Communication is threatening to become an increasing bottleneck towards performance scaling in the post exascale era as bytes-per-FLOP ratios continue to decline. We describe bandwidth steering in HPC to take advantage of emerging photonic switches for efficiently changing the connectivity of the lower layers in a hierarchical topology to reconstruct locality that was lost from system fragmentation and was impossible to recover with task placement. This allows for more aggressive oversubscription of the higher layers to reduce cost with no performance penalty. We demonstrate bandwidth steering with a scalable algorithm in an experimental testbed and at system scale using simulations. At the system scale, bandwidth steering reduces static power consumption per unit throughput by 51% and dynamic power consumption by 10% compared to a reference topology. In addition, bandwidth steering reduces average network latency by up to 87% and improves the average throughput by an average of 4.3x.

As the field of Quantum Computing grows, various levels of abstraction must be developed to make it easier for users to adopt. Sitting in between a control processor and the digital/analog interface for a classical control system, we propose a layer called qFirm. This layer will provide the vital interface for converting quantum instructions into the analog signals sent to control the quantum device, as well as reading the results of the device. This will provide the essential glue logic for the classical control stack for a quantum control system.

Extensive research in material science together with outstanding engineering efforts allowed quantum technology to be significantly improved hence enabling continuing scaling of quantum circuit size. However, quantum circuit scaling itself does not guarantee any practical use without appropriate progress on the part of classical control hardware and software. To operate such a large-scale universal quantum computer with thousands of qubits, extensive classical computational resources will be required. Control hardware includes multiple layers each of which is responsible for a specific set of tasks, e.g. controllers, digital-analogue and analogue-digital converters, filters, waveform generators, etc. At this early stage of quantum architecture development, there is no clear understanding of where the upcoming challenges will be addressed through the entire stack of complex digital and analogue circuits. However, we expect that control processor will become a crucial part for successful implementation and adoption of future quantum computers.
In our talk, we discuss the challenges that classical processors will face while controlling future large-scale quantum systems. We discuss how these challenges will affect processor micro-architecture to guarantee on time quantum gate execution, continuing qubit state measurement, store and analysis, support massive parallelism and perform advanced bit manipulations on the top of the measured data.

As we approach the end of Moore’s law modern, complex, HPC systems are increasingly relying upon specialized accelerators in order to deliver continued performance increases for specific computational workloads. Developers of these accelerators, especially in in many low volume scientific applications, face a stark choice: spend millions on a commercial license for processors and other IP, or face the significant risk and of developing custom hardware. Rapid prototyping methods need to be explored in order to make the design, verification and programming tools for these new accelerators more accessible to the broader scientific community. To increase access and innovation while reducing cost there has been a consistent march towards open source solutions for each of these components including Facebook’s Open Compute Project and Intel’s OpenHPC effort, as well as a burgeoning community surrounding RISC-V based processors.
Looking beyond accelerators that may be tightly integrated with HPC systems we see opportunities for open source hardware to include programmable logic embedded within high performance sensors and detectors for aggressive data reduction or being used in conjunction with FPGA and other reconfigurable computing based platforms. This talk will explore the emerging open source hardware effort as well as showcase new platforms for the rapid generation of future HPC accelerators.

An increasing number of technologies are being proposed to preserve digital computing performance scaling as lithographic scaling slows. These technologies include new devices, specialized architectures, memories, and 3D integration. Currently, no end-to-end tool flow is available to rapidly perform architectural-level evaluation using device-level models and for a variety of emerging technologies at once. We propose PARADISE: An open-source comprehensive methodology to evaluate emerging technologies with a vertical simulation flow from the individual device level all the way up to the architectural level. To demonstrate its effectiveness, we use PARADISE to perform end-to-end simulation and analysis of heterogeneous architectures using CNFETs, TFETs, and NCFETs, along with multiple hardware designs. To demonstrate its accuracy, we show that PARADISE has only a 6% mean deviation for delay and 9% for power compared to previous studies using commercial synthesis tools.

