Correctness for Scientific Computing Systems | 24 571

Opportunity Information: Apply for 24 571

Correctness for Scientific Computing Systems (CS2) is a joint NSF and DOE research grant program aimed at making correctness a first-class requirement in scientific computing, on par with performance. The opportunity targets the full scientific computing stack, from low-level numerical libraries and compiler/toolchain components up through large-scale, multi-physics simulation codes and newer end-to-end scientific workflows. The core idea is to move beyond the traditional emphasis on speed alone and instead develop methods, tools, and evidence that scientific computing systems behave as intended when they run, even at scale and under the practical constraints of high-performance environments.

In this program, "correctness" is defined broadly and pragmatically as the satisfaction of desired behavioral properties during execution. In scientific computing, that includes preventing or controlling a wide range of failure modes across software and hardware, such as excessive numerical rounding and loss of significance, floating-point exceptions, data races and deadlocks in parallel code, memory faults, and mismatches or violations at module boundaries where components interact through specified interfaces. CS2 also recognizes that many scientific applications are inherently statistical, incorporate randomized or probabilistic algorithms, or work with uncertain data. Because of that, the program allows (and in some cases effectively requires) probabilistic notions of correctness, where guarantees may be expressed in terms of likelihoods, bounds, or distributions rather than absolute deterministic statements.

A key theme in the solicitation is that correctness guarantees are always relative to a model of the system, and real systems often deviate from that model due to misspecifications, approximations, or defects. When those deviations happen, it can be hard to diagnose what went wrong and why, especially in complex HPC environments. CS2 therefore emphasizes not only proving or establishing correctness with respect to a model, but also developing ways to probe, observe, and isolate discrepancies between the model and real execution. That can include techniques that combine static reasoning with runtime checks, monitoring, or other diagnostic approaches that help pinpoint where the assumptions break down.

CS2 requires close, sustained collaboration between two complementary research communities. One required area of expertise is scientific computing, interpreted broadly to include scientific modeling and simulation, data management and analysis for simulations and experiments, and numerical computation libraries and related infrastructure. The second required area is formal reasoning and mechanized methods for establishing program properties, also interpreted broadly to include formal verification (automatic, interactive, and hybrid approaches), runtime verification, type systems, abstract interpretation, programming languages and program logic, compiler methods, concurrency reasoning, stochastic reasoning, and rigorous testing approaches such as static/dynamic testing and property-based testing. The program is explicitly structured so proposals are not meant to be single-discipline efforts; each project must include leadership coverage for both scientific computing and formal reasoning/mechanized proving.

Eligibility is limited to specific U.S.-based organization types. Proposals may be submitted by U.S. non-profit, non-academic organizations that are directly tied to research or educational activities (for example, independent research labs, museums, observatories, or professional societies), by U.S. accredited two- and four-year Institutions of Higher Education (including community colleges) acting on behalf of their faculty, and by DOE National Laboratories. For personnel eligibility, by the submission deadline each PI, co-PI, or other senior/key person must either be a DOE National Laboratory employee or hold an eligible academic/research appointment at a U.S.-based campus of an eligible submitting organization (typically tenured/tenure-track, or a primary full-time paid research or teaching position, with limited exceptions such as approved family or medical leave). Individuals whose primary appointment is at a for-profit non-academic organization or at an overseas branch campus of a U.S. institution are not eligible. In addition to these formal rules, the solicitation includes a specific team-composition requirement: every project must have at least one PI/co-PI with scientific computing expertise and at least one PI/co-PI with formal reasoning and mechanized proving expertise.

Administratively, this is a discretionary grant opportunity led by the National Science Foundation in partnership with the Department of Energy, listed under funding opportunity number 24-571, with CFDA/assistance listings 47.070 and 81.049. The original closing date provided is 2024-08-13. The solicitation text provided does not specify an award ceiling or the expected number of awards, so applicants would need to consult the full announcement for budget ranges, project duration norms, and any additional submission requirements.

• The National Science Foundation in the science and technology and other research and development sector is offering a public funding opportunity titled "Correctness for Scientific Computing Systems" and is now available to receive applicants.
• Interested and eligible applicants and submit their applications by referencing the CFDA number(s): 47.070, 81.049.
• This funding opportunity was created on 2024-05-10.
• Applicants must submit their applications by 2024-08-13. (Agency may still review applications by suitable applicants for the remaining/unused allocated funding in 2026.)
• Eligible applicants include: Others.

Apply for 24 571

[Watch] Creating a grant proposal using the step-by-step wizard inside the applicant portal: