소스 정보
- 저장소
- ROCm/rocm-systems
- 최근 소스 활동
- 2026년 8월 6일 15:04
- 감지된 SKILL.md 언어
- 영어
- 스타
- 494
- 포크
- 398
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
메뉴
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
SOC 직업 분류 기준
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/ROCm/rocm-systems --skill validate명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SKILL.md 표시 중
| 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.