skills.rest
lightrain-a
lightrain-aCOMMUNITY
Explore all skills by this author →
GitHub

system-profile

Profile performance targets and emit structured bottleneck reports.

Updated May 20, 2026
Search Tags:#reporting#instrumentation#gpu#performance profiling#bottleneck analysis#interconnect#memory overhead
One-click install
npx skills add https://github.com/lightrain-a/medtrace-aris --skill system-profile-lightrain-a
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: system-profile
Source: https://github.com/lightrain-a/medtrace-aris/tree/main/.vendor/aris/skills/system-profile
Command: npx skills add https://github.com/lightrain-a/medtrace-aris --skill system-profile-lightrain-a

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It helps you identify performance bottlenecks and wasted resources by profiling a target (script/process/GPU/memory/interconnect) and turning the results into actionable recommendations.

Core Features & Use Cases

  • End-to-end profiling guidance: Selects appropriate profiling strategies based on the target type (CPU, memory, GPU, interconnect, or a running service).
  • Tool + instrumentation approach: Uses external tools when available and writes targeted instrumentation code when external tooling alone is insufficient.
  • Structured performance reports: Produces reports covering CPU overhead, memory overhead, interconnect communication, and GPU compute to support clear next steps.
  • Use Case: When training and serving models, run a system profile on a specific PID or on “vllm serving” to find whether delays come from CPU overhead, memory replication, PCIe/NVLink transfer issues, NCCL communication inefficiency, or low SM utilization.

Quick Start

Run the system-profile skill and provide the target you want to analyze, such as a process ID, a training script name, or the phrase "vllm serving".

Frequently Asked Questions about system-profile

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I identify GPU compute bottlenecks and CPU overhead during model serving?▼

Performance profiling a running process involves selecting suitable external profiling tools or adding targeted code instrumentation to capture metrics. This approach surfaces CPU overhead, memory usage, GPU compute, and interconnect communication bottlenecks for actionable optimization recommendations.

What's the best way to analyze NCCL communication inefficiency and PCIe transfer issues?▼

Profiling interconnect communication and GPU compute requires parsing a user-provided target and selecting appropriate external profiling tools. The system-profile skill captures outputs to profile artifacts and emits structured results, including an instrumentation changelog for bottleneck analysis.

Can I profile a specific PID to find memory overhead and low SM utilization issues?▼

Yes, you can profile a specific PID by providing the process ID as the target. The skill parses the target, runs suitable external profiling tools or adds code instrumentation, and captures GPU compute and memory overhead metrics to produce structured performance reports.

How do I profile a training script for benchmarking and troubleshooting bottlenecks?▼

To profile a training script for benchmarking and troubleshooting bottlenecks, provide the script name as the target. The skill selects appropriate profiling strategies, applies targeted code instrumentation or external tools, and emits structured performance reports with an instrumentation changelog.

When do I need targeted code instrumentation instead of external profiling tools?▼

You need targeted code instrumentation for performance profiling when external tooling alone is insufficient to capture the required metrics. The skill uses external profiling tools when available and writes targeted instrumentation code to ensure comprehensive coverage of CPU, memory, GPU, and interconnect bottlenecks.

Related Skills

Explore More Software Engineering Skills→
duypham2801duypham2801

system-profile

Profile CPU, GPU, memory, and interconnect bottlenecks in system execution.

Community
Advanced
nateanatea
4

performance-analysis

Detect and profile performance bottlenecks in Claude Flow swarms.

Community
Advanced
jandan138jandan138

system-profile

Profile Python scripts, processes, GPU workloads, and interconnect topologies into structured reports.

Community
Advanced
zhuyingqinzhuyingqin
1

system-profile

Profile scripts, services, or GPU workloads to identify performance bottlenecks.

Community
Advanced