メニュー
Use when running GPU programs against cosim and classifying outcomes. Requires exact program identity, fresh-session runner use, verdict artifacts, and provenance checks. After cosim-gpu-flow-plan for planned test tasks.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use as the main entry for cosim debugging: gem5 crashes, assertions, hangs, timeouts, address translation faults, non-PASS test results, guest-side waits, and QEMU exits that may be secondary to gem5. Collect logs, preserve live state when needed, route to gem5 model references, and decide the first failing component.
Use when building gem5 or QEMU, checking build provenance, or preserving gem5 incremental builds across worktrees. Delegates to scripts/cosim_build.sh; never hand-roll Docker, scons, or QEMU make commands.
Use when delegating independent code review, design review, or focused bug investigation to an AI reviewer. Provides delegation templates and records reviewer evidence; higher-level review flow remains in cosim-gpu-review or cosim-gpu-rlcr-loop.
Use when the guest disk image needs modification without a full rebuild. Mount with guestmount, add files, systemd services, kernel module config.
Use as the first step for non-trivial cosim testing, debugging, review, or iterative implementation tasks. Do not use for ordinary repository maintenance such as commit splitting, submodule pointer updates, author config, or documentation-only cleanup.
Use when you need to send commands to a running cosim guest. Console pipe interaction, 9p share mount, GPU status check, manual driver load.
| name | cosim-gpu-test |
| description | Use when running GPU programs against cosim and classifying outcomes. Requires exact program identity, fresh-session runner use, verdict artifacts, and provenance checks. After cosim-gpu-flow-plan for planned test tasks. |
Execute GPU programs against cosim and classify results. The test runner owns launching, guest execution, artifact routing, source snapshot, provenance, matrix rows, verdicts, and cleanup.
Use cosim-gpu-launch only for manual boot or interactive inspection. Use this
skill whenever a program result, matrix row, verification result, or PR-grade
test artifact is required.
bash scripts/run_cosim_tests.sh vector_add
bash scripts/run_cosim_tests.sh --all
bash scripts/run_cosim_tests.sh HIP-Basic/FloydWarshall
bash scripts/run_cosim_tests.sh --test-timeout 120 --output-dir artifacts/<slug> vector_add
bash scripts/run_cosim_tests.sh --hang-env --output-dir artifacts/<slug> bit_extract_sync
scripts/run_cosim_tests.sh has two modes:
pure_test: fresh cosim session, guest workload, classification, artifact
archive, cleanup.hang_env: launch cosim, mount the shared repository, build the target,
write category=hang_env_ready, and leave the console pipe alive for
cosim-gpu-guest.Normal tests always archive evidence and clean the live session.
Before launching a named program, prove exact program identity. Read
references/program-identity-and-guest-env.md for local kernels, ROCm
examples, variants, timeout estimation, device-side printf, or
GUEST_TEST_PREFIX.
Before automated rows, repeats, environment matrices, non-default binaries, or
evidence intended to prove uncommitted source, read
references/run-manifest-and-provenance.md.
For long diagnostic rows, performance rows, matrix runs, or interrupted test
automation, read references/diagnostic-and-matrix-runs.md.
Use cosim-gpu-build for build readiness and binary provenance decisions. Do
not duplicate build metadata rules here.
When program identity, manifest rows, and required provenance checks are
complete, run scripts/run_cosim_tests.sh directly. Do not ask before standard
test execution, repeats, interrupt-mode matrix rows, timeout probes, post-fix
reruns inside the current scope, standard incremental build, or standard
preflight cleanup.
Ask only before changing target program, changing fixed environment values, using a nonstandard binary, full or cold rebuilds, deletion outside repository cleanup scripts, or switching from normal test mode to live debug mode when the target is not already known to hang.
Tests are classified by scripts/classify_runs.py --json. The agent must read
[COSIM_VERDICT] or verdict.json from the test artifact and report the
outcome verbatim. Do not override verdict artifacts by manually interpreting
logs.
| Outcome | Meaning | Action |
|---|---|---|
PASS | Program ran, output validated, exit 0 | Record and proceed |
FAIL | Anything not PASS | Route by verdict reason |
Use verdict reason for routing. Examples: timeout, wrong result, simulator
death, boot timeout, or invocation error. If QEMU or gem5 exits before the
guest test starts, treat it as launcher/backend evidence, not a program result.
For existing logs:
python3 scripts/classify_runs.py --log-dir artifacts/<slug>/logs [--matrix artifacts/<slug>/matrix.tsv]
cosim-gpu-flow-plan --type testcosim-gpu-buildbash scripts/cosim_cleanup.shcosim-gpu-debugartifacts/<slug>/, matrix rows,
guest logs, gem5 log, QEMU log, verdict, source snapshot, and binary
provenanceFinal matrices must prove program identity, outcome, guest log, artifact path, gem5 commit, gem5 binary path, and gem5 binary hash for every accepted row. Manifest values are pre-run contracts, not proof of execution.