The growing demand for performance, energy efficiency, and domain-specific acceleration has driven a shift toward heterogeneous computing platforms. While hyperscaler cloud platforms offer broad access to compute resources, they often lack the flexibility required by advanced research workloads. In response, research institutions are increasingly deploying self-managed heterogeneous clusters, which remain difficult to maintain, configure, and use effectively. The Heterogeneous Accelerated Compute Clusters (HACCs) initiative, supported by the AMD University Program, addresses this by equipping select institutions with clusters composed of CPUs, GPUs, and adaptive devices for high-performance computing and AI research. This paper presents two complementary tools developed for such systems: HACC Management and Development. HACC Management is an Ansible-based automation framework for provisioning and maintaining HACC-like infrastructure in a reproducible and scalable way. HACC Development ( hdev ) is a lightweight Bash-based CLI that streamlines the development and deployment of heterogeneous applications. It orchestrates both environment and infrastructure setup and provides ready-to-use templates—example projects for user reference—that enable rapid iteration on real hardware. We demonstrate the design and capabilities of both tools using ETH Zurich's HACC as a case study. Our approach reduces operational overhead, enables faster prototyping, and provides a foundation for community-driven expansion of heterogeneous workflow support.
Moya et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: