Maciej Pawlik, Klemens Noga, Maciej Czuchry, Jacek Budzowski, Łukasz Flis, Patryk Lasoń, Marek Magryś, Michał Sterzel

ACC Cyfronet UST, Krakow, Poland

Title: Evaluation of ARM based system for HPC workloads, a case study.

Abstract:ARM is a well known CPU architecture, up to this point, used primarily in portable devices. Application of ARM CPUs was mostly motivated by features which include low power consumption and flexible licensing model. Over the recent years ARM has grown to feature a mature software ecosystem and performance comparable to CPUs used in stationary devices like workstations and servers. According to the community, ARM has potential to address some of challenges which arose in the field of High Performance Computing. In the current day HPC, vendors and researchers alike are working on lowering power consumption, increasing performance per dollar and increasing density. This paper presents experience from a case study, which was done at ACC Cyfronet UST, where we evaluated ARM based system designed to provide computing services. The evaluation consisted of integrating the test system with a running HPC cluster and subjecting it to typical workloads. Integration allowed us to test system’s compatibility with supporting services like storage and networking. Our experiences demonstrate that storage based on Lustre filesystem needs substantial changes in configuration to work with ARM servers properly. The computing performance was measured and proved to be comparable to that offered by most popular vendors. While we experienced some difficulties, it was shown that it is possible to build and run an HPC cluster based on ARM architecture.

Evaluation of ARM based system for HPC workloads, a case study (paper)

Karthee Sivalingam, Alfio Lazzaro, Nina Mujkanovic

HPE HPC/AI EMEA Research Lab

Title: Optimising AI training deployments using Graph compilers and containers

Abstract: AI applications based on Deep Neural Networks (DNN) have become popular in solving nontrivial problems like image analysis and speech recognition. An AI workload usually incorporates an Extract, Transform and Load (ETL) pipeline, data movement, and execution of DNN graphs.Deep learning (DL) models are usually represented as computational graphs, with nodes representing tensor operators, and edges the data dependencies between them. AI training deployments can be optimised with target-specific libraries, Graph compilers, and by improving data movement. Graph compilers aim to optimise the execution of a DNN graph by generating an optimised code for a target hardware/backend, thus accelerating training and deployment of DNN models. Heterogenous Cloud and HPC infrastructures further increase the complexity of deploying and optimising AI training workloads.
In the SODALITE (a Horizon 2020 project), we address this problem by providing tools to enable simpler and faster development, deployment, operation, and execution of applications in heterogenous HPC and cloud computing environments. As part of this project, we are developing an Application Optimiser component that uses a performance model of infrastructure and applications (based on benchmarks) to optimise its deployment and runtime for heterogenous infrastructure and hardware. Using input from a data scientist which defines the configurations and optimisations to be enabled, the Application Optimiser component will select the framework, graph compiler, and target-specific libraries before building an optimised container.
In this talk, we present a review of different AI frameworks and supported graph compilers. We also compare the performance of different frameworks and graph compilers using standard benchmarks when deployed using containers. We describe how AI training deployments in heterogenous targets will be optimised by the Application Optimiser.

Uros Ignjacevic

General Manager of Sun Data World

Title: Ethiopia: The formation of a digital powerhouse of Africa

Abstract: With abundant renewable energy sources, including Africa’s largest hydroelectric $4.5 Billion, Grand Ethiopian Renaissance Dam, currently under construction, Ethiopia is emerging as a global player in providing highly competitive and reliable HPC services. Using abundant green energy Ethiopia has the potential to create a zero net carbon footprint while building its economy on the pillars of information technology and using advanced high speed computing for better and faster data analysis and research, creating prediction models for agriculture, an improved health system, the development of science and more. Join me in a presentation on Ethiopia, the land of origins, and its future in advanced computing.

Chin Fang

CEO of Zettar Inc.

Title: The historical 1st Poland-Singapore data transfer production trial over CAE-1, a behind the scene look

Abstract: From mid-October to early November 2019, ICM, Warsaw, Poland, ASTAR Computational Resource Centre (A*CRC), Singapore, and Zettar Inc. jointly carried out a historical 1st data transfer production trial from ICM to A*CRC, across the then world’s newest 100Gbps international connection, CAE-1, covering a distance 12,375 miles. The talk takes a look at the motivation, the unusual preparation, the set up at both the Poland side and the Singapore side, and a few key engineering accomplishments. Also explained is the three different types of data movement solutions, why this endeavor is a production trial instead of a demo, together with its significance in the context of modern data movement in the hybrid cloud environment.

Vladimir Brusic

School of Computer Science, University of Nottingham Ningbo China

Title: Single cell transcriptomics – new challenges for Big Data analytics

Abstract: Transcriptomics is the study of complete set of gene expression (expressed RNA) produced by transcription of DNA in a given biological sample. Biological samples typically contain mixtures of cells. The analysis of transcriptome aims to identify genes that show differences in expression between samples that represent different biological conditions, such as different cell types, different activation status, developmental stages, or disease states. This knowledge provides insight into biological processes, understanding the role of genes in biological processes, and changes related to various biological conditions. Transcriptomics has many practical applications in all fields of life sciences. Prominent areas of application include developmental biology, immunology, virology, agriculture and food science, and medicine. Clinical applications of transcriptomics include discovery of biomarkers, disease diagnosis and prognosis, selection and optimization of therapies, and disease monitoring. A great promise of clinical transcriptomics is the possibility of screening for multiple diseases, understanding the disease causes, estimating the disease progression, and assessing likely responses to specific treatment, all in a single transcriptomics experiment.

Bulk transcriptomics measures expression of tens of thousands of gene from the overall mixture of cells. On the other hand, single-cell transcriptomics (SCT) determines levels of expression of tens of thousands of genes in individual cells from a given biological sample. The main advantages of bulk sequencing is the ability to perform “deep“ sequencing, allowing detection of transcripts that are present in minute concentrations. However, bulk transcriptomics will miss differential expression of transcripts between cell types and subtypes present in the sample. SCT offers significant advantages as compared to bulk sequencing, including ability to profile different cell types and subtypes and identification of novel or rare cell types. The tradeof of SCT sequencing is that it is more “shallow“ in comparison to bulk sequencing, resulting in lower transcript coverage, noisier data, and larger variability than bulk sequencing. Cell populations that have very similar bulk transcriptomes often show remarkably variable transcriptome profiles of single cells because of inherent biological variability, a mix of cell-cycle stages, random variability, and shallow sequencing.

Nicolas Tonello

Constelcom Ltd

Title: Constellation® – Supercomputing at your fingertips – Delivering HPC power and expertise to all

Abstract: Whilst supercomputers at HPC Centres represent the best in class, access to on-premises HPC remains siloed and challenging for both users and HPC centres. The challenges facing users who could benefit from the advanced performance the systems bring include technical as well as operational issues, from the need for advanced computing expertise to having the access and knowledge to run software to take full advantage of supercomputers, all the way to being able to register on the systems in the first place and finding the resources. Centres are also facing challenges in addressing users in their own professional environments and customs and in successfully delivering their full expertise. Constelcom will present Constellation®, its web accessible and application agnostic platform for self-managed, easy and secure access to High Performance Computing. Constellation® has been conceived to bridge the gap between users and HPC Centres, making it possible for users and organisations to plug into and collaborate with HPC at their fingertips and for HPC Centres to easily manage their systems and resources whilst effortlessly acquiring and providing their in-house expertise to users.

Eden Figueroa

Stony Brook University

Title: Towards the Quantum Internet: Building an entanglement-sharing quantum network

Abstract: The goal of quantum communication is to transmit quantum states between distant sites. The key aspect to achieve this goal is the generation entangled states over long distances. Such states can then be used to faithfully transfer classical and quantum states via quantum teleportation. This is an exciting new direction which establishes the fundamentals of a new quantum internet. The big challenge, however, is that the entanglement generated between two distant sites decreases exponentially with the length of the connecting channel. To overcome this difficulty, the new concepts of entanglement swapping, and quantum repeater operation are needed.
In this talk we will show our progress towards building a quantum network of many quantum devices capable of distributing entanglement over long distances connecting Stony Brook University and the Brookhaven National Laboratory in Long Island, New York. We will show how to produce photonic quantum entanglement in the laboratory and how to store it and distribute it by optically manipulating the properties of room temperature atomic clouds. Finally, we will discuss our recent experiments in which several quantum devices are already interconnected forming elementary quantum cryptographic and quantum repeater networks. Finally, we will discuss how these milestones can form the backbone of a future quantum-information-enhanced internet.

Makoto Taiji

RIKEN Center for Biosystems Dynamics Research Osaka

Title: MDGRAPE-4A: A Special-purpose computer system for Molecular Dynamics simulations

Abstract: We have developed several special-purpose computer systems for particle simulations called “GRAPE (GRAvity PipE)”. The GRAPE systems are dedicated or quasi-general-purpose accelerators attached to host computers. The efficient parallelization in GRAPE systems provided us high-performance at low cost and low energy. However, as the performance increases, it requires a more powerful host machine – a PC cluster. Thus, the bottleneck in strong scaling lies in host systems, not in accelerators. This is often the case for GPU cluster systems. To overcome the bottleneck, we are currently developing the “MDGRAPE-4”, the special-purpose computer system for molecular dynamics (MD) simulations. The MDGRAPE-4 has similar structure with Anton by D. E. Shaw Research. It consists of System-on-Chip LSIs with general-purpose cores, dedicated pipelines, network interfaces, and memory units. By the integration of these elements we aim to achieve a simulation speed of a few ten microsecond per step by its low latency memory access and networking.

Łukasz Orłowski

Co-Founder and CTO of Archanan

Title: Tomorrow’s Supercomputers, Yesterday’s Practices: Applications of cloud-backed large system emulation for supercomputer software development

Abstract: Software development for supercomputers has been flawed since the early days of supercomputing. Typically, a programmer user develops their software on their workstation and then submits it to a debug queue for testing. Waiting for the test job to run in the debug queue often proves futile, as the run may not provide any information useful in debugging or troubleshooting.
This approach has many shortcomings:

• The development workstation and the target supercomputer are two entirely different machines, at best they may have the same processor architecture.
• There is no network communication during computation in a workstation environment, only socket-to-socket, at best
• The debug queue is around 10% of the entire system resources, what if the target deployment requires 25% of the target system?
• Debuggers are often licensed on a per-node basis, making debugging in a large environment prohibitively expensive and incompatible with cloud.

Supercomputers are expensive and, by design, are meant to run production codes. It makes more sense to run development, testing, and validation workloads in the cloud.

James K. Gimzewski

University of California Los Angeles

Title: Emergent Atomic Switch Networks for Neuroarchitectonics:

Abstract: Artificial realizations of the mammalian brain alongside their integration into electronic components are explored through neuromorphic
architectures, neuroarchitectonics, on CMOS compatible platforms. Exploration of neuromorphic technologies continue to develop as an alternative computational paradigm as both capacity and capability reach their fundamental limits with the end of the transistor-driven industrial phenomenon of Moore’s law. Here, we consider the electronic landscape within neuromorphic technologies and the role of the atomic switch as a model device. We report the fabrication of an atomic switch network (ASN) showing critical dynamics and harness criticality to perform benchmark signal classification and Boolean logic tasks. Observed evidence of biomimetic behavior such as synaptic plasticity and fading memory enable the ASN to attain a cognitive capability within the context of artificial neural networks. Single atomic switches have been shown to exhibit both long-term and short-term memorization based on their history of excitation dynamics. When connected together in a massively dense network using nanoarchitectonics they additionally display emergent behavior observable using multielectrode arrays similar to EEG to follow their spatio-temporal dynamics. A number of architectures such as Reservoir computation (RC) are promising methods to implement neuromorphics.

Jaroslaw Jung & Krzysztof Halagan

Lodz University of Technology

Title: Technology of Real-World Analyzers (TAUR) and its practical application

Abstract: As a result of many years of research and testing, at Lodz University of Technology the concept of technology of Real-World Analyzers (in polish: Technologia Analizatorów Układów Rzeczywistych TAUR) was developed [1]. Devices built in this technology can contain up to several hundred million operational cells placed in 3D face centered cubic (fcc) network nodes. These machines are a fully parallel data processing systems equipped with low-latency communication channels, designed for simulations of huge amount of relatively simple elements working in parallel and interacting locally. Devices can support solving problems from various fields of science and technology such as e.g. molecular simulations, artificial intelligence or data encryption.
In frame of TAUR technology the device called Analyzer of Real Complex Systems (in polish: Analizator Rzeczywistych Układów Złożonych – ARUZ) was designed and constructed [2]. This machine is located in Poland in the city of Łódź in BioNanoPark. ARUZ is based on 26 000 reconfigurable FPGA devices (Field Programmable Gate Array) connected in 3D fcc network and working as signal processing units.
Currently, ARUZ is intensively used to analyze computationally demanding complex macromolecular systems. Simulation of soft matter systems in chemistry and polymer physics (e.g. polymer liquids containing brushes, bottle-brushes, polymer stars and networks) is pretty challenging because must involve a variety of time scales and broad range of sizes – from Fickian–like dynamics to segmental motion of chains and mass-center diffusion. An unique simulation method in this field, called Dynamic Lattice Liquid (DLL) model, was proposed by Tadeusz Pakula [3]. The DLL correctly represents the dynamics of the studied molecular structure, taking into account the diffusion restrictions associated with obstacles occurring in the system. The DLL algorithm was implemented in ARUZ allowing to study the dynamics of molecular systems on the time scale difficult to access (or unreachable in some cases) for existing computing systems.

