How to set up performance benchmarking and profiling on an STM32 MCU

Goal
Measure execution time and optimize code performance on an STM32(What is STM32?) microcontroller using:

• Cycle counters
• GPIO pin toggling
• STM32CubeIDE's built-in profiler (SWV/SWO)

Required Tools

Method 1: Cycle Counting Using DWT (Data Watchpoint & Trace)
Enable DWT Cycle Counter (for Cortex-M3/M4/M7)

c

#include "core_cm4.h"  // Adjust to your core type

void DWT_Init(void) {
  CoreDebug->DEMCR |= CoreDebug_DEMCR_TRCENA_Msk;  // Enable trace
  DWT->CYCCNT = 0;                                 // Reset cycle counter
  DWT->CTRL |= DWT_CTRL_CYCCNTENA_Msk;             // Enable counter
}

Example Usage

c

DWT_Init();

uint32_t start = DWT->CYCCNT;
myFunction();
uint32_t end = DWT->CYCCNT;

printf("Cycles: %lu\n", end - start);

Multiply by 1 / SystemCoreClock to convert cycles to time.

Method 2: GPIO Toggling + Oscilloscope/Analyzer
Example

c

HAL_GPIO_WritePin(GPIOB, GPIO_PIN_0, GPIO_PIN_SET);  // Start timing
myFunction();
HAL_GPIO_WritePin(GPIOB, GPIO_PIN_0, GPIO_PIN_RESET); // Stop timing

You can observe the pin on a logic analyzer to measure function duration down to microseconds or less.

Method 3: STM32CubeIDE SWV/SWO Profiling
Requirements:

• MCU with SWO pin
• Use ST-Link or J-Link debugger
• Connect SWO pin (often on PB3 or similar, depending on your MCU)

Steps:

1. Enable SWV I/O in STM32CubeMX (under SYS > Debug > Serial Wire)

2. In Project Settings > Debug Configuration:

• Go to SWV Settings
• Enable Trace and Profiling

1. Launch debugger and open:

• SWV Console
• SWV Data Trace / Performance Analyzer

Profiling Views:

• Function Execution Time
• Instruction Count
• CPU Usage over Time

Optimization Tips Based on Profiling

Optional: Use CoreMark or Dhrystone Benchmarks

• These industry benchmarks help compare MCU performance objectively.
• You can run CoreMark on STM32 and see DMIPS/MHz.
• ST sometimes provides optimized CoreMark examples in STM32Cube repos.

Summary