gpu-kernel-bottleneck-analysis

Name: gpu-kernel-bottleneck-analysis
Availability: InStock
Author: Alibaba

Analyzes GPU kernel performance using Roofline models and profiling data.

71|22|Updated Jun 23, 2026

Search Tags:#performance #optimization #profiling #gpu #cuda #roofline

Ready to run

npx skills add https://github.com/alibaba/atrex-kernel-agent --skill gpu-kernel-bottleneck-analysis

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: gpu-kernel-bottleneck-analysis
Source: https://github.com/alibaba/atrex-kernel-agent/tree/main/skills/gpu-kernel-bottleneck-analysis
Command: npx skills add https://github.com/alibaba/atrex-kernel-agent --skill gpu-kernel-bottleneck-analysis

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This skill addresses the difficulty of identifying performance bottlenecks in GPU kernels by providing a structured, evidence-based diagnostic framework that moves beyond guesswork.

Core Features & Use Cases

Roofline Analysis: Calculate compute and memory utilization against theoretical hardware peaks.
Evidence-Based Diagnosis: Extract concrete metrics from profiling tools like NCU or ROCprof to pinpoint stalls, bank conflicts, or low occupancy.
Use Case: When a custom CUDA kernel underperforms, use this skill to determine if it is memory-bound or compute-bound and identify the specific ISA-level bottleneck preventing higher throughput.

Quick Start

Use the gpu-kernel-bottleneck-analysis skill to profile the current kernel and generate a report identifying the primary performance bottleneck.

Frequently Asked Questions about gpu-kernel-bottleneck-analysis

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

FAQPage Schema

How do I diagnose if my CUDA kernel is compute-bound or memory-bound?▼

To diagnose if a CUDA kernel is compute-bound or memory-bound, calculate its TFLOPS and bandwidth utilization against a hardware-specific Roofline model. This skill extracts concrete metrics from NVIDIA NCU profiling data to pinpoint the specific bottleneck.

What is GPU Roofline analysis and how does it identify performance stalls?▼

GPU Roofline analysis is a diagnostic framework that measures compute and memory utilization against theoretical hardware peaks. By parsing profiling summaries, it identifies specific ISA-level stalls, bank conflicts, or low occupancy preventing higher throughput.

Can I use this to profile AMD ROCprof data for kernel underperformance?▼

Yes, you can profile AMD ROCprof data to identify kernel underperformance. The analysis supports diagnostic workflows for both compute-bound and memory-bound kernels by parsing ROCprofiling summary files to extract actionable optimization evidence.

How do I find ISA-level bottlenecks preventing higher GPU kernel throughput?▼

To find ISA-level bottlenecks, execute the provided Python benchmarking tools to generate profiling data, then parse the summary files. This evidence-based approach moves beyond guesswork to pinpoint exact stalls or low occupancy issues.

Do I need to run Python benchmarking tools before analyzing GPU kernel performance?▼

Yes, you need to run Python benchmarking tools to generate the profiling summary files required for analysis. The skill parses these extracted metrics to determine if the kernel is memory-bound or compute-bound against the hardware Roofline.