publications

2024

1. RadicalPy: A Tool for Spin Dynamics Simulations

   Lewis M. Antill, and Emil Vatai

   Journal of Chemical Theory and Computation, Oct 2024

   Abs DOI

   Radical pairs (electron–hole pairs, polaron pairs) are transient reaction intermediates that are found and exploited in all areas of science, from the hard realm of physics in the form of organic semiconductors, spintronics, quantum computing, and solar cells to the soft domain of chemistry and biology under the guise of chemical reactions in solution, biomimetic systems, and quantum biology. Quantitative analysis of radical pair phenomena has historically been successful by a few select groups. With this in mind, we present an intuitive open-source framework in the Python programming language that provides classical, semiclassical, and quantum simulation methodologies. A radical pair kinetic rate equation solver, Monte Carlo-based spin dephasing rate estimations, and molecule database functionalities are implemented. We introduce the kine-quantum method, a new approach that amalgamates classical rate equations, semiclassical, and quantum techniques. This method resolves the prohibitively large memory requirement issues of quantum approaches while achieving higher accuracy, and it also offers wavelength-resolved simulations, producing time- and wavelength-resolved magnetic field effect simulations. Model examples illustrate the versatility and ease of use of the software, including the new approach applied to the magnetosensitive absorption and fluorescence of flavin adenine dinucleotide photochemistry, spin–spin interaction estimation from molecular dynamics simulations on radical pairs inside reverse micelles, radical pair anisotropy inside proteins, and triplet exciton pairs in anthracene crystals. The intuitive interface also allows this software to be used as a teaching or learning aid for those interested in the field of spin chemistry. Furthermore, the software aims to be modular and extensible, with the aim to standardize how spin dynamics simulations are performed.

2. POSTER: ORCHA: A Performance Portability System For Flash-X — A Multiphysics Application Software

   Youngjun Lee, Wesley Kwiecinski, Klaus Weide, and 6 more authors

   In , Oct 2024
3. On the Scalability of Computing Genomic Diversity Using SparkLeBLAST: A Feasibility Study

   Ritvik Prabhu, Bernard Moussad, Karim Youssef, and 2 more authors

   In 28th Annual IEEE High Performance Extreme Computing Virtual Conference, Oct 2024

   Outstanding Student Paper Award

   Abs

   Studying the genomic diversity of viruses can help us understand how viruses evolve and how that evolution can impact human health. Rather than use a laborious and tedious wet-lab approach to conduct a genomic diversity study, we take a computational approach, using the ubiquitous NCBI BLAST and our parallel and distributed SparkLeBLAST, across 53 patients (∼40,000,000 query sequences) on Fugaku, the world’s fastest homogeneous supercomputer with 158,976 nodes, where each code contains a 48-core A64FX processor and 32 GB RAM.

4. POSTER: Evaluation of Vectorization Methods on Arm SVE Using the Exo Language

   Rin Iwai,  Emil Vatai,  Jens Domke, and 1 more author

   In IEEE Cluster 2024, Oct 2024
5. POSTER: Quantifying the Effects of Copious 3D-Stacked Cache on HPC Workloads

   Emil Vatai, Jens Domke, Balazs Gerofi, and 9 more authors

   In IEEE Symposium on Low-Power and High-Speed Chips and Systems (COOL Chips 26), Oct 2024

6. 

   Tadashi: Enabling AI-Based Automated Code Generation With Guaranteed Correctness

   Emil Vatai, Aleksandr Drozd, Ivan R. Ivanov, and 2 more authors

   Oct 2024

   Abs DOI

   Frameworks and DSLs auto-generating code have traditionally relied on human experts developing them to have in place rigorous methods to assure the legality of the applied code transformations. Machine Learning (ML) is gaining wider adoption as a means to auto-generate code optimised for the hardware target. However, ML solutions, and in particular black-box DNNs, provide no such guarantees on legality. In this paper we propose a library, Tadashi, which leverages the polyhedral model to empower researchers seeking to curate datasets crucial for applying ML in code-generation. Tadashi provides the ability to reliably and practically check the legality of candidate transformations on polyhedral schedules applied on a baseline reference code. We provide a proof that our library guarantees the legality of generated transformations, and demonstrate its lightweight practical cost. Tadashi is available at https://github.com/vatai/tadashi/.

2023

1. At the Locus of Performance: Quantifying the Effects of Copious 3D-Stacked Cache on HPC Workloads

   Jens Domke, Emil Vatai, Balazs Gerofi, and 9 more authors

   ACM Trans. Archit. Code Optim., Dec 2023

   Abs DOI

   Over the last three decades, innovations in the memory subsystem were primarily targeted at overcoming the data movement bottleneck. In this paper, we focus on a specific market trend in memory technology: 3D-stacked memory and caches. We investigate the impact of extending the on-chip memory capabilities in future HPC-focused processors, particularly by 3D-stacked SRAM. First, we propose a method oblivious to the memory subsystem to gauge the upper-bound in performance improvements when data movement costs are eliminated. Then, using the gem5 simulator, we model two variants of a hypothetical LARge Cache processor (LARC), fabricated in 1.5 nm and enriched with high-capacity 3D-stacked cache. With a volume of experiments involving a broad set of proxy-applications and benchmarks, we aim to reveal how HPC CPU performance will evolve, and conclude an average boost of 9.56\texttimes for cache-sensitive HPC applications, on a per-chip basis. Additionally, we exhaustively document our methodological exploration to motivate HPC centers to drive their own technological agenda through enhanced co-design.

2. A Tool and a Methodology to Use Macros for Abstracting Variations in Code for Different Computational Demands

   Anshu Dubey, Youngjun Lee, Tom Klosterman, and 1 more author

   Future Generation Computer Systems, Jul 2023

   Abs DOI

   Scientific software used on high-performance computing platforms is in a phase of transformation because of the combined increase in the heterogeneity and complexity of models and hardware platforms. Having separate implementations for different platforms can easily lead to combinatorial explosions; therefore, the computational science community has been looking for mechanisms to express code through abstractions that can be specialized for different platforms. Most existing approaches use template meta-programming in C++, and are, therefore language specific. We have developed a tool that uses customized expansion of macros to mimic some of C++ behaviour in other languages. It enables unification of any code variants that may be necessary to run efficiently on different target architectures and different computational environments through use of macros with multiple alternative definitions and ability to arbitrate on definition selection for expansion. Combined with two other tools, a custom runtime, and a user specified recipe translator, our custom macroprocessor becomes a part of an overall performance portability solution that does not depend on any specific programming language. We also use macros as code-shorthand that lets code snippets become building blocks that allow variations in control flow to explore performance options. We demonstrate use of macros in Flash-X, a multiphysics multicomponent code with many Fortran legacy components derived from an earlier community code FLASH.

2022

1. Why Globally Re-shuffle? Revisiting Data Shuffling in Large Scale Deep Learning

   Truong Thao Nguyen, François Trahay, Jens Domke, and 5 more authors

   In 2022 IEEE International Parallel and Distributed Processing Symposium (IPDPS), May 2022

   Abs DOI

   Stochastic gradient descent (SGD) is the most prevalent algorithm for training Deep Neural Networks (DNN). SGD iterates the input data set in each training epoch processing data samples in a random access fashion. Because this puts enormous pressure on the I/O subsystem, the most common approach to distributed SGD in HPC environments is to replicate the entire dataset to node local SSDs. However, due to rapidly growing data set sizes this approach has become increasingly infeasible. Surprisingly, the questions of why and to what extent random access is required have not received a lot of attention in the literature from an empirical standpoint. In this paper, we revisit data shuffling in DL workloads to investigate the viability of partitioning the dataset among workers and performing only a partial distributed exchange of samples in each training epoch. Through extensive experiments on up to 2,048 GPUs of ABCI and 4,096 compute nodes of Fugaku, we demonstrate that in practice validation accuracy of global shuffling can be maintained when carefully tuning the partial distributed exchange. We provide a solution implemented in PyTorch that enables users to control the proposed data exchange scheme.

