Code Profile Analyzer

Analyze execution-time and stack usage profiles for generated code

Since R2023a

expand all in page

Description

Use the Code Profile Analyzer app to analyze execution-time and stack usage profiles produced by software-in-the-loop (SIL) or processor-in-the-loop (PIL) execution of generated code:

  • View execution times of tasks and functions, visualize the function-call stack at a specific simulation time, and compare execution times from different runs.

  • View stack usage for tasks and functions and compare values from different runs.

If you are using Embedded Coder®, you can also:

  • Analyze execution-time profiles produced by XCP-based external mode simulations.

  • Configure code profiling for a model hierarchy.

  • Identify and analyze performance hotspots in the generated code. This workflow uses the coder.profile.test.runTests function.

  • Identify and analyze critical paths in the generated code. This workflow uses the coder.profile.test.analyzePath function.

Code Profile Analyzer app

Open the Code Profile Analyzer App

If you are generating code from a Simulink® model, click the SIL/PIL tab. In the Results gallery, under Execution Profiling Results or Stack Profiling Results, click the Code Profile Analyzer button.

If you are generating code from MATLAB® code, terminate the SIL/PIL execution in the Command Window. The code generator produces a message in the Command Window that contains a link to the execution or stack profiling report. Click that link.

Alternatively, for both Simulink and MATLAB workflows, open the Code Profile Analyzer from the Command Window after the SIL/PIL execution has been terminated by entering coder.profile.show or coder.profile.show(profileVariable).

For the Simulink workflow, profileVariable is a workspace variable (coder.profile.ExecutionTime or coder.profile.ExecutionStack object) that you specify through the Workspace variable (CodeExecutionProfileVariable) or Stack workspace variable (CodeStackProfileVariable) configuration parameter.

For the MATLAB workflow, obtain profileVariable by using the getCoderExecutionProfile or getCoderStackProfile function.

If the app is already open, you can use the General tab to import single simulation results by selecting the workspace variable that contains your execution-time or stack usage measurements.

You cannot use the General tab to import aggregated results from multiple simulations. Instead, use the command coder.profile.show(profileAggregate) where profileAggregate is a coder.profile.ExecutionTimeSet or coder.profile.ExecutionStackSet object.

Code Profile Analyzer Workflows

The type of simulation results that you import determines the tabs displayed by the app and the workflows that you can perform.

When you import results using a coder.profile.ExecutionTime object, the Time Profiling tab displays tabs as described in this table.

TabPurpose
Task Execution

Show execution-time metrics of tasks. For each task, inspect execution times (that is, minimum, average, and maximum execution times) and the corresponding CPU utilization, which enables you to determine whether overall the tasks can run on the target hardware without execution-time issues.

See View and Compare Code Execution Times.

Function Execution

Show execution-time metrics of generated code functions to identify hotspots and determine functions that require rework.

See View and Compare Code Execution Times.

Function-Call Stack

Show function-call stack at specific simulation time. If you identify that a task is taking longer than expected, you can display its function-call step for each execution instance. This feature helps you to identify performance bottlenecks.

See View and Compare Code Execution Times.

Comparison

Compare execution-time metrics of two simulations to identify performance regressions.

See View and Compare Code Execution Times.

Task Validation

Perform CPU feasibility analysis to determine formally whether the task set can run without execution-time issues. Analyze the effect of task-to-CPU mapping.

See Validate Task-to-CPU Mapping.

CPU Usage

For each task, display execution times across the available CPUs. This feature shows how tasks are allocated on the system and can help you to improve load balancing.

Available only when these conditions apply:

  • Simulation is XCP external mode

  • Multi-tasking execution is enabled

  • Target application runs on an operating system

See Validate Task-to-CPU Mapping.

When you import results using a coder.profile.ExecutionTimeSet object, the Time Profiling tab displays tabs as described in this table.

TabPurpose
Cumulative Results

Show task execution-time metrics from different tests. If you run a suite of test cases, for each task, you can identify the longest and average execution time. Then, you can identify the test case that produces the most demanding execution.

Flame Graph

Show flame graph for longest task executions. This view helps you to understand what happens during a task execution instance, enabling you to identify performance bottlenecks.

Comparison

Compare execution-time metrics of two simulations to identify performance regressions.

Task Validation

Perform CPU feasibility analysis to determine formally whether the task set can run without execution-time issues. Analyze the effect of task-to-CPU mapping.

See Validate Task-to-CPU Mapping.

When you import results using a coder.profile.ExecutionStack object, the Memory Profiling tab displays tabs as described in this table.

TabPurpose
Stack Memory

Show stack usage metrics of tasks and functions. For each task, view a summary of the minimum, average, and maximum values of stack memory usage. Within a task, see how much memory is required for each executed function.

See View and Compare Stack Usage Metrics.

Function-Call Stack

Show function-call stack at a specific simulation time. If you observe that memory allocation for a task is higher than expected, you can display its function-call stack for each execution instance. This feature helps you to identify performance bottlenecks.

See View and Compare Stack Usage Metrics.

Comparison

Compare stack usage metrics of two simulations to identify performance regressions.

See View and Compare Stack Usage Metrics.

When you import results using a coder.profile.ExecutionStackSet object, the Memory Profiling tab displays tabs as described in this table.

TabPurpose
Cumulative Results

Show task stack usage metrics from different tests. If you run a suite of test cases, for each task, you can identify the maximum value of stack usage. Then, you can identify the test case that produces the most demanding execution.

Comparison

Compare stack usage metrics of two simulations to identify performance regressions.

If your model is configured to produce Performance Monitoring Unit (PMU) metrics, then when you import results, the Code Profiling tab displays tabs as described in this table.

TabPurpose
Function Execution

Show PMU metrics of generated code functions to identify hotspots and determine functions that require rework.

See Generate Profiles Using PMU Metrics.

Comparison

Compare PMU metrics of two simulations to identify performance regressions.

See Generate Profiles Using PMU Metrics.

Whether you import data or not, you can also use the Code Profile Analyzer app to configure code profiling in a model hierarchy. See Configure Execution-Time Profiling in Model Hierarchy Using Code Profile Analyzer and Configure Stack Usage Profiling in Model Hierarchy Using Code Profile Analyzer.

Examples

expand all

This table gives further workflow examples.

ExampleUse the app to:
Aggregate Execution-Time Profiles to Identify Worst Execution

Process an aggregate of execution-time profiles. For example, identify:

  • The longest execution for each task

  • Performance hotspots in generated code

Analyze Execution Times of Critical Paths

View results of a critical path analysis of generated code.

Visualize Task Scheduling in XCP External Mode Simulation

View the overrun rate (number of overrun events / number of calls) as a percentage.

Aggregate Stack Usage Profiles to Identify Most Demanding Execution

Process an aggregate of stack usage profiles. For example, identify the most demanding execution for each task.

Related Examples

Version History

Introduced in R2023a

See Also