Domain Project

We are pleased to announce that the DOMAIN project was presented at HiPEAC 2025 in the delegates’ poster session with a dedicated poster. The event provided an ideal opportunity to introduce our platform’s goals and technical architecture to the high-performance and embedded computing community.

The poster titled “DOMAIN: Taming Heterogeneous Computing Complexity with Full-Stack Governance of Domain-Specific Languages” highlighted our approach to managing the growing complexity of heterogeneous computing systems through a model-driven, DSL-centered workflow. We presented the full-stack design of the DOMAIN platform, from high-level language engineering down to execution on hybrid architectures combining CPUs, GPUs, and FPGAs.

Key aspects included our Model-to-Model and Model-to-Text transformation chains, which allow application-level DSLs to be automatically translated into architecture-specific binaries through a common intermediate representation based on the Actor Model. We also showcased the hybrid runtime environment, equipped with memory tracking, dynamic optimization, and decision-making modules driven by Answer Set Programming.

Preliminary results from a case study in speculative stream processing—comparing our system to Apache Spark using NYC Taxi trip data—demonstrated promising improvements in execution efficiency. We also presented our ongoing efforts to extend DOMAIN’s applicability to domains such as discrete-event simulation, biological reactions, and demographic modeling.

We thank the HiPEAC community for the interest and stimulating discussions, and we look forward to continued collaboration and feedback as the DOMAIN platform evolves.