Imported from MicroYY/GPUDriverLearningRoadmap (AGENTS.md). Install upstream with npx skills add MicroYY/GPUDriverLearningRoadmap. Copyright stays with the author.

Agent Instructions

Project Overview

This repository is a GPU driver learning roadmap organized as course-style documentation:

• docs/ contains the preface, long-term roadmap, source-reading checklist, and weekly retrospective template.
• sessions/ contains numbered learning sessions.
• labs/ contains numbered labs that correspond to the sessions.
• sources/ contains upstream source code managed as Git submodules (linux, mesa, libdrm, and igt-gpu-tools).

The main repository is documentation-first. Treat sources/ as reference material unless the user explicitly asks to inspect or modify submodule contents.

Working Guidelines

• Preserve the existing Chinese documentation style in README.md, docs/, sessions/, and labs/.
• Keep session and lab numbering aligned. If adding or renaming a session or lab, update the matching index files and roadmap links.
• Prefer small, reviewable Markdown edits. Do not reformat unrelated sections.
• Use relative Markdown links for files inside the repository.
• Be careful with sources/: it is large and externally maintained. Avoid broad searches there unless source inspection is required; prefer scoped paths such as sources/linux/drivers/gpu/drm when searching.
• Do not commit generated build outputs, editor files, or local environment directories. Follow .gitignore.

Session and Lab Writing Standard

• Treat sessions/ as lecture notes only.
• Session files should focus on explanation, concepts, boundaries, call paths, and source-reading guidance. Prefer more substantive teaching content over task-management prose.
• Session files should link to the corresponding lab, but should not include execution-oriented sections such as deliverables, checklists, completion criteria, or duplicated lab steps.
• Treat labs/ as hands-on execution pages only.
• Lab files should be concise code-experiment pages. Prefer a single 代码实验 section with concrete files, commands, and TODO order.
• Put the actual experiments and TODO order before references, setup notes, run commands, or background context.
• For code labs, include a short 相关代码 pointer under each experiment when helpful, naming the relevant files, structs, and test functions students should inspect before implementing.
• Prefer concrete commands, code, file paths, and executable steps in labs instead of generic advice.
• Avoid sections such as deliverables, checklists, completion criteria, recommended output formats, note templates, or next-step summaries unless the user explicitly asks for them.
• When a lab has a code/ directory, keep student-facing exercise instructions and run commands in the main lab Markdown file rather than a separate code/README.md, unless the user asks for a standalone code README.
• Avoid repeating the same structure across session and lab. In particular, do not duplicate deliverables, checklists, completion standards, or broad learning summaries in both places.

Useful Commands

There is no package manager or project test suite configured for this repo. For documentation changes, use focused checks:

git status --short
rg -n "term-or-link" README.md docs sessions labs

Submodules are initialized with:

git submodule update --init --depth 1 --filter=blob:none

For a smaller user-space source checkout:

git submodule update --init --depth 1 --filter=blob:none sources/mesa sources/libdrm sources/igt-gpu-tools

Windows native filesystems may warn about case conflicts in a full Linux kernel checkout. Prefer WSL/Linux filesystems for sources/linux, or use the sparse checkout described in README.md.

AGENTS.md Maintenance

After any non-trivial project change, check whether this file should be updated.

Update AGENTS.md when a change affects:

• project structure or important directories
• install, development, test, build, sync, or deploy commands
• required environment variables, secrets, credentials, or proxy expectations
• external integrations or API workflows
• generated files or files agents should not edit
• coding conventions, architectural patterns, or testing expectations
• recurring pitfalls discovered while working on the project

Do not update AGENTS.md for small bug fixes, one-off implementation details, or information already obvious from nearby code, package scripts, or existing docs.