699 citations
- Nvidia (United Kingdom)GB28 papers
- Stanford UniversityUS12 papers
- University of California, BerkeleyUS11 papers
- California Institute of TechnologyUS10 papers
- Seattle UniversityUS10 papers
- University of TorontoCA10 papers
- University of WashingtonUS10 papers
- Georgia Institute of TechnologyUS9 papers
- Massachusetts Institute of TechnologyUS9 papers
- Argonne National LaboratoryUS8 papers
- Lawrence Berkeley National LaboratoryUS8 papers
- Carnegie Mellon UniversityUS7 papers
9 papers · 1 filter
Wattchmen: Watching the Wattchers -- High Fidelity, Flexible GPU Energy Modeling
Brandon Tran, Matthias Maiterth, Woong Shin +2
Modern GPU-rich HPC systems are increasingly becoming energy-constrained. Thus, understanding an application's energy consumption becomes essential. Unfortunately, current GPU ener…
DEMOTIC: A Differentiable Sampler for Multi-Level Digital Circuits
Arash Ardakani, Minwoo Kang, Kevin He +4
Efficient sampling of satisfying formulas for circuit satisfiability (CircuitSAT), a well-known NP-complete problem, is essential in modern front-end applications for thorough test…
DRC-Coder: Automated DRC Checker Code Generation Using LLM Autonomous Agent
Chen-Chia Chang, Chia-Tung Ho, Yaguang Li +2
In the advanced technology nodes, the integrated design rule checker (DRC) is often utilized in place and route tools for fast optimization loops for power-performance-area. Implem…
Virgo: Cluster-level Matrix Unit Integration in GPUs for Scalability and Energy Efficiency
Hansung Kim, Ruohan Richard Yan, Joshua You +2
Modern GPUs incorporate specialized matrix units such as Tensor Cores to accelerate GEMM operations, which are central to deep learning workloads. However, existing matrix unit des…
FuseMax: Leveraging Extended Einsums to Optimize Attention Accelerator Design
Nandeeka Nayak, Xinrui Wu, Toluwanimi O. Odemuyiwa +3
Attention for transformers is a critical workload that has recently received significant "attention" as a target for custom acceleration. Yet, while prior work succeeds in reducing…
CiMLoop: A Flexible, Accurate, and Fast Compute-In-Memory Modeling Tool
Tanner Andrulis, Joel S. Emer, Vivienne Sze
Compute-In-Memory (CiM) is a promising solution to accelerate Deep Neural Networks (DNNs) as it can avoid energy-intensive DNN weight movement and use memory arrays to perform low-…