The USA Exascale Computing Project (ECP) is focused on accelerating the delivery of a capable exascale computing ecosystem that delivers 50 times more computational science and data analytic application power than possible with DOE HPC systems such as Titan (ORNL) and Sequoia (LLNL). As next generation applications and experiments grow in concurrency and in complexity, the data produced often grows to extreme levels, limiting scientific knowledge discovery. In this presentation, I will talk about the new set of applications and experiments which push the edge of scientific data processing and simulation. I will present some of the exciting new research in this area to cope with this tsunami of data, along with the challenges in implementing these effectively on next-generation computer architectures. In this presentation I will also focus on the ADIOS framework a next generation to ingest, reduce, and move data on HPC systems and over the WAN to other computational resources. I will also focus on in situ data processing infrastructure and next generation data compression algorithms.
※紹介されるツールのチュートリアルが同日午後に行われます。
[ Time table ]13:00-16:00　ADIOS2 Tutorial
Lecturers: Scott Klasky (ORNL), Norbert Podhorszki (ORNL)
興味を持たれた方は、ADIOS2 Adaptable I/O Frameworkチュートリアル 開催（4月15日・神戸）をご覧下さい。

米国における気候数値モデル開発の一例をご紹介いたします。発表者は米国エネルギー省SciDACプロジェクトを通してE3SMという全球モデル開発に参加しています。E3SMはエネルギー問題を解決するために，高解像度の地球環境シミュレーションを実施し、水循環，雪氷-海洋系，そして生物圏の理解を進めることを目的に開発されているモデルです。計算能力向上のために，機械学習やGPUコンピューティングにも力を入れており、米国エネルギー省の次世代スーパーコンピュータ「Shasta」で計算が実行される予定です。この次世代機はクレイ社によって発表されたばかりで，AMD，Intel，ARM，そしてGPUといった様々な計算コアが実装される予定です。モデル開発者は，まだ見ぬ次世代機に適用するため様々な手法を考えています。

Agent-Based Model (ABM) is a promising tool for city-scale traffic simulation to understand the complex behaviour of the entire urban transportation system under different scenarios. In the ABM, traffic is intuitively simulated as movements and interactions between large numbers of agents, each capable of finding the route for an individual traveller or vehicle. In this talk, the development of such an ABM simulation tool will be presented to reproduce the traffic patterns of the city of San Francisco. The model features a detailed road network and hour-long simulation time step to capture realistic variations in traffic conditions. Agent speed is determined according to a simplified volume-delay macroscopic relationship, which is more efficient than applying microscopic rules (e.g., car following) for evaluating city-scale traffic conditions. Two particular challenges of building such an ABM will be discussed in particular: data availability and computational cost. The key inputs to the ABM are sourced from standard and publicly available datasets, including the travel demand surveys published by local transport authorities and the road network data from the OpenStreetMap. In addition, an efficient priorityqueue based Dijkstra algorithm is implemented to overcome the computational bottleneck of agent routing. The ABM is designed to run on High Performance Computing (HPC) clusters, thereby improving the computational speed significantly. Preliminary validation of the ABM is conducted by comparing its results with a published model. Overall, the ABM has been demonstrated to run efficiently and produce reliable results. Use cases of the ABM tool will be demonstrated through two examples, including evaluating the value of real-time traffic information and assessing the outcomes of complex network-level emission mitigation measures.

Effective use of data management techniques for analysis and visualization of massive scientific data is a crucial ingredient for the success of any supercomputing center and cyberinfrastructure for data-intensive scientific investigation. In the progress towards exascale computing, the data movement challenges have fostered innovation leading to complex streaming workflows that take advantage of any data processing opportunity arising while the data is in motion. This technology finds practical use in a number of industry applications including precision agriculture and tele-medicine. In this talk I will present a number of techniques developed at the Center for Extreme Data Management Analysis and Visualization (CEDMAV) that allow to build a scalable data movement infrastructure for fast I/O while organizing the data in a way that makes it immediately accessible for analytics and visualization. In addition, I will present a topological analytics framework that allows processing data in-situ and achieve massive data reductions while maintaining the ability to explore the full parameter space for feature selection. Overall, this leads to a flexible data streaming workflow that allows working with massive simulation models without compromising the interactive nature of the exploratory process that is characteristic of the most effective data analytics and visualization environment.
※ストリーミング型大規模データ可視化ツールOpenViSUSのチュートリアルを3月20日に行います。
詳細はOpenViSUS ストリーミング型大規模データ可視化チュートリアル（2019年3月20日）をご覧下さい。