skills.rest
kindofluke
kindoflukeCOMMUNITY
Explore all skills by this author →
GitHub

profile-analyzer

Analyze Sirius GPU profiles from Nsight Systems to identify bottlenecks and regression risks.

Updated Mar 18, 2026
Search Tags:#gpu#memory-bandwidth#profile-analysis#nsys#nsight-systems#kernel-occupancy#operator-attribution
One-click install
npx skills add https://github.com/kindofluke/albatross-data --skill profile-analyzer-kindofluke
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: profile-analyzer
Source: https://github.com/kindofluke/albatross-data/tree/main/.vibe/skills/profile-analyzer
Command: npx skills add https://github.com/kindofluke/albatross-data --skill profile-analyzer-kindofluke

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Sirius GPU profiling from Nsight Systems data helps identify bottlenecks and regression risks across runs.

Core Features & Use Cases

  • Generate GPU performance reports including kernel occupancy, memory bandwidth, and operator attribution.
  • Compare profiled and non-profiled runs to detect regressions and validate improvements.
  • Workflow guidance for full analysis and quick inspection of existing profiles.

Quick Start

Run the profiler on your Nsight Systems data to generate a comparative report and identify kernel hotspots.

Frequently Asked Questions about profile-analyzer

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

FAQPage Schema
How do I analyze NVIDIA Nsight Systems profiles to identify GPU bottlenecks?▼

Analyze NVIDIA Nsight Systems profiles by evaluating kernel occupancy and memory bandwidth to identify GPU bottlenecks. This process parses Sirius workload data to pinpoint operator attribution issues and detect regression risks across different execution runs.

What is the best way to compare GPU profile runs to detect performance regressions?▼

The best way to detect performance regressions is by comparing profiled and non-profiled GPU runs. This comparison highlights variations in kernel occupancy and memory bandwidth, validating whether recent changes introduced regression risks or performance improvements.

How does operator attribution work when analyzing Sirius GPU performance profiles?▼

Operator attribution works by mapping GPU performance metrics to specific operators within Sirius workloads. Analyzing Nsight Systems profiles isolates individual kernel execution times and memory bandwidth usage, identifying which operators cause the most significant bottlenecks.

Can I generate machine-readable metadata alongside human-readable GPU performance reports?▼

Yes, you can generate machine-readable metadata alongside human-readable GPU performance reports. Analyzing Nsight Systems profiles outputs a comprehensive report.md for users and summary.json plus metadata.json for automated pipeline processing.

Do I need Nsight Systems data to analyze kernel occupancy and memory bandwidth?▼

Yes, you need Nsight Systems data to analyze kernel occupancy and memory bandwidth. The analysis requires Nsight profile data specifically captured from Sirius workloads to accurately evaluate performance bottlenecks and operator attribution.

Related Skills

Explore More Data & Analytics Skills→
HKUDSHKUDS
46.8k

cli-anything-nsight-graphics

Automate GPU capture, trace, and analysis via NVIDIA Nsight Graphics CLI.

Official
Intermediate
KYRIE66nbKYRIE66nb

configure-telegram

Configure Telegram bot notifications for OMX session events and tasks.

Community
Intermediate
sirius-dbsirius-db
1.0k

profile-analyzer

Analyze NVIDIA Nsight Systems profiles for GPU performance and regressions.

Official
Advanced
liangdabiaoliangdabiao
264

funnel-analysis

Analyze multi-step funnels to compute conversion rates and identify bottlenecks.

Community
Advanced