Hal Finkel | Kalyan Perumalla

Competitive Portfolios for Advanced Scientific Computing Research: Data Management and Visualization

Sat, 01 Feb 2025 00:00:00 +0000

Solicitation PDF

Original at OSTI
Cached local copy

Selected Pages

Selected Extracts

The SC ASCR program hereby announces its interest in advanced scientific computing research portfolios for accelerating discovery and innovation in support of the DOE mission. ASCR seeks to invest in DOE National Laboratory-led portfolios that balance long-term, high-impact research along with the ability to aggressively respond to, and take advantage of, emerging science and technology trends. The ASCR Computer Science (CS) research program [1] supports long-term, basic research that enables computing and networking at extreme scales and the understanding of extreme-scale and complex data from both simulations and experiments. ASCR, in tandem with industry and others, has made highly successful investments to ensure U.S. leadership in high performance computing (HPC), which resulted in Exascale systems that are enabling scientific discovery and decision support through data integration, simulation and modeling [2].

To ensure continued leadership in delivering on the promise of computational science, and drive innovation in energy-efficient and versatile HPC for science, ASCR seeks to invest in DOE National Laboratory-led portfolios that:

Support long-term, high-impact CS research,
Aggressively respond to, and take advantage of, emerging science and technology needs and trends including Artificial Intelligence (AI), and
Collaborate with a diverse community of the most-promising academic and industry partners

SUPPLEMENTARY INFORMATION

Scientific research driven by Artificial Intelligence (AI)-enabled technologies is not only making scientists more productive but promises to change how scientists find the most-promising ideas to investigate in the future [3]. This requires deep changes in the methods available, and algorithms developed, to store, search, retrieve, analyze, and visualize scientific data. Past efforts which focused primarily on storing and analyzing data quickly only in specifically-anticipated contexts are giving way to discovery-optimized techniques which prioritize supporting AI- enabled investigation and the aggregation of curated data sets of many kinds.

In this context, ASCR seeks innovative research with vision beyond its current investments in HPC data management, storage [4], and scientific visualization [5] that will help enable the development of the next-generation energy-efficient and capable computing systems [6] and approaches enabling accelerated scientific discovery. This can include the use of new hardware, software, algorithms, and other related technologies that are currently at early stages of development.

While proposed work can leverage software from prior research efforts where they add significant value, ASCR is primarily looking for new research efforts in scientific data management, storage, and visualization. These efforts should build on the best available open platforms and befit the future of energy-efficient AI-driven scientific discovery where data management and visualization are fast, efficient, and flexible.

Exploratory Research for Extreme-Scale Science (EXPRESS)

Sat, 01 Feb 2025 00:00:00 +0000

Solicitation PDF

Original at OSTI
Cached local copy

Selected Pages

Selected Extracts

B) Local and Campus-Area Quantum Networking for Next Generation Parallel Quantum Computing

Quantum networking involves effective communication of quantum information among geographically distributed quantum systems, separated by short or long distances. Sources of quantum information in the network communication include qubits used in different forms of quantum computing or output from various quantum sensing devices.

This topic seeks advancements in quantum networking over short distances to enable parallel quantum computing within a building and integration of quantum information sources to storage and quantum computing across a campus area. The interconnection of different co-located quantum computing systems is aimed at increasing the scale of quantum computation (e.g., aggregate number of qubits) and at progressing towards an architecture of flexible connectivity of quantum devices across a laboratory or university campus, potentially composing heterogeneous quantum computing hardware, including broadening of networking from exchange of physical qubits to logical qubits.

Research Area

The specific aim of this topic is to support quantum science needed to effectively scale quantum computing and enable flexible exchanges of coherent quantum information via locally networked heterogeneous quantum systems. Proposals must address one or more research advancements in the aforementioned directions in local and campus-area quantum networking. Research must be aimed at advancing our understanding of aspects, such as core concepts, devices, architectures, integration, and interfaces, that are necessary for a quantum counterpart to the current infrastructures of classical local and campus area networks within the scientific and other facilities.

