - Plessl, Christian
ID - 43228
T2 - 2023 IEEE 31st Annual International Symposium on Field-Programmable Custom Computing Machines (FCCM)
TI - Computing and Compressing Electron Repulsion Integrals on FPGAs
ER -
TY - JOUR
AB - The non-orthogonal local submatrix method applied to electronic structure–based molecular dynamics simulations is shown to exceed 1.1 EFLOP/s in FP16/FP32-mixed floating-point arithmetic when using 4400 NVIDIA A100 GPUs of the Perlmutter system. This is enabled by a modification of the original method that pushes the sustained fraction of the peak performance to about 80%. Example calculations are performed for SARS-CoV-2 spike proteins with up to 83 million atoms.
AU - Schade, Robert
AU - Kenter, Tobias
AU - Elgabarty, Hossam
AU - Lass, Michael
AU - Kühne, Thomas
AU - Plessl, Christian
ID - 45361
JF - The International Journal of High Performance Computing Applications
KW - Hardware and Architecture
KW - Theoretical Computer Science
KW - Software
SN - 1094-3420
TI - Breaking the exascale barrier for the electronic structure problem in ab-initio molecular dynamics
ER -
TY - GEN
AB - Viscous hydrodynamics serves as a successful mesoscopic description of the
Quark-Gluon Plasma produced in relativistic heavy-ion collisions. In order to
investigate, how such an effective description emerges from the underlying
microscopic dynamics we calculate the hydrodynamic and non-hydrodynamic modes
of linear response in the sound channel from a first-principle calculation in
kinetic theory. We do this with a new approach wherein we discretize the
collision kernel to directly calculate eigenvalues and eigenmodes of the
evolution operator. This allows us to study the Green's functions at any point
in the complex frequency space. Our study focuses on scalar theory with quartic
interaction and we find that the analytic structure of Green's functions in the
complex plane is far more complicated than just poles or cuts which is a first
step towards an equivalent study in QCD kinetic theory.
AU - Ochsenfeld, Stephan
AU - Schlichting, Sören
ID - 50172
T2 - arXiv:2308.04491
TI - Hydrodynamic and Non-hydrodynamic Excitations in Kinetic Theory -- A Numerical Analysis in Scalar Field Theory
ER -
TY - GEN
AB - Memory Gym presents a suite of 2D partially observable environments, namely
Mortar Mayhem, Mystery Path, and Searing Spotlights, designed to benchmark
memory capabilities in decision-making agents. These environments, originally
with finite tasks, are expanded into innovative, endless formats, mirroring the
escalating challenges of cumulative memory games such as ``I packed my bag''.
This progression in task design shifts the focus from merely assessing sample
efficiency to also probing the levels of memory effectiveness in dynamic,
prolonged scenarios. To address the gap in available memory-based Deep
Reinforcement Learning baselines, we introduce an implementation that
integrates Transformer-XL (TrXL) with Proximal Policy Optimization. This
approach utilizes TrXL as a form of episodic memory, employing a sliding window
technique. Our comparative study between the Gated Recurrent Unit (GRU) and
TrXL reveals varied performances across different settings. TrXL, on the finite
environments, demonstrates superior sample efficiency in Mystery Path and
outperforms in Mortar Mayhem. However, GRU is more efficient on Searing
Spotlights. Most notably, in all endless tasks, GRU makes a remarkable
resurgence, consistently outperforming TrXL by significant margins. Website and
Source Code: https://github.com/MarcoMeter/endless-memory-gym/
AU - Pleines, Marco
AU - Pallasch, Matthias
AU - Zimmer, Frank
AU - Preuss, Mike
ID - 50221
T2 - arXiv:2309.17207
TI - Memory Gym: Towards Endless Tasks to Benchmark Memory Capabilities of Agents
ER -
TY - CHAP
AU - Alt, Christoph
AU - Kenter, Tobias
AU - Faghih-Naini, Sara
AU - Faj, Jennifer
AU - Opdenhövel, Jan-Oliver
AU - Plessl, Christian
AU - Aizinger, Vadym
AU - Hönig, Jan
AU - Köstler, Harald
ID - 46191
SN - 0302-9743
T2 - Lecture Notes in Computer Science
TI - Shallow Water DG Simulations on FPGAs: Design and Comparison of a Novel Code Generation Pipeline
ER -
TY - GEN
AB - This preprint makes the claim of having computed the $9^{th}$ Dedekind
Number. This was done by building an efficient FPGA Accelerator for the core
operation of the process, and parallelizing it on the Noctua 2 Supercluster at
Paderborn University. The resulting value is
286386577668298411128469151667598498812366. This value can be verified in two
steps. We have made the data file containing the 490M results available, each
of which can be verified separately on CPU, and the whole file sums to our
proposed value.
AU - Van Hirtum, Lennart
AU - De Causmaecker, Patrick
AU - Goemaere, Jens
AU - Kenter, Tobias
AU - Riebler, Heinrich
AU - Lass, Michael
AU - Plessl, Christian
ID - 43439
T2 - arXiv:2304.03039
TI - A computation of D(9) using FPGA Supercomputing
ER -
TY - GEN
AB - We investigate the early time development of the anisotropic transverse flow
and spatial eccentricities of a fireball with various particle-based transport
approaches using a fixed initial condition. In numerical simulations ranging
from the quasi-collisionless case to the hydrodynamic regime, we find that the
onset of $v_n$ and of related measures of anisotropic flow can be described
with a simple power-law ansatz, with an exponent that depends on the amount of
rescatterings in the system. In the few-rescatterings regime we perform
semi-analytical calculations, based on a systematic expansion in powers of time
and the cross section, which can reproduce the numerical findings.
