About the conference
As supercomputers become more and more powerful, the number and diversity of applications that can be tackled with these machines grows. Unfortunately, the architectural complexity of these supercomputers grows as well, with heterogeneous processors, multiple levels of memory hierarchy, and many ways to move data and synchronize between processors. The MPI+X programming model, use of which is considered by many to be standard practice, demands that a programmer be expert in both the application domain and the low-level details of the architecture(s) on which that application will be deployed, and the availability of such superhuman programmers is a critical bottleneck. Things become more complicated when evolution and change in the underlying architecture translates into significant re-engineering of the MPI+X code to maintain performance.
Numerous alternatives to the MPI+X model exist, and by raising the level of abstraction on the application domain and/or the target architecture, they offer the ability for “mere mortal” programmers to take advantage of the supercomputing resources that are available to advance science and tackle urgent real-world problems. However, compared to the MPI+X approach, these alternatives generally lack two things. First, they aren’t as well known as MPI+X and a domain scientist may simply not be aware of models that are a good fit to their domain. Second, they are less mature than MPI+X and likely have more functionality or performance “potholes” that need only be identified to be addressed.
PAW-ATM is a forum for discussing HPC applications written in alternatives to MPI+X. Its goal is to bring together application experts and proponents of high-level languages to present concrete example uses of such alternatives, describing their benefits and challenges.
Scope and Aims
The PAW-ATM workshop is designed to be a forum for discussion of supercomputing-scale parallel applications and their implementation in programming models outside of the dominant MPI+X paradigm. Papers and talks will explore the benefits (or perhaps drawbacks) of implementing specific applications with alternatives to MPI+X, whether those benefits are in performance, scalability, productivity, or some other metric important to that application domain. Presenters are encouraged to generalize the experience with their application to other domains in science and engineering and to bring up specific areas of improvement for the model(s) used in the implementation.
In doing so, our hope is to create a setting in which application authors, language designers, and architects can present and discuss the state of the art in alternative scalable programming models, while also wrestling with how to increase their effectiveness and adoption. Beyond well-established HPC scientific simulations, we also encourage submissions exploring artificial intelligence, big data analytics, machine learning, and other emerging application areas.
Topics of interest include, but are not limited to:
- Novel application development using high-level parallel programming languages and frameworks.
- Examples that demonstrate performance, compiler optimization, error checking, and reduced software complexity.
- Applications from artificial intelligence, data analytics, bioinformatics, and other novel areas.
- Performance evaluation of applications developed using alternatives to MPI+X and comparisons to standard programming models.
- Novel algorithms enabled by high-level parallel abstractions.
- Experience with the use of new compilers and runtime environments.
- Libraries using or supporting alternatives to MPI+X.
- Benefits of hardware abstraction and data locality on algorithm implementation.
Papers that include description of applications that demonstrate the use of alternative programming models will be given higher priority.
Submissions are solicited in 2 categories:
- Full-length papers presenting novel research results: Full-length papers will be published in the workshop proceedings. Submitted papers must describe original work that has not appeared in, nor is under consideration for, another conference or journal. Papers shall be eight (8) pages minimum and not exceed ten (10) pages including text, appendices, and figures, but excluding bibliography and acknowledgments. Appendix pages related to the reproducibility initiative dependencies, namely the Artifact Description (AD) and Artifact Evaluation (AE), are not included in the page count.
- Extended abstracts summarising preliminary/published results: Extended abstracts will be evaluated separately and will not be included in the published proceedings; they are intended to propose timely communications of novel work that will be formally submitted elsewhere at a later stage, and/or of already published work that would be of interest to the PAW-ATM audience in terms of topic and timeliness. Extended abstracts shall not exceed four (4) pages.
When deciding between manuscript submissions with similar merit, submissions whose focus relates more directly to the key themes of the workshop (application studies, computing at scale, high-level alternatives to MPI+X) will be given priority over those that don't.
Submissions shall be submitted through Linklings using 10pt font in the IEEE format.
PAW-ATM follows the reproducibility initiative of SC23. Submissions shall include an Artifact Description (AD) appendix. The appendix pages related to the reproducibility initiative dependencies are not included in the page count.
AI-SPRINT at the 6th Annual Parallel Applications Workshop, Alternatives To MPI+X
Daniele Lezzi, Senior Researcher at Barcelona Supercomputing Center and member of the technical team in AI-SPRINT, is one of the Programme Committe Chairs of the workshop.