Early Career Research Program (ECRP)

Wed, 01 Jan 2025 00:00:00 +0000

Solicitation PDF

Original at OSTI
Cached local copy

Selected Pages

Selected Extracts

Computer Science: Systems

Programming Models and Environments: Innovative programming models for developing applications on next-generation platforms, exploiting unprecedented parallelism, heterogeneity of memory systems (e.g. non-uniform memory access [NUMA], non-coherent shared memory, high-bandwidth memory [HBM], scratchpads, and heterogeneity of processing (e.g. graphics processing units [GPUs], field- programmable gate arrays [FPGAs], coarse-grained reconfigurable architectures [CGRAs], other types of accelerators, big-small cores, processing in memory, and near memory, etc.), with particular emphasis on making it easier to program at scale. All phases of the software-development cycle are relevant, including but not limited to, design, implementation, verification, optimization, and integration. Particularly welcome are methods that infuse artificial intelligence/machine learning into the programming environment.
Operating and Runtime Systems: System software that provides intelligent, adaptive resource management and support for highly-parallel software and workflow-management systems, and that facilitates effective and efficient use of heterogeneous computing technologies, including diverse execution models, processors, accelerators, memory, and storage systems. Target workloads include modeling and simulation, data analysis, and the processing of large- scale, streaming data from experiments.
Performance Portability and Co-design: Methods that support performance portability, which provides the ability to efficiently use diverse kinds of hardware platforms with minimal changes to the application source code, and/or hardware/software co-design, which is a method for designing and/or adapting both hardware and software design as part of a holistic process. These methods include automated and semi-automated refinements from high-level specification of an application and/or hardware design to low-level code, optimized when compiled and/or, for software, at runtime, to different HPC platforms. The focus is on enabling performance portability of, and/or the design of future hardware for, applications developed for extreme-scale computing and beyond.
Memory-Aware Systems: Advances in memory technologies are creating new opportunities and challenges where it is unclear how to best introduce or abstract memory awareness and composition. Memory is evolving in highly asymmetric and distributed directions, with new industry standards greatly expanding memory sharing and capacities to much larger sizes, largely in backward-compatible system architectures. Research is needed to uncover new possibilities for solving larger scientific computing problems with such highly asymmetric and distributed memory architectures. Innovations in algorithms, software interfaces, programming languages and models are needed to also effectively exploit new processing-in-memory architectures that are emerging as a relatively newer paradigm for scientific computing. Memory safety needs to be revisited in new research aimed at a more fundamental level of programming languages, runtimes, and operating systems, considering the multi-developer and shared nature of modern scientific programming eco-systems. The smoothening of the spectrum from volatile to non- volatile memories needs to be investigated for revisiting out-of-core algorithms to expand the limits of scientific computing. On-the-fly compression and decompression needs investigation for increasing the problem sizes without detriment to performance. The intersection of machine learning with memory systems opens the potential for new solutions, including smarter ML- informed cache prefetching and replacement policies potentially customizable for specific scientific applications via signatures and other mechanisms.

Applications are not restricted to a single Systems topic above and may span all of them, provided the scope of work remains appropriate for the program.

Continuation of Solicitation for the Office of Science Financial Assistance Program

