IEEE Computer 2026 — "Bridging the Gap in Neuromorphic Co-Design with the SANA-FE Co-Simulation Framework"
Copyright (c) 2026 — The University of Texas at Austin. Produced under contract #2317831 to National Technology and Engineering Solutions of Sandia, LLC (contract No. DE-NA0003525 with the U.S. Department of Energy).
This artifact reproduces all four case studies from the paper using the SANA-FE architectural simulator and LASANA circuit surrogate models.
| # | Study | Script |
|---|---|---|
| 1 | Loihi concurrency | synapse.py |
| 2 | Loihi-Indiveri (analog neurons) | indiveri.py |
| 3 | Loihi-IMAC (analog synaptic crossbars) | crossbar.py |
| 4 | Loihi-Indiveri-IMAC (fully analog pipeline) | combined.py |
This artifact is available for download on GitHub. To install locally, run:
git clone https://github.com/SLAM-Lab/computer26-artifact.git computer26 && cd computer26
git submodule update --init --recursivedocker build -t computer26 .
docker run --rm -v $(pwd)/results:/code/results computer26Requires: cmake, g++, python3, and the packages listed in the Dockerfile.
bash build_and_run.shResults are written to results/.
.
├── build_and_run.sh # top-level build and run script
├── Dockerfile
├── data/ # pre-trained models and binary assets (read-only)
├── results/ # experiment outputs, generated at runtime
├── sana-fe/ # architectural simulator (subrepository)
└── lasana-plugins/ # LASANA plug-ins (subrepository)
build_and_run.sh respects two environment variables:
| Variable | Default | Description |
|---|---|---|
QUICK |
true |
Reduced run for faster results; set to false for full fidelity |
JOBS |
nproc |
Parallel make jobs |
QUICK=false docker run --rm -v $(pwd)/results:/code/results computer26Note that a full run may take a long time (>10 hours) to complete, even on a good machine.
| Tool | Purpose |
|---|---|
| SANA-FE | Neuromorphic architectural simulator |
| LASANA-Plugins | Hardware surrogate plug-ins |
| LASANA | Circuit surrogate modeling framework |