This is the GitHub organization for the MICAS research group at KU Leuven. Here you can find open-source projects made by us and our collaborators.
Snax is an end-to-end, open-source multi-accelerator cluster design framework. It consists of the snax-mlir compile toolchain and the snax cluster hardware framework.
SNAX-MLIR Repo
SNAX Cluster Repo
HeMAiA Repo
ZigZag Repo | Documentation | Tutorial
ZigZag targets rapid DSE for DNN accelerator platforms supporting an broad set of hardware architectures and workload scheduling scenarios beyond other existing frameworks.
The latest version of ZigZag includes support for modeling of analog and digital in-memory computing accelerators and estimating both peak/workload performance at the macro and system level.
Stream Repo | Documentation | Tutorial
Stream is an extension of ZigZag capable of modeling multi-core DNN acceleration employing fine-grained layer-fused processing.
ZigZag-LLM Repo
ZigZag-LLM is a framework to rapidly model Large Language Models on dedicated, single-core accelerators and facilitates early identification of energy bottlenecks within the hardware architecture.
DeFiNes Repo
DeFiNes extends ZigZag to enable the DSE of cross-layer depth-first scheduling (a.k.a. layer fusion, or cascaded execution)
HTVM Repo
HTVM is a neural network compiler based on Dory and TVM that allows for efficient neural network deployment on heterogenous TinyML platforms with scratchpad-memory accelerators.
Floating point units (add, mul, FMA) for transprecision computing in arbitrary FP formats, wrapped and tested in Chisel.
TinyVers Repo
TinyVers: A 0.8-17 TOPS/W, 1.7 μW-20 mW, Tiny Versatile System-on-chip with State-Retentive eMRAM for Machine Learning Inference at the Extreme Edge (VLSI Technology and Circuits'23)
DIANA: An End-to-End Energy-Efficient Digital and ANAlog Hybrid Neural Network SoC (ISSCC'22)
DIANA: An End-to-End Hybrid DIgital and ANAlog Neural Network SoC for the Edge (JSSC'23)
DPU Repo
DPU: DAG Processing Unit for Irregular Graphs With Precision-Scalable Posit Arithmetic in 28 nm (JSSC'22)
DPU-v2 Repo
DPU-v2: Energy-efficient execution of irregular directed acyclic graphs (MICRO'22)