Tue, 01 Oct 2024 00:00:00 +0000

Announcement

Press Release

Solicitation PDF

Original at OSTI
Cached local copy

Selected Pages

Selected Extracts

Network-Offloaded Acceleration for Distributed/Parallel Computing: Programmable and computation-enabled network interfaces present the opportunity to exploit computational power closer to the network to complement the capabilities of CPUs, GPUs, and other computational components. Note that the programmable network interfaces include both edge accelerators as well as devices in core interconnects in parallel platforms or transport planes in distributed settings. Application behavioral information may be exploited, both in terms of dynamic learning as well as mathematically predefined primitives such as distributed reductions and other offloaded synchronization operations. New methods, algorithms, software, and interfaces are needed to effectively exploit asynchronous and autonomous capabilities of network hardware beyond traditional data-transfer functionalities. Of interest are new conceptual approaches, algorithmic support, application programming interfaces, and use cases in HPC scientific applications.
Computer Science Fundamentals Accounting for Thermodynamics and Energy: Unprecedented levels of modern computation, including areas such as artificial intelligence and machine learning (AI/ML) training, have now made computation a very large consumer of energy in the Nation and the world. Much of modern computer science, and the understanding it provides regarding the fundamental properties of algorithms, does not account for the underlying thermodynamic and information-theoretic reality of computation. As “Beyond Moore” devices are explored along with their corresponding ultra-efficient computer architectures, and the programming paradigms appropriate for these new computing technologies, a better understanding is needed of both potential ultra-efficient computer architectures and the energy-aware properties of algorithms executed on them. Ultra-efficient computer architectures include, but are not limited to, those based on reversible and asymptotically-adiabatic approaches. Investigations combining thermodynamics and information theory, computer architecture, reversible computing and algorithmic properties are sought to advance our ability to design new, energy-efficient approaches to scientific computation.
Memory-Aware Systems: Advances in memory technologies are creating new opportunities and challenges where it is unclear how to best introduce or abstract memory awareness and composition. Memory is evolving in highly asymmetric and distributed directions, with new industry standards greatly expanding memory sharing and capacities to much larger sizes, largely in backward- compatible system architectures. Research is needed to uncover new possibilities for solving larger scientific-computing problems with such highly asymmetric and distributed memory architectures. Innovations in algorithms, software interfaces, programming languages and models are needed to also effectively exploit new processing-in-memory architectures that are emerging as a paradigm for scientific computing. Memory safety needs to be revisited in fundamental research on programming languages, runtimes, and operating systems, considering the multi-developer and shared nature of modern scientific programming eco- systems. The smoothening of the spectrum from volatile to non-volatile memories needs to be investigated for revisiting out-of-core algorithms to expand the limits of scientific computing. On-the-fly compression and decompression needs investigation for increasing the problem sizes without detriment to performance. The intersection of machine learning (ML) with memory systems opens the potential for new solutions, including smarter ML-informed cache prefetching and replacement policies potentially customizable for specific scientific applications via signatures and other mechanisms.
Quantum Networking: This topic involves innovative research in quantum networking concepts, systems, and protocols by which quantum networking applies in scientific discovery, including, but not limited to, distribution of quantum information from sensors, quantum networking in support of interconnected or scalable quantum computing systems, and blind/cloud quantum computing. Networking can span heterogeneous systems or homogeneous systems (such as all-photonic) and parallel quantum processing (in co-located or local-area settings) and distributed quantum communications (at metropolitan or wide-area scales). Possible topics include quantum networking areas as presented in “Report for the ASCR Workshop on Basic Research Needs in Quantum Computing and Networking,” https://doi.org/10.2172/2001045.

Advancements in Artificial Intelligence for Science

Thu, 01 Feb 2024 00:00:00 +0000

Solicitation PDF

Original at OSTI
Cached local copy

Selected Pages

Selected Extracts

The DOE SC program in Advanced Scientific Computing Research (ASCR) hereby announces its interest in basic computer science and applied mathematics research in the fundamentals of Artificial Intelligence (AI) for science. Specifically, advancements in this area are sought that can enable the development of:

Foundation models for computational science;
Automated scientific workflows and laboratories;
Scientific programming and scientific-knowledge-management systems;
Federated and privacy-preserving training for foundation and other AI models for science; and
Energy-efficient AI algorithms and hardware for science.

The development of new AI techniques applicable to multiple scientific domains can accelerate progress, increase transparency, and open new areas of exploration across the scientific enterprise.

Research Area 1: …

…

Research Area 2: AI Innovations for Scientific Knowledge Synthesis and Software Development

The state-of-the-art in knowledge synthesis and programming tools are changing rapidly, fueled by AI Large Language Models (LLMs) trained on text, source code, and other data sources. New AI-driven tools are currently not trustworthy; do not systematically understand mathematical and physical principles; cannot properly ingest and understand scientific literature and data; and do not produce consistent, verified, uncertainty-quantified, reproducible results. In addition to addressing those challenges, there may be particular advantages in such tools building up knowledge and context over many interactions with a user or group of users. However, incremental training of AI systems over long time horizons, and the representation of knowledge in AI systems robust to changes in the underlying AI models, remain critical challenges. This research area seeks fundamental advancements in knowledge synthesis and programming tools for science. Moreover, realizing AI systems that can truly understand, and assist with, all aspects of the scientific process requires innovation in many areas, including multimodality, tool use, deeper reasoning and planning, memory, and external interaction. For additional background, see Chapter 2, “AI Foundation Models for Scientific Knowledge Discovery, Integration, and Synthesis,” Chapter 6, “AI for Programming and Software Engineering,” Chapter 12, “Mathematics and Foundations,” and Chapter 14, “Data Ecosystem,” of the AI For Science, Energy, and Security report [1].

