Workshops Archive

Porting Codes to the NextSilicon Maverick-2 Dataflow Architecture: Early Experience, Best Practice and Performance


Workshop: Twelfth International Workshop on Heterogeneous High-Performance Reconfigurable Computing (H2RC 2026)

Authors: Chekuri Choudary (NextSilicon Inc.); Steffen Christgau, Daniel Gecasek, and Ronaldo Rodrigues Pela (Zuse Institute Berlin); Kevin Olson and Mathias Pütz (NextSilicon Inc.); and Florian Schintke, Lewin Stein, and Thomas Steinke (Zuse Institute Berlin)

Abstract: Porting parallel code to accelerators such as GPUs, FPGAs, or AI Accelerators often requires major changes. In contrast, the new NextSilicon Maverick-2 dataflow architecture and framework transparently accelerate OpenMP code without explicit offloading statements, offering a practical energy-efficient high-performance computing (HPC) platform. We present preliminary best practice, porting experience, and early performance results for the Maverick-2 architecture, its programming models, and the OpenMP-based code-to-device development workflow. The in-house proxy apps used in the benchmark cover several application domains typical for the tier-2 HPC system at NHR@ZIB: computational fluid dynamics, large-eddy simulations, molecular dynamics, and electronic structure theory.


Back to Twelfth International Workshop on Heterogeneous High-Performance Reconfigurable Computing (H2RC 2026) Archive Listing

Back to Full Workshop Archive Listing