2021

1. Matrix Engines for High Performance Computing: A Paragon of Performance or Grasping at Straws?

   Jens Domke, Emil Vatai, Aleksandr Drozd, and 8 more authors

   In 2021 IEEE International Parallel and Distributed Processing Symposium (IPDPS), May 2021

   Abs DOI

   Matrix engines or units, in different forms and affinities, are becoming a reality in modern processors; CPUs and otherwise. The current and dominant algorithmic approach to Deep Learning merits the commercial investments in these units, and deduced from the No. 1 benchmark in supercomputing, namely High Performance Linpack, one would expect an awakened enthusiasm by the HPC community, too. Hence, our goal is to identify the practical added benefits for HPC and machine learning applications by having access to matrix engines. For this purpose, we perform an in-depth survey of software stacks, proxy applications and benchmarks, and historical batch job records. We provide a cost-benefit analysis of matrix engines, both asymptotically and in conjunction with state-of-the-art processors. While our empirical data will temper the enthusiasm, we also outline opportunities to “misuse” these dense matrix-multiplication engines if they come for free.

2020

1. Diamond Matrix Powers Kernels

   Emil Vatai, Utsav Singhal, and Reiji Suda

   In Proceedings of the International Conference on High Performance Computing in Asia-Pacific Region, Jan 2020

   Abs DOI

   Matrix powers kernel calculates the vectors Akv, for k = 1, 2,..., m and they are the heart of various scientific computations, including communication avoiding iterative solvers. In this paper we propose diamond matrix powers kernel - DMPK, which has the purpose to apply the "diamond tiling" stencil algorithm to general matrices. It can also be considered as an extension of the PA1 and PA2 algorithms, introduced by Demmel et al. Our approach enables us to control the balance between the amount of communication avoidance and redundant computation inherently present in communication avoiding algorithms. We present a proof of concept implementation of the algorithm using MPI routines. The experiments we performed show that the control of the amount of computation and communication is achievable, and with more thorough optimisations, DMPK is a promising alternative to existing MPK approaches.

2019

1. POSTER: Communication Avoiding and Memory Hierarchy in Large Computations

   Emil Vatai, Utsav Singhal, and Suda Reiji

   In 4th International Symposium on Research and Education of Computational Science (RECS), Jan 2019

2014

1. The Largest Known Cunningham Chain of Length 3 of the First Kind

   Gabor Farkas, Gabor E Gevay, Antal Jarai, and 1 more author

   Studia Univ. Babes-Bolyai Mathematica, Jan 2014

   Abs

   Cunningham chains of length n of the first kind are n long sequences of prime numbers p1, p2, . . . , pn so that pi+1 = 2pi + 1 (for 1 ≤ i < n). In [3] we have devised a plan to find large Cunningham chains of the first kind of length 3 where the primes are of the form pi+1 = (h0 + cx) ⋅ 2e+i - 1 for some integer x with h0 = 5 775, c = 30 030 and e = 34 944. The project was executed on the non-uniform memory access (NUMA) supercomputer of NIIF in Pe\textasciiacutecs, Hungary. In this paper we report on the obtained results and discuss the implementation details. The search consisted of two stages: sieving and the Fermat test. The sieving stage was implemented in a concurrent manner using lockfree queues, while the Fermat test was trivially parallel. On the 27th of April, 2014 we have found the largest known Cunningham chain of length 3 of the first kind which consists of the numbers 5110664609396115 ⋅ 234944+j - 1 for j = 0, 1, 2.

2013

1. Inverse Sieve

   Emil Vatai

   Annales Universitatis Scientiarum Budapestinensis de Rolando Eötvös nominatae Sectio Computatorica, Jan 2013

2011

1. Cache Optimized Linear Sieve

   Antal Járai, and Emil Vatai

   Acta Univ. Sapientiae, Inform., Jan 2011

2010

1. Cache Optimized Sieve

   Antal Járai, and Emil Vatai

   In 8th Joint Conference on Math and Computer Science MaCS 2010, Jan 2010
2. POSTER: Nagyharékonyságú Számítások: Eratosthenészi Szita

   Emil Vatai

   In Hallgatók a Kutatásban, Tudományos Előadássorozat, Jan 2010