Additionally, investigations into AI-driven tools for science should be conceptualized accounting for the iterative and collaborative processes that define modern science and scientific-software development. Accordingly, research proposed in this area is encouraged to address the relevant Priority Research Directions (PRDs) from the Basic Research Needs in The Science of Scientific Software Development and Use report [6], which are PRD 1, “Develop next-generation tools to enhance developer productivity and software sustainability,” PRD 2, “Develop methodologies and tools to comprehensively improve team-based scientific software development and use,” and PRD 3, “Develop methodologies, tools, and infrastructure for trustworthy software-intensive science.

Methods proposed for investigation should use any appropriate techniques that might be necessary to accomplish their goals, including, but not limited to, machine learning, natural- language processing, formal reasoning, instrumentation, data management, and compiler technology. The sustainability and explainability of scientific software are critically important to the scientific process, and as a result, particular consideration should be given to maximizing the extent to which human programmers understand and/or trust the outputs of these methods.

Research Area 3: …

…

Research Area 4: …

…

Research Area 5: …

…

Competitive Portfolios for Advanced Scientific Computing Research

Mon, 01 Jan 2024 00:00:00 +0000

Awards

Solicitation PDF

Original at OSTI
Cached local copy

Selected Pages

Selected Extracts

To ensure continued leadership in delivering on the promise of computational science, and drive innovation in energy-efficient and versatile high-performance computing for science, ASCR seeks to invest in DOE National Laboratory-led portfolios that:

Support long-term, high-impact research,
Aggressively respond to, and take advantage of, emerging science and technology trends, and
Collaborate with a diverse community of the most-promising academic and industry partners.

Research Proposals: Each Laboratory is limited to leading one proposal in response to this Announcement. The Principal Investigator (PI) must be a Laboratory division director or a manager with equivalent supervisory responsibilities. The proposal narrative (at most 30 pages) must provide a Laboratory vision and management plan for the portfolio of capabilities stemming from the proposed research and development in scientific computing. The narrative is comprised of one or more research Thrusts. Each Thrust must have a Laboratory Senior/Key Personnel (SKP) as the Lead along with other SKPs and researchers. Overall, the proposal must describe the research Thrusts and integration Tasks needed to enable new scientific computing- based capabilities that address national priorities in energy, the environment, and national security. The proposal should describe how the overall vision and each Thrust take advantage of the responding Laboratory’s, and each partnering institution’s, distinctive expertise and capabilities.

Research Thrusts: A Thrust is a distinct, focused area of basic research in scientific computing.

Applied Mathematics: Single-facet Thrusts require and build on research expertise in a core area such as s linear algebra and nonlinear solvers, discretization and meshing, multi-scale mathematics, discrete mathematics, optimization, complex systems, emergent phenomena, and applied analysis methods including but not limited to analysis of large- scale data, uncertainty quantification, and error analysis, or related topics. [4, pg. 281]
Computer Science: Single-facet Thrusts require and build on research expertise in a core area such as programming languages, high-performance computing tools, peta- to exa- scale scientific data management and scientific visualization, distributed computing infrastructure, programming models for novel computer architectures, and automatic tuning for improving code performance, or related topics. [4, pg. 281]
Advanced Computing Technologies and Testbeds: Access to resources to test and develop new tools, libraries, languages, etc. is an important enabling capability [4, pg. 281]. Thrusts focused on the establishment and development of testbeds1 that offer promising paths to versatile energy-efficient computing, addressing among other challenges, the data storage and movement requirements of artificial intelligence and simulation workloads may be proposed in response to this Announcement. Computing hardware should be interpreted broadly to include computational, memory, networking,