Introduction

Currently, MindStudio Advisor provides self-owned and ecosystem repositories to analyze the performance of models and operators, and enables one-click performance tuning to implement efficient performance tuning capabilities.

• Self-owned repositories provide the following functions: Roofline model-based operator bottleneck identification and tuning suggestion, timeline-based AI CPU operator tuning, operator fusion recommendation, TransData operator identification, and operator tuning analysis.
• The performance tuning functions of ecosystem repositories are developed by ecosystem developers using Python. You can call the corresponding API of Advisor to analyze the performance of models and operators provided by ecosystem developers.

  The current version of MindStudio IDE supports only the function of creating ecosystem repositories. You can develop the ecosystem repository code in the IDE. The Advisor-based analysis on ecosystem repositories is coming soon.

This sample describes how to use MindStudio Advisor to identify performance bottlenecks of models and operators and provide tuning suggestions.