kali20gakki/msAgent / skills/host-cpu/msagent-perf-cpu-bind-tune
msagent-perf-cpu-bind-tune
Flagged昇腾 NPU 多卡任务的 Host CPU 绑核方案求解器:覆盖 PyTorch 训练/离线推理与 vLLM-Ascend、SGLang 推理服务的 rank/worker/engine 分核,基于 NUMA 拓扑与 卡↔进程映射输出可执行、可回滚、可验证的绑核计划(taskset 前缀 / cpuset / OMP 环境变量),自动判定事前(env 片段)与事中(在线 sched_setaffinity)双形态。 触发:绑核方案、核怎么分、per-rank 绑核、NUMA 绑定方案、cpuset 划分、多卡分核、 CPU range 冲突、TTFT 或吞吐受 CPU 影响。 不用于:绑核没生效/绑核异常诊断(→ perf-cpu-config-audit);运行期是否落核 (→ perf-cpu-tracing);变慢根因归因(→ perf-cpu-perf-triage)。
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.1 KB |
| references/checks.md | markdown | 1.4 KB |
| references/framework-profiles.md | markdown | 5.1 KB |
| references/mechanisms.md | markdown | 2.8 KB |
| references/permissions.md | markdown | 5.4 KB |
| references/process-roles.md | markdown | 2.8 KB |
| resources/plan_template.json | config | 4.7 KB |
| resources/report.css | other | 7.6 KB |
| resources/report_template.html | other | 3.9 KB |
| scripts/apply_plan.py | script | 11.8 KB |
| scripts/cpu_list.py | script | 1.4 KB |
| scripts/discover.py | script | 16.8 KB |
| scripts/emit_plan.py | script | 19.8 KB |
| scripts/feasibility.py | script | 4.7 KB |
| scripts/model.py | script | 9.5 KB |
| scripts/npu_query.py | script | 16.2 KB |
| scripts/plan.py | script | 10.7 KB |
| scripts/preflight.py | script | 5.0 KB |
| scripts/procfs.py | script | 9.0 KB |
| scripts/report.py | script | 31.0 KB |
| scripts/roles.py | script | 1.5 KB |
| scripts/topology.py | script | 16.8 KB |
| scripts/unit_alloc.py | script | 8.2 KB |
| scripts/verify_config.py | script | 5.7 KB |
| scripts/workspace.py | script | 2.5 KB |
| scripts/zone_alloc.py | script | 3.7 KB |
references/process-roles.md
View on GitHub# 进程/线程分类规则
## 进程类型
| 类型 | 识别 | 绑核决策 |
| --- | --- | --- |
| 训练主进程 | `npu-smi` 卡占用 + `cmdline` 含 `torchrun` / `msrun` / `verl` / `MindSpeed` | 业务区,按 rank 分核;RL 按 actor/rollout 角色分核 |
| 推理进程 | `npu-smi` 卡占用 + `cmdline` 含 `vllm serve` / `sglang.launch_server` | 业务区,按实例分核;调度/分词线程与计算核分离 |
| DataLoader worker | 主进程 fork,不占卡;comm 与主进程同为 `python`,**当前仅按线程 `comm` 关键字识别,PPID 识别待实现**;未占卡的 worker 不作为独立 unit | 同 node 独立 IO 子核集 |
| HCCL 通信线程 | 线程 `comm` 含 `hccl` | 通信核集(1~2 核) |
| NPU runtime 线程 | `comm` 含 `AOE_RTKB` | 与 comm 同核集;`acl_thread` 归 compute(见下表) |
| OMP 计算线程 | 主进程线程池,受 `OMP_*` 控制 | 计算主体 |
| 监控/采集 | `npu-smi` 自身、`node_exporter`、`telescope` | 干扰隔离区(仅同 cgroup 可迁) |
| 容器 runtime | `containerd-shim` / `runc` | 系统中断区,不主动迁移 |
| 日志/agent | `fluentd` / `filebeat` | 干扰隔离区 |
| 编译/交互 | `gcc` / `pip` / `jupyter` | 告警并建议移出,不自动处理 |
## 线程角色映射(discover.classify_thread)
| 角色 | comm 关键字 | 说明 |
| --- | --- | --- |
| `comm` | `hccl`、`aoe_rtkb` | HCCL 通信 + NPU runtime;独立 1~2 核 |
| `io` | `zmq`、`tokenizer`、`detokenizer`、`dataloader`、`network` | 消息队列/分词/数据加载;独立 1 核 |
| `compute` | `vllm`、`caffe`、`python`、`pt_autograd`、`acl_thread`、`compl`、`omp` | 计算主体 |
| `main` | 其他(`plog`、`trace`、`monitor`、`ras`、`watchdog` 等) | 主线程/运行时管理,保留 1 核 |
## 绑定执行语义
| 角色 | apply | 执行方式 |
| --- | --- | --- |
| main | exec | 绑主线程(`tid == pid`) |
| comm | exec | 执行时枚举 `/proc/<pid>/task`,comm 命中关键字后细绑 |
| io | exec | 同上 |
| compute | exec | 绑定到 compute 核集(与 `OMP_NUM_THREADS` 核数一致,OMP 在该核集内自绑);旧计划标 `inherit` 时用 `--bind-compute` 强制绑定 |
所有目标核先与 cgroup `cpuset.cpus.effective` 取交集;plan 不保存 tid 快照。
## Primary NPU 判定
`npu-smi info` 中同一 pid 可能出现在多张卡上。按进程内存取最大条目作为 primary NPU:
- 48GB 条目 = 主卡;
- 117MB 条目 = 控制/共享内存,不作为主卡;
- `npus` 保留全部卡列表,便于报告展示。
## 权限降级
- `numa_maps` 不可读(非进程 owner / 容器权限)→ `view_limited`,远端页占比标 `not_observable`;
- `Cpus_allowed_list` 不可读 → 该 unit 跳过,并在报告中标注;
- `npu-smi` 不可用/失败/无设备 → 直接 blocked,不生成 plan。
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |