Chapter 18 of FreeRTOS: Multi-Core Scheduling (SMP Extension)

Table of Contents

• 1. Overview of SMP Architecture
• 1.1 Basic Concepts
• 1.2 FreeRTOS-SMP Features
• 2. Inter-Core Task Migration Experiment
• 2.1 Experiment Objectives
• 2.2 Experiment Steps
• 2.3 Experiment Result Analysis
• 3. Protection of Shared Resources in Multi-Core Systems
• 3.1 Common Issues
• 3.2 Comparison of Solutions
• 3.3 Example Code
• 4. Key API Analysis
• 4.1 SMP Extension APIs
• 4.2 Configuration Parameters
• 5. Performance Optimization Suggestions

1. Overview of SMP Architecture

1.1 Basic Concepts

#define configNUMBER_OF_CORES 2 // Dual-core configuration example

• Symmetric Multi-Processing (SMP) characteristics:
• All cores have equal access to shared memory
• Unified task scheduling view
• Dynamic load balancing

1.2 FreeRTOS-SMP Features

Feature	Description
Global Task Queue	All cores share the task ready queue
Affinity	Supports task binding to specific cores
Lock-Free Scheduling Algorithm	Reduces inter-core contention

2. Inter-Core Task Migration Experiment

2.1 Experiment Objectives

• Observe dynamic migration of tasks between cores
• Validate load balancing mechanism

2.2 Experiment Steps

void vTaskMigrationDemo(void* pvParams) {
    for (;;) {
        // Print current running core ID
        UBaseType_t coreId = xPortGetCoreID();
        printf("Task running on Core %d\n", coreId);
        vTaskDelay(pdMS_TO_TICKS(500));
    }
}

2.3 Experiment Result Analysis

• Migration Trigger Conditions:
• Uneven core load
• Task voluntarily relinquishing CPU
• Scheduler tick interrupt

3. Protection of Shared Resources in Multi-Core Systems

3.1 Common Issues

• Cache Coherency issues
• Priority Inversion amplified across cores
• Increased Probability of Deadlock

3.2 Comparison of Solutions

Method	Applicable Scenarios	Overhead
Spinlock	Short critical sections	Low
Mutex	Long critical sections	Medium
Inter-Core Interrupt (IPI)	Urgent synchronization needs	High

3.3 Example Code

// Using SMP-compatible mutex
SemaphoreHandle_t xCrossCoreMutex = xSemaphoreCreateMutex();

void vSafeAccessResource(void) {
    if (xSemaphoreTake(xCrossCoreMutex, portMAX_DELAY) == pdTRUE) {
        // Safely access shared resource
        xSemaphoreGive(xCrossCoreMutex);
    }
}

4. Key API Analysis

4.1 SMP Extension APIs

// Set task affinity
vTaskCoreAffinitySet(TaskHandle_t xTask, const UBaseType_t uxCoreAffinityMask);

// Get current core ID
BaseType_t xPortGetCoreID(void);

4.2 Configuration Parameters

# FreeRTOSConfig.h key configurations
#define configUSE_CORE_AFFINITY 1
#define configUSE_SMP 1
#define configNUM_CORES 2

5. Performance Optimization Suggestions

1. Cache-Friendly Design:

• Keep frequently accessed data in a single core’s cache
• Use <span>vTaskCoreAffinitySet</span> to limit migration

Lock Granularity Control:

• Fine-grained locks are better than coarse-grained locks
• Consider read-write locks as alternatives to mutexes

Monitoring Tools:

• Use <span>uxTaskGetSystemState()</span> to monitor load
• Track <span>xPortGetCoreID()</span> to analyze task distribution

Best Practice: For time-critical tasks, it is recommended to run on fixed cores to reduce migration overhead