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 |
scripts/verify_config.py
View on GitHub# -------------------------------------------------------------------------
# This file is part of the MindStudio project.
# Copyright (c) 2026 Huawei Technologies Co.,Ltd.
#
# MindStudio is licensed under Mulan PSL v2.
# You can use this software according to the terms and conditions of the Mulan PSL v2.
# You may obtain a copy of Mulan PSL v2 at:
#
# http://license.coscl.org.cn/MulanPSL2
#
# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
# See the Mulan PSL v2 for more details.
# -------------------------------------------------------------------------
"""配置生效校验 / 现状 gap 分析。
用法:
python3 verify_config.py <plan.json> [--threads] [--bind-compute]
plan-only 时输出"现状 vs 目标"的 gap(预期有 FAIL);
apply 后复跑,全部 PASS 才算落地。
线程校验在执行时枚举 /proc/<pid>/task 并按 comm 归类,不依赖 plan 生成时的 tid。
目标核先与 cgroup cpuset.effective 取交集,避免"假 FAIL / 假 PASS"。
本脚本只读(不修改亲和),无特权要求;未知参数经 argparse 直接报错,
避免静默忽略导致的"校验口径与预期不符"。
"""
import argparse
import json
import sys
import procfs
import roles
import topology
def _cpus(text):
return set(topology.parse_cpulist(text))
def _clip(target, cgroup_text):
cgroup = _cpus(cgroup_text) if cgroup_text else set()
if not cgroup:
return set(target), False
clipped = set(target) & cgroup
return clipped, clipped != set(target)
def _actual(pid):
status = procfs.read_status(pid) or {}
return _cpus(status.get("Cpus_allowed_list") or "")
def _parse_args(argv):
p = argparse.ArgumentParser(description="bind-tune 配置生效校验(只读)")
p.add_argument("plan", help="hostcpu_plan_*.json")
p.add_argument("--threads", action="store_true", help="逐线程校验(按 comm 归类)")
p.add_argument("--bind-compute", action="store_true",
help="强制校验 plan 标 inherit 的 compute 绑定")
return p.parse_args(argv)
def main(argv):
args = _parse_args(argv)
with open(args.plan, "r", encoding="utf-8") as f:
plan = json.load(f)
results = []
ok = True
for unit in plan.get("units") or []:
pid = unit.get("pid")
if not pid:
# 无 pid 无法校验,必须显式记失败,避免「零校验即 PASS」
ok = False
results.append({
"unit_id": unit.get("unit_id"),
"pid": None,
"role": "unit",
"expect": unit.get("cpuset") or "",
"actual": "",
"match": False,
"clipped": False,
"threads": [],
"error": "unit 无 pid(未启动或进程信息缺失),无法校验",
})
continue
binding = {b.get("role"): b for b in (unit.get("thread_binding") or [])}
cgroup = procfs.read_cgroup_cpuset(pid)
main_expect, main_clipped = _clip(
_cpus((binding.get("main") or {}).get("cpuset")
or unit.get("main_cpuset") or unit.get("cpuset")),
cgroup,
)
main_actual = _actual(pid)
main_match = main_actual == main_expect
if not main_match:
ok = False
results.append({
"unit_id": unit.get("unit_id"),
"pid": pid,
"role": "main",
"expect": topology.cpus_to_str(main_expect),
"actual": topology.cpus_to_str(main_actual),
"match": main_match,
"clipped": main_clipped,
"threads": [],
})
if not args.threads:
continue
tasks = [t for t in procfs.list_tids(pid) if t != pid]
for role in ("comm", "io", "compute"):
item = binding.get(role) or {}
enabled = item.get("apply") != "inherit" or args.bind_compute
if not enabled or item.get("apply") == "skip" or not item.get("cpuset"):
continue
expect, clipped = _clip(_cpus(item.get("cpuset")), cgroup)
mismatches = []
matched = 0
total = 0
for tid in tasks:
if roles.classify(procfs.read_comm(pid, tid)) != role:
continue
total += 1
actual = _actual(tid)
if actual == expect:
matched += 1
elif len(mismatches) < 10:
mismatches.append({
"tid": tid,
"comm": procfs.read_comm(pid, tid),
"actual": topology.cpus_to_str(actual),
})
role_match = total == 0 or not mismatches
if not role_match:
ok = False
results[-1]["threads"].append({
"role": role,
"expect": topology.cpus_to_str(expect),
"matched": matched,
"total": total,
"match": role_match,
"clipped": clipped,
"mismatches": mismatches,
})
if not results:
# 空计划不能按「零失败」判为通过:没有任何单元被校验过
print(json.dumps({"ok": False, "error": "计划中没有 unit,未执行任何校验", "results": []},
ensure_ascii=False, indent=2))
return 1
print(json.dumps({"ok": ok, "results": results}, ensure_ascii=False, indent=2))
return 0 if ok else 1
if __name__ == "__main__":
sys.exit(main(sys.argv[1:]))
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 |