Menü
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Basierend auf der SOC-Berufsklassifikation
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
npx skills add https://github.com/ROCm/rocm-systems --skill validateDer Befehl bleibt in einer Zeile. Scrollen Sie horizontal, um ihn vor dem Kopieren vollständig zu prüfen.
Sie bevorzugen eine lokale Kopie? Laden Sie die Dateien herunter, die SkillsMP derzeit vorliegen.
SKILL.md wird angezeigt
| name | validate |
| description | Run full rocDecode build and test pipeline after code changes |
| allowed-tools | ["Bash(cmake *)","Bash(make *)","Bash(test/validate.sh *)","Bash(test/build_samples.sh *)","Bash(nproc)"] |
Run the full rocDecode validation pipeline. Execute these steps in order, stopping only
if a step fails. Run from the rocDecode project root (the directory containing this
repo's build/, samples/, and test/ directories).
IMPORTANT: Run each command as a SEPARATE Bash tool call. Do not chain commands with && or |.
The build derives its compiler, install prefix, and CTest data path from ROCM_PATH
(default /opt/rocm). If ROCm is installed elsewhere (e.g. a TheRock or custom build),
ROCM_PATH must be set. NOTE: skills run cmake in a NON-interactive shell that does not
source ~/.bashrc, so set ROCM_PATH in ~/.profile (or export it before launching Claude
Code) — otherwise the build silently falls back to /opt/rocm. Step 1 verifies this and
fails fast with guidance if the toolchain is missing.
Conformance test data (Phase 4) is located via the ROCDECODE_CONFORMANCE_DIR
environment variable, which defaults to $HOME/rocDecodeConformance. It must contain the
per-codec subdirectories AvcConformance, Av1Conformance, HevcConformance, and
Vp9Conformance. Codecs whose directory is missing are reported as WARNING and skipped,
so the pipeline still runs without the full data set.
Run: test/validate.sh --check-rocm
This confirms ROCM_PATH points at a usable ROCm install (the compiler exists) and that
the install prefix is writable before any build work — make install installs into
ROCM_PATH, so a read-only prefix like a system /opt/rocm would need sudo, which this
skill does not use. If it fails, fix ROCM_PATH (see the note above; point it at a
user-writable install) and re-run before continuing. Do not retry without changing
ROCM_PATH — if it cannot be resolved, stop and ask the developer.
The build/ directory is not checked into the repo, so on a fresh checkout it must be
created and configured first. cmake -B build -DENABLE_EXTENDED_TESTS=ON is idempotent —
it creates and configures build/ if missing, and is a cheap no-op if it is already
configured. -DENABLE_EXTENDED_TESTS=ON enables the additional FFmpeg-based CTest cases
(the CTest phase covers 6 tests without it, 15 with it); those extra tests are only added
when FFmpeg is found, so the flag is harmless on machines without FFmpeg.
Run these four commands separately:
cmake -B build -DENABLE_EXTENDED_TESTS=ONmake clean -C buildmake -j -C buildmake install -C build(make -j lets make choose parallelism without a $(nproc) command substitution,
which would otherwise trigger a permission prompt on every run.)
Run: test/build_samples.sh
Run: test/validate.sh
Report the final summary from validate.sh to the user.
Decode and triage HRR capture archives with full GPU replay by default (Linux). Builds hrr-playback when missing. On Windows: full native GPU replay via triage_archive.ps1 + ensure_playback.ps1 (PowerShell); Docker replay requires Linux or WSL2. Never edits source. Print finding summary in the chat reply.
Profiles a target application using rocprofv3 with program counter sampling enabled and then analyzes the results. Use when the user has a program that runs on an AMD GPU and asks to perform PC sampling, to determine the runtime performance characteristics of their application, to determine stall reasons, or to determine hotspots in the code.
Use when writing, planning, or improving unit tests for low-level transport or systems code — especially when reasoning about branch coverage, test gaps, identifying which uncovered paths are worth pursuing, or deciding when a feature's test suite is ready to merge.