Aug 2026· Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication· 0 citations· 27 references
Computer Science
TL;DR
VibeNIC is a SmartNIC framework co-designed at every layer for an LLM developer, and a case study on a stateful HBM-augmented UDP datapath delivers a working end-to-end design in hours.
Abstract
FPGA SmartNICs combine wire-speed performance with full datapath customizability, yet remain among the hardest to program. LLM coding agents ease development but stumble on system-scale FPGA design: physical overhead and long toolchain runtime, far from the network function, degrade convergence and design quality across the agentic loop. While agentic development tolerates higher intrinsic complexity, a framework must eliminate the incidental complexity of FPGA design-space exploration. VibeNIC is a SmartNIC framework co-designed at every layer for an LLM developer. Its shell absorbs board-level complexity into a timing-closed base NIC behind standard interfaces, shrinking each iteration to a user-partition-only recompile and leaving most design space reconfigurable. A building-block library atop it lets the agent compose rather than microarchitect, grounded by documentation, examples, and prompts (DEPs) in shell and library contracts. A case study on a stateful HBM-augmented UDP datapath delivers a working end-to-end design in hours.
A unified analysis of attacks on chiplet systems, the integration of Large Language Models into Electronic Design Automation (EDA) flows, and how LLM systems can advance hardware security efforts for modern systems, including chiplets is provided.
J. Knechtel, Ozgur Sinanoglu, Paul V. Gratz et al.· 0 citations
FPGAgent is the first task-specification-to-executable HLS generation framework experimentally validated on a well-established benchmark, and the value of end-to-end validation is demonstrated.
Tian-Yun Wang, Wenjie Wang, Jian-Guo Yao et al.· 0 citations
Can large language models generate not just correct, but fast hardware? This paper investigates the question in financial FPGA design, where 5-10 nanoseconds of latency determines competitive advantage and designs iterate continuously as protocols, strategies, and regulations evolve. FinHardBench, a benchmark of 33 fin...
Weimin Fu, He-Jia Zhang, Ming-Hao Shao et al.· 0 citations
Shift-left DFT methodologies enable early insertion of test logic at the RTL stage, but conventional flows rely on DFT EDA tools to directly insert CODECs, TDRs, MBIST, scan routers, pin multiplexing, boundary scan, and TAP controllers into design RTL, enforcing serial workflows between design and DFT teams, resulting...
Janus is presented, an LLM-assisted framework that synthe-sizes custom instructions integrated into the Ibex RISC-V core while keeping correctness outside the agent, demonstrating a practical path for using LLMs to explore ISA specialization without making the agent part of the trusted correctness boundary.
This paper presents LithOS Bare-Metal (LithOS-BM), a Real-Time Operating System (RTOS) partially compliant with the ARINC 653 APEX standard, designed for small embedded systems where partitioning or virtualization is unnecessary or infeasible, and rapid response times are critical.Integrated Modular Avionics consolidat...
Miguel Gotor-Ramos, Carlos Cuesta-Martínez, Miguel Masmano-Tello et al.· Jornadas de Automática· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.