Antonino Tumeo

Pacific Northwest National Laboratory

Title: The Data-Model Convergence: a case for Software Defined Architectures

Abstract: High Performance Computing, data analytics, and machine learning are often considered three separate and different approaches. Applications, software and now hardware stacks are typically designed to only address one of the areas at a time. This creates a false distinction across the three different areas. In reality, domain scientists need to exercise all the three approaches in an integrated way. For example, large scale simulations generate enormous amount of data, to which Big Data Analytics techniques can be applied. Or, as scientist seek to use data analytics as well as simulation for discovery, machine learning can play an important role in making sense of the disparate source’s information. Pacific Northwest National Laboratory is launching a new Laboratory Directed Research and Development (LDRD) Initiative to investigate the integration of the three techniques at all level of the high-performance computing stack, the Data-Model Convergence (DMC) Initiative. The DMC Initiative aims to increase scientist productivity by enabling purpose-built software and hardware and domain-aware ML techniques.

In this talk, I will present the objectives of PNNL’s DMC Initiative, highlighting the research that will be performed to enable the integration of vastly different programming paradigms and mental models. I will then make the case for how reconfigurable architectures could represent a great opportunity to address the challenges of DMC. In principle, the possibility to dynamically modify the architecture during runtime could provide a way to address the requirement of workloads that have significantly diverse behaviors across phases, without losing too much flexibility or programmer productivity, with respect to highly heterogeneous architectures composed by sea of fixed application specific accelerators. Reconfigurable architectures have been explored since long time ago, and arguably new software breakthroughs are required to make them successful. I will thus present the efforts that the DMC initiative is launching to design a productive toolchain for upcoming novel reconfigurable systems. I will then proceed to discuss the influences and relations of the DMC initiative with other projects ongoing at PNNL that are exploring methods to accelerate, through custom hardware/software stacks and advanced design automation approaches, machine learning and data analytics.

Trilce Estrada

University of New Mexico

Title: Graphic Encoding of Macromolecules for Quantitative Classification of Protein Structure and Representation of Conformational Changes

Abstract: The function of a protein depends on its three-dimensional structure. Computational approaches for protein function prediction, and more generally macromolecular analysis are limited by the expressiveness and complexity of protein representation formats. Partial structural representations and representations that rely on homology alignments are both computationally expensive and do not scale with the number of molecules, as three-dimensional matching is an NP-hard problem. Being able to represent heterogeneous macromolecules in a homogeneous, easy-to-compare, and easy-to-analyze format has the potential to disrupt the way and scale at which molecular analysis is done today. This talk introduces a generalizable and homogeneous representation of macromolecules that explicitly encodes tertiary structural motifs and their relative distance as a proxy of their interaction. The final goal of this encoding is to expose intra- and inter-molecular structural patterns in a scalable way, i.e., that does not require performing alignments, homology calculations, or other expensive operations between pairs, or sets, of proteins. To demonstrate the effectiveness of this encoding, we also present an image processing system based on deep convolutional neural networks that is able to use our graphic representation to perform high throughput protein function prediction and interpretation of protein folding.

Wilhelmina Nekoto

Data Engineer, Research Scientist (Computer Vision)

Title: Automated Wildlife Monitoring and Poacher Detection in Namibian Communal Conservancies

Abstract: Manual systems still underpin many of the Namibian economic sectors, hindering effective collec-tion, storage and processing of data, and hence affecting informed decision making, to social challenges such as poverty, underdevelopment and crime. Although, pushing for the development of supercomputing in a developing country may seem overly ambitious, doing it at the right scale provides research resources necessary in addressing many of the global challenges and presents opportunities for the economic growth. In this talk we aim to identify applications in which HPC-AI powered systems can be used for development planning, to impact the growth of economy and human capital. Using new technologies to create and provide datasets for research that will fuel scientific discoveries and provide the key to addressing social and economical challenges in Namibia and beyond.

Jay Lofstead

Scalable System Software Sandia National Laboratories

Title: Memory vs. Storage Software and Hardware: The Shifting Landscape

Abstract: New memory technologies, such as the persistent memory modules supported by the latest Intel chips offer a persistent storage device accessible on the memory bus. NVMe and other high performance storage devices offer extreme performance at increasingly affordable prices. With these technology shifts, what software we use for different tasks and what we can afford to do is changing. This talk will explore how things are changing and where new opportunities to enable science exist.

Jay Lofstead

Scalable System Software Sandia National Laboratories

Title: Containers and Data-Centric Computing

Abstract: Data-centric computing in the cloud frequently uses containerized applications for easier and more flexible deployment. This compute-focused structure works for data-centric computing, but is not ideal. Provenance and workflow systems track the connection between files and their origins, but this needs to be done using a less fragile approach. By extending file systems and the operating system to incorporate these provenance features, understanding where analyzed and processed data came from becomes easier.

Ziogas Alexandros Nikolaos

ETH Zürich

Title: A Data-Centric Approach to Extreme-Scale Ab initio Dissipative Quantum Transport Simulations

Abstract: The computational efficiency of a state of the art ab initio quantum transport (QT) solver, capable of revealing the coupled electro-thermal properties of atomically-resolved nano-transistors, has been improved by up to two orders of magnitude through a data centric reorganization of the application. The approach yields coarse-and fine-grained data-movement characteristics that can be used for performance and communication modeling, communication-avoidance, and dataflow transformations. The resulting code has been tuned for two top-6 hybrid supercomputers, reaching a sustained performance of 85.45 Pflop/s on 4,560 nodes of Summit (42.55% of the peak) in double precision, and 90.89 Pflop/s in mixed precision. These computational achievements enable the restructured QT simulator to treat realistic nanoelectronic devices made of more than 10,000 atoms within a 14× shorter duration than the original code needs to handle a system with 1,000 atoms, on the same number of CPUs/GPUs and with the same physical accuracy.

Eliu Huerta

University of Illinois at Urbana-Champaign

Title: Accelerated Artificial Intelligence for Big-data Physics Experiments

Abstract: The cyberinfrastructure needs of large-scale facilities share a common thread of computational challenges, namely, datasets of ever increasing complexity and volume, and the need for novel signal-processing algorithms to extract trustworthy scientific results in real-time with oversubscribed computational resources. I will discuss the emergence of accelerated artificial intelligence algorithms as a solution to address computational grand challenges in big-data physics experiments.

Piotr Bala

ICM UW

Title: High Performance Services for genetic research

Abstract: Recent progress in genetic research was possible thanks to the progress in the experimental techniques. However such progress will not be possible without computational methods developed in the last decades. One of the milestones is BLAST (Basic Local Alignment Search Tool) – an essential algorithm that researchers use for sequence alignment analysis. Amongst different implementations of the BLAST algorithm, the NCBI-BLAST is the most popular. It can run on a single multithreading node. However, the volume of nucleotide and protein data is fast-growing, making single node insufficient. Therefore it is more and more important to develop high-performance computing solutions, which could help researchers to analyze genetic data in a fast and scalable way. In this lecture, we will present the execution of the BLAST algorithm on HPC clusters and supercomputers in a massively parallel manner using thousands of processors. The PCJ (Parallel Computing in Java) library has been used to implement the optimal splitting up of the input queries, work distribution and search management. It is used with the non-modified NCBI-BLAST package which is an additional advantage for the users. The result application PCJ-BLAST is responsible for the reading sequence for comparison, splitting it up and starting multiple NCBI-BLAST executables. Since I/O performance could limit sequence analysis performance we will also address this problem. Part of the talk will be dedicated to the PCJ which is key technology enabling efficient parallelization of the BLAST and other algorithms. We will present a general idea of the library as a parallelization tool in the PGAS paradigm as well as implementation details. Examples of successful parallelization of different algorithms from HPC, BigData and AI application domains will be provided. We will show that using Java and PCJ library it is possible to perform sequence analysis using hundreds of nodes in parallel. Since genomics research is getting significant attention in Europe we will present the Human Exposome Assessment Platform (HEAP) which is Horizon 2020 project starting 1st January 2020. It will create a global research resource that enables ethical and efficient management and processing of massive data from geographically distributed large-scale population cohorts. Also, a standardized, integrated and generic informatics platform will be created, It will provide PCJ-BLAST and other highly scalable genetics applications to the wide research community.

