Janus 4.0

Quantum-classical Heterogeneous Architecture and Flexible Scheduling — framework repository accompanying the ISCA 2026 tutorial.

Tutorial landing page: https://janusq.github.io/ISCA_2026_Tutorial/

Quickstart

docker pull janusq/janus4:isca2026
docker run --rm --platform linux/amd64 -p 127.0.0.1:8888:8888 janusq/janus4:isca2026

Open JupyterLab at http://localhost:8888/lab. The single Janus4 image contains the numbered topic tree and topic-specific kernels, including qtenon for Qtenon, artery for ARTERY, adaptiveqc for AdaptDQC, chocoq for Choco-Q, and qram for EXP-QRAM.

Apple Silicon / ARM Mac

The published tutorial image is currently linux/amd64 only because the bundled Qtenon RISC-V toolchain and Verilator simulator artifacts are built for x86_64 Linux. On Apple Silicon Macs, run the same image through Docker's cross-architecture emulation layer:

# Docker Desktop or Colima both work. With Colima:
colima start --cpu 4 --memory 8 --disk 40

docker pull --platform linux/amd64 janusq/janus4:isca2026
docker run --rm \
  --platform linux/amd64 \
  -p 127.0.0.1:8888:8888 \
  janusq/janus4:isca2026

Do not switch the command to --platform linux/arm64: there is no native ARM image for this tag. Inside the container, uname -m will report x86_64 even on an ARM Mac. That is expected. Normal tutorial notebook runs use checked-in Qtenon replay captures and a baked executed-notebook smoke cache, so the emulation overhead is modest; opting into fresh Qtenon live simulation with QTENON_RUN_LIVE_SIM=1 or QTENON_RUN_PAPER_EXPERIMENT=1 will be much slower.

Topics

• 2-qtenon/ — Topic 2: low-latency quantum-classical hybrid control on a RISC-V + RoCC accelerator. Hybrid-loop and paper-figure walkthroughs use checked-in replay captures by default, with a baked notebook smoke cache in the Janus4 image.
• 3-adaptDQC/ — Topic 3: adaptive distributed quantum computing with local Qiskit simulation in the adaptiveqc kernel.
• 4-EXP-QRAM/ — Topic 4: bucket-brigade QRAM circuit construction, simulation, and processed data visualization in the qram kernel.
• 5-artery/ — Topic 5: ARTERY-style low-latency quantum feedback with S21/IQ analysis, trajectory prediction, compressed tutorial data in the artery kernel, hardware interface notes, and GUI/UDP demos.
• 6-Choco-Q/ — Topic 6: constrained binary optimization with Choco-Q in the chocoq kernel.

Adding a new topic

1. Create a new top-level directory using the tutorial number and topic name (e.g. 7-new-topic/).
2. Include a README.md, your source code, and a topic environment recipe or kernel setup that can be folded into the root Dockerfile.
3. Open a pull request against main.

The framework-wide tutorial Docker image is built from the root Dockerfile:

docker build --platform linux/amd64 -t janusq/janus4:isca2026 .

Related repositories

• JanusQ/JanusQ — Janus 3.0 framework (HPCA 2025 tutorial)
• JanusQ/HPCA_2025_Tutorial — Janus 3.0 tutorial site
• JanusQ/ISCA_2026_Tutorial — Janus 4.0 tutorial landing site (React SPA)

License

MIT. See LICENSE.