AU - Borghini, Nicolas
AU - Borrell, Marc
AU - Roch, Hendrik
ID - 32177
T2 - arXiv:2201.13294
TI - Early time behavior of spatial and momentum anisotropies in kinetic theory across different Knudsen numbers
ER -
TY - GEN
AB - We test the ability of the "escape mechanism" to create the anisotropic flow
observed in high-energy nuclear collisions. We compare the flow harmonics $v_n$
in the few-rescatterings regime from two types of transport simulations, with
$2\to 2$ and $2\to 0$ collision kernels respectively, and from analytical
calculations neglecting the gain term of the Boltzmann equation. We find that
the even flow harmonics are similar in the three approaches, while the odd
harmonics differ significantly.
AU - Bachmann, Benedikt
AU - Borghini, Nicolas
AU - Feld, Nina
AU - Roch, Hendrik
ID - 32178
T2 - arXiv:2203.13306
TI - Even anisotropic-flow harmonics are from Venus, odd ones are from Mars
ER -
TY - JOUR
AU - Hou, W
AU - Yao, Y
AU - Li, Y
AU - Peng, B
AU - Shi, K
AU - Zhou, Z
AU - Pan, J
AU - Liu, M
AU - Hu, J
ID - 32183
IS - 1
JF - Frontiers of materials science
SN - 2095-025x
TI - Linearly shifting ferromagnetic resonance response of La0.7Sr0.3MnO3 thin film for body temperature sensors
VL - 16
ER -
TY - JOUR
AU - Wojciechowski, M
ID - 32234
JF - Data Brief
SN - 2352-3409
TI - Dataset for random uniform distributions of 2D circles and 3D spheres.
VL - 43
ER -
TY - THES
AU - Lass, Michael
ID - 32414
TI - Bringing Massive Parallelism and Hardware Acceleration to Linear Scaling Density Functional Theory Through Targeted Approximations
ER -
TY - GEN
AB - The Julia programming language has evolved into a modern alternative to fill existing gaps in scientific computing and data science applications. Julia leverages a unified and coordinated single-language and ecosystem paradigm and has a proven track record of achieving high performance without sacrificing user productivity. These aspects make Julia a viable alternative to high-performance computing's (HPC's) existing and increasingly costly many-body workflow composition strategy in which traditional HPC languages (e.g., Fortran, C, C++) are used for simulations, and higher-level languages (e.g., Python, R, MATLAB) are used for data analysis and interactive computing. Julia's rapid growth in language capabilities, package ecosystem, and community make it a promising universal language for HPC. This paper presents the views of a multidisciplinary group of researchers from academia, government, and industry that advocate for an HPC software development paradigm that emphasizes developer productivity, workflow portability, and low barriers for entry. We believe that the Julia programming language, its ecosystem, and its community provide modern and powerful capabilities that enable this group's objectives. Crucially, we believe that Julia can provide a feasible and less costly approach to programming scientific applications and workflows that target HPC facilities. In this work, we examine the current practice and role of Julia as a common, end-to-end programming model to address major challenges in scientific reproducibility, data-driven AI/machine learning, co-design and workflows, scalability and performance portability in heterogeneous computing, network communication, data management, and community education. As a result, the diversification of current investments to fulfill the needs of the upcoming decade is crucial as more supercomputing centers prepare for the exascale era.
AU - Churavy, Valentin
AU - Godoy, William F
AU - Bauer, Carsten
AU - Ranocha, Hendrik
AU - Schlottke-Lakemper, Michael
AU - Räss, Ludovic
AU - Blaschke, Johannes
AU - Giordano, Mosè
AU - Schnetter, Erik
AU - Omlin, Samuel
AU - Vetter, Jeffrey S
AU - Edelman, Alan
ID - 36879
TI - Bridging HPC Communities through the Julia Programming Language
ER -
TY - JOUR
AB - AbstractTailored nanoscale quantum light sources, matching the specific needs of use cases, are crucial building blocks for photonic quantum technologies. Several different approaches to realize solid-state quantum emitters with high performance have been pursued and different concepts for energy tuning have been established. However, the properties of the emitted photons are always defined by the individual quantum emitter and can therefore not be controlled with full flexibility. Here we introduce an all-optical nonlinear method to tailor and control the single photon emission. We demonstrate a laser-controlled down-conversion process from an excited state of a semiconductor quantum three-level system. Based on this concept, we realize energy tuning and polarization control of the single photon emission with a control-laser field. Our results mark an important step towards tailored single photon emission from a photonic quantum system based on quantum optical principles.
AU - Jonas, B.
AU - Heinze, Dirk Florian
AU - Schöll, E.
AU - Kallert, P.
AU - Langer, T.
AU - Krehs, S.
AU - Widhalm, A.
AU - Jöns, Klaus
AU - Reuter, Dirk
AU - Schumacher, Stefan
AU - Zrenner, Artur
ID - 40523
IS - 1
JF - Nature Communications
KW - General Physics and Astronomy
KW - General Biochemistry
KW - Genetics and Molecular Biology
KW - General Chemistry
KW - Multidisciplinary
SN - 2041-1723
TI - Nonlinear down-conversion in a single quantum dot
VL - 13
ER -