Document Dragon GPU sharing, MPS and MPI execution

[tpn]

Summary

Explain how cuPhoton assigns work to GPUs, uses Dragon queues and MPI collectives, and validates completion. Cover optional GPU sharing, MPS connection checks and external service ownership. Add five diagrams and links to the implementation, and preserve the existing Dragon transport measurements in a separate page.

Why this is needed

The launch guide shows how to run a job. These pages explain what its processes do, which GPU each worker uses, and where the data goes. They distinguish multiple workers sharing a GPU from the default placement, and the shared worker lifecycle from the separate xPois paths.

Validation

• Compared the descriptions with the executors and component adapters; rendered and visually checked all five Mermaid diagrams.
• Checked 92 relative links and heading anchors.
• make ci-lint, make test-cpu and make build passed during the original documentation validation: 2,862 CPU tests passed, 210 skipped for GPU/optional-runtime support. The run emitted 40 Astropy FITS keyword warnings.
• Documentation-only change relative to main. No new MPS performance or multi-node qualification.

Repository hygiene

• No credentials, private paths, local datasets, notebook outputs, or generated run artifacts were added.
• Large binary artifacts are excluded or intentionally tracked through an approved storage plan.
• Security-sensitive information is not included.
• User-visible behavior changes are documented.
• New dependencies and their licenses are documented (none added here).

[coderabbitai]

Important

Review skipped

Auto reviews are limited based on label configuration.

🏷️ Required labels (at least one) (1)

• ai-review

---

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

• Configuration used: Repository: NVIDIA/cuPhoton/.coderabbit.yaml
• Review profile: CHILL
• Plan: Enterprise
• Run ID: 97a69ef2-0827-4f33-bd4f-3977b37139d7

---

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

_{Comment @coderabbitai help to get the list of available commands.}

[melo-gonzo]

LGTM, with one nonblocking correction inline about MPI import ordering. The remaining reviewed architecture descriptions and local links match the implementation.