skills.rest
mlc-ai
mlc-aiOFFICIAL
Explore all skills by this author →
GitHub

analyze-nsys-profile

Query Nsight Systems profiles to measure compute/communication overlap and inspect per-rank streams.

327|31|Updated Mar 21, 2026
Search Tags:#profiling#gpu#overlap#trace-analysis#nsys#trace-diagnostics
One-click install
npx skills add https://github.com/mlc-ai/pith-train --skill analyze-nsys-profile
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: analyze-nsys-profile
Source: https://github.com/mlc-ai/pith-train/tree/main/.agents/skills/analyze-nsys-profile
Command: npx skills add https://github.com/mlc-ai/pith-train --skill analyze-nsys-profile

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Query a captured Nsight Systems profile to measure compute/communication overlap and inspect per-rank streams.

Core Features & Use Cases

  • Targeted analysis primitives to measure overlap and categorize streams.
  • Bottleneck localization by identifying exposed communication in DualPipeV stages.
  • Reusable workflow for traces from workspace/capture-nsys-profile.

Quick Start

Provide a captured .nsys-rep file (e.g., workspace/capture-nsys-profile/pithtrain_node0.nsys-rep) and ask for overlap analysis and bottleneck identification.

Frequently Asked Questions about analyze-nsys-profile

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

FAQPage Schema
How do I measure compute and communication overlap in an Nsight Systems trace?▼

You can analyze compute and communication overlap in an Nsight Systems trace by querying a captured .nsys-rep file. The Skill exports the profile to SQLite and applies targeted analysis primitives to measure overlap and inspect per-rank streams.

How do I locate exposed communication bottlenecks in DualPipeV stages?▼

You can locate exposed communication bottlenecks in DualPipeV stages by running trace diagnostics on the .nsys-rep file. The Skill applies targeted analysis primitives to identify exposed communication and pinpoint bottlenecks within the pipeline stages.

What is the best way to inspect per-rank streams from an .nsys-rep file?▼

Inspecting per-rank streams from an .nsys-rep file is done by exporting the trace to SQLite using provided scripts. This enables the construction of targeted queries to categorize streams and assess overlap across ranks.

Do I need a specific file format to analyze GPU trace diagnostics?▼

Yes, you need a captured .nsys-rep file located in the workspace/capture-nsys-profile directory. This specific Nsight Systems profile format is required to export data to SQLite and perform the overlap analysis.

Related Skills

Explore More Data & Analytics Skills→
yo-stevenyo-steven

perf-nsight-systems

Profile CPU, GPU, and communication timelines from .nsys-rep captures.

Community
Intermediate
michaeldubumichaeldubu

perf-analyzer

Parse and analyze CUDA, CPU, and memory benchmark outputs across CSV, JSON, and text formats.

Community
Advanced
GindaChenGindaChen
71

nsys-ai

Analyze NVIDIA Nsight Systems profiles to identify GPU bottlenecks and propose fixes.

Community
Advanced
m0atm0at
1

cuda

Analyze CUDA kernel performance bottlenecks with Nsight Systems and Nsight Compute.

Community
Advanced