Joanna Trylska

Centre of New Technologies University of Warsaw

Title: Functional dynamics of biomolecules with supercomputers

Abstract: Proper association and recognition of biomolecules determines many life-sustaining processes occurring in a crowded environment of cells. In addition, biomolecules are intrinsically flexible and their dynamics is related to function. For example, flexibility of the ribosome, a protein synthesis molecular machine, is crucial for the fidelity of this synthesis in cells. Thus, not surprisingly, aminoglycoside antibiotics that affect bacterial protein synthesis hinder the functional flexibility of the ribosome. Therefore, understanding the dynamics of biomolecular recognition is indispensable in any drug design process. Binding of drugs to their intracellular targets requires conformational rearrangements of both binding partners in order to form a tight complex and inhibit the function of the target. Transport of drugs through cellular membranes is also a highly dynamic process.

I will talk about the importance of investigating biomolecular flexibility in drug design and transport. Using HPC resources we apply various molecular modeling methods and molecular dynamics simulations to investigate the processes of diffusion and flexibility of biomolecules. I will show examples of biological systems whose dynamics modulates their function and is crucial for drug design.

Yuefan Deng

Stony Brook University, NY

Title: Fast and Accurate Multiscale Modeling of Platelets Guided by Machine Learning

Abstract: Multiscale modeling in biomedical engineering is gaining momentum because of progress in supercomputing, applied mathematics, and quantitative biomedical engineering. For example, scientists in various disciplines have been advancing, slowly but steadily, the simulation of blood including its flowing and the physiological properties of such components as red blood cells, white blood cells, and platelets. Platelet activation and aggregation stimulate blood clotting that results in heart attacks and strokes causing nearly 20 million deaths each year. To reduce such deaths, we must discover new drugs. To discover new drugs, we must understand the mechanism of platelet activation and aggregation. To model platelets’ dynamics involves setting the basic space and time discretization in huge ranges of 5-6 orders of magnitudes, resulting from the relevant fundamental interactions at atomic, to molecular, to cell, to fluid scales. To achieve the desired accuracy at the minimal computational costs, we must select the correct physiological parameters in the force fields as well as the spatial and temporal discretization, by machine learning. We demonstrate our results of speeding up a multiscale platelet aggregation simulation, while maintaining desired accuracies, by orders of magnitude, compared with traditional algorithm that uses the smallest of temporal and spatial scales in order to capture the finest details of the dynamics. We present our analyses of the accuracies and efficiencies of the representative modeling. We will also outline the general methodologies of multiscale modeling of cells at atomic resolutions guided by machine learning.

Michael Bader

Technical University of Munich

Title: Towards Exascale Hyperbolic PDE Engines – Are We Skating to Where the Puck Will Be?

Abstract: Exascale supercomputers are “ante portas”. They pose novel challenges to simulation software, such as how to cope with heterogeneous performance of compute resources, increased failure rates and respective demands on resiliency, or with scalability on multiple levels of parallelism, which are at the same time suffering from unpredictable performance. CPU/GPU power will continue to outperform memory and communication hardware, which asks for algorithms that feature high arithmetic intensity and minimize data movement across the entire memory hierarchy. Can we successfully develop simulation software that will run efficiently on the supercomputers of 2030?

This talk will present on recent work to better prepare simulation software for exascale, focusing on two concrete packages: ExaHyPE and SeisSol. ExaHyPE is an engine to solve hyperbolic PDE systems. While it provides flexibility with respect to the tackled PDEs, it focuses on high-order discontinuous Galerkin (with a-posteriori Finite-Volume-based limiting) on tree-structure Cartesian meshes as the underlying numerical scheme. I will outline ExaHyPE’s code generation approach to tailor the engine to different needs of application, algorithm and code optimisation experts, and will highlight a fine-grain task-offloading strategy which can respond to performance fluctuations in hardware and combines with an MPI rank replication approach. Similar, I will present recent work on the earthquake simulation package SeisSol. SeisSol allows several modelling variants for seismic wave propagation and earthquake sourcing. It relies on code generation for its performance critical element-local small tensor operations, to cope with the complexity of realising several model variants on different hardware.

Rob Knight

University of California San Diego

Title: The Human Microbiome: Big Challenges, Big Data, Big Compute

Abstract: Our ability to sequence DNA has improved many orders of magnitude in speed and cost over the past 15 years. This enabling technology allows us to read out the human microbiome (the genes from the bacteria, fungi, viruses, and other tiny organisms that live on and in our bodies) on a massive scale, associating these microbes with a range of diseases from the expected (gut microbes are involved in inflammatory bowel disease) to the surprising (they are also involved in autism). Much of the data interpretation relies on shared resources backed by large compute. Here I discuss our rebuilding the tree of life on SDSC’s Comet, and efforts to predict the structures of all the proteins in the gut microbiome on IBM’s World Community Grid, and discuss the implications of these resources for building a user interface to our microbiomes that allows us to shape it for life-long health.

Happy Sithole

The Council for Scientific and Industrial Research

Title: The role of Cyber-Infrastructure in Development of Africa

Abstract: South Africa has invested in cyber-infrastructure since 2006, and has developed a National Integrated Cyber-Infrastructure (NICIS), which provides high-speed networks, large data research capabilities and high performance computing. In an endeavour to ensure that these facilities have maximum impact in Africa, South Africa has developed programs to include the rest of the African continent in Cyber-Infrastructure Development. In this talk, I will outline the developmental challenges that the country faced and how they were addressed. I will also provide examples of success in industry facilitation using HPC and access to Data. The talk will also cover some of the developments in the other African countries. We will talk of the technology options and strategies on HPC technology choices and the rational behind those decisions.

Laura Boykin

Cassava Virus Action Project

Title: Utility of Real time portable genome sequencing and HPC for global food security

Abstract: Portable DNA sequencing technology has great potential to reduce the risk of community crop failure and help improve livelihoods of millions of people, especially in low resourced communities. Crop losses due to plant viral diseases and pests are major constraints on food security and income for millions of households in sub-Saharan Africa (SSA). Such losses can be reduced if plant diseases and pests are correctly diagnosed and identified early. Currently, accurate diagnosis for definitive identification of plant viruses and their vectors in SSA mostly relies on standard PCR and next generation sequencing technologies (NGS). However, it can take up to 6 months before results generated using these approaches are available. The long time taken to detect or identify viruses impedes quick within-season decision making necessary for early action, crop protection advise and disease control measures by farmers. This ultimately compounds the magnitude of crop losses and food shortages suffered by farmers. For the first time globally, the MinION portable pocket DNA sequencer was used to sequence whole plant virus genomes on the farm in Uganda, Tanzania and Kenya and we have reduced the time to diagnosis to 3 hours. In this talk I will outline how we have used the Oxford Nanopore portable genomics technology to identify the begomoviruses causing the devastating cassava mosaic disease, which is ravaging smallholder farmers’ crops in sub-Saharan Africa leaving millions food insecure. I will also cover the gaps in computing we have identified with our on farm genomic sequencing with the hopes of engaging with people in attendance at the Supercomputing Frontiers Europe 2020 into the future.

Calista Redmond

Chief Executive Officer, RISC-V Foundation

Title: The Reality and Tremendous Opportunity of Custom, Open Source Processing

Abstract: The growth of human and business interaction with technology continues to explode. At the literal heart of that technology sits a silicon core, combined with general and specific instructions and connections. The insane cost, risk, development time, necessary volumes, and limited computing demands kept the lucrative chip opportunity within reach of just a handful of companies — focused mostly on general purpose processors. New computing needs in various power and performance dimensions have increased demand and competition for custom processors. This pressure is quietly and rapidly disrupting the processor industry. An Open source approach to processors now reduces risk and investment, with accelerated time to market, and opens the opportunity to thousands of possible custom processors. Learn about the trends, opportunities, and examples — from smart watches to supercomputers!

FLORIN MANAILA

Senior AI Architect and Inventor at IBM Systems Hardware Europe

Title: HPC Transformation with AI

Abstract: The session will present HPC challenges in fueling machine learning and deep learning into the simulations. Besides, we will present a user-centric view of IBM Watson ML Community Edition and the newly IBM inference system IC922 adoption into AIops of large HPC clusters (from deployment to inference).