Using PerFlow

There are two main ways for developers to implement specific analysis tasks with PerFlow’s APIs: using built-in models and customize models with built-in passes.

Using Models

Developers can directly use built-in models to obtain related performance analysis reports.

example1

Building Customized Models

PerFlow provides a built-in performance analysis pass library for building customized models. For scenarios where analysis tasks have already been designed, the following example shows a complete process of implementation.

example2

For scenarios in which developers do not know what analysis to apply, PerFlow supports an interactive mode. It is advisable to first use a general built-in analysis pass, such as hotspot detection. The output of the previous pass will provide some insights to help determine or design the next passes. Then analysts can add other analysis passes into models step by step. Finally, models are generated according to detailed analysis.

Writing Customized Passes

If built-in passes cannot satisfy the demands, developers need to write customized passes and combine these user-defined passes with other built-in passes to build models.

example3