Before you begin, install VisualGDB 5.3 or later and ensure you have the latest version of the STM32 package and OpenOCD.<\/p>\n
- \n
- Start Visual Studio and launch the VisualGDB Embedded Project Wizard:
![\"01-prjname\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/01-prjname.png\")
<\/a><\/li>\n- Select “New Project -> Embedded Binary” and click “Next”:
\n<\/a>![\"02-newprj\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/02-newprj.png\")
<\/a><\/li>\n- On the next page\u00a0choose an ARM toolchain and\u00a0select\u00a0the device you want to target. In this\u00a0tutorial we will use the STM32H7-Nucleo board that features the\u00a0STM32H743ZI microcontroller:
![\"03-device\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/03-device.png\")
<\/a><\/li>\n- On the next page select\u00a0“STM32CubeMX Samples -> FreeRTOS_ThreadCreation”:
![\"04-threads\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/04-threads.png\")
<\/a><\/li>\n- Connect your STM32H7-Discovery board to\u00a0your computer via USB:
![\"board\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/board.jpg\")
<\/a><\/li>\n- VisualGDB should automatically recognize the on-board ST-Link programmer and display it in the “Debug using” selector:
![\"05-stlink\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/05-stlink.png\")
<\/a>Once\u00a0the ST-Link is selected,\u00a0press “Finish” to\u00a0create the project.<\/li>\n- As of v1.2.0, the STM32H7 SDK from ST contains incorrect clock initialization code in the FreeRTOS example. To fix it, replace the SystemClock_Config() function with the following one taken from the HTTP Server example:\n
static void SystemClock_Config(void)\r\n{\r\n RCC_ClkInitTypeDef RCC_ClkInitStruct;\r\n RCC_OscInitTypeDef RCC_OscInitStruct;\r\n HAL_StatusTypeDef ret = HAL_OK;\r\n \r\n \/*!< Supply configuration update enable *\/\r\n MODIFY_REG(PWR->CR3, PWR_CR3_SCUEN, 0);\r\n \r\n \/* The voltage scaling allows optimizing the power consumption when the device is \r\n clocked below the maximum system frequency, to update the voltage scaling value \r\n regarding system frequency refer to product datasheet. *\/\r\n __HAL_PWR_VOLTAGESCALING_CONFIG(PWR_REGULATOR_VOLTAGE_SCALE1);\r\n \r\n while (!__HAL_PWR_GET_FLAG(PWR_FLAG_VOSRDY)) {}\r\n \r\n \/* Enable HSE Oscillator and activate PLL with HSE as source *\/\r\n RCC_OscInitStruct.OscillatorType = RCC_OSCILLATORTYPE_HSE;\r\n RCC_OscInitStruct.HSEState = RCC_HSE_BYPASS;\r\n RCC_OscInitStruct.HSIState = RCC_HSI_OFF;\r\n RCC_OscInitStruct.CSIState = RCC_CSI_OFF;\r\n RCC_OscInitStruct.PLL.PLLState = RCC_PLL_ON;\r\n RCC_OscInitStruct.PLL.PLLSource = RCC_PLLSOURCE_HSE;\r\n \r\n RCC_OscInitStruct.PLL.PLLM = 4;\r\n RCC_OscInitStruct.PLL.PLLN = 400;\r\n RCC_OscInitStruct.PLL.PLLP = 2;\r\n RCC_OscInitStruct.PLL.PLLR = 2;\r\n RCC_OscInitStruct.PLL.PLLQ = 4;\r\n \r\n RCC_OscInitStruct.PLL.PLLVCOSEL = RCC_PLL1VCOWIDE;\r\n RCC_OscInitStruct.PLL.PLLRGE = RCC_PLL1VCIRANGE_2;\r\n ret = HAL_RCC_OscConfig(&RCC_OscInitStruct);\r\n if (ret != HAL_OK)\r\n {\r\n asm(\"bkpt 255\");\r\n }\r\n \r\n \/* Select PLL as system clock source and configure bus clocks dividers *\/\r\n RCC_ClkInitStruct.ClockType = (RCC_CLOCKTYPE_SYSCLK | RCC_CLOCKTYPE_HCLK | RCC_CLOCKTYPE_D1PCLK1 | RCC_CLOCKTYPE_PCLK1 | \\\r\n RCC_CLOCKTYPE_PCLK2 | RCC_CLOCKTYPE_D3PCLK1);\r\n \r\n RCC_ClkInitStruct.SYSCLKSource = RCC_SYSCLKSOURCE_PLLCLK;\r\n RCC_ClkInitStruct.SYSCLKDivider = RCC_SYSCLK_DIV1;\r\n RCC_ClkInitStruct.AHBCLKDivider = RCC_HCLK_DIV2;\r\n RCC_ClkInitStruct.APB3CLKDivider = RCC_APB3_DIV2; \r\n RCC_ClkInitStruct.APB1CLKDivider = RCC_APB1_DIV2; \r\n RCC_ClkInitStruct.APB2CLKDivider = RCC_APB2_DIV2; \r\n RCC_ClkInitStruct.APB4CLKDivider = RCC_APB4_DIV2; \r\n ret = HAL_RCC_ClockConfig(&RCC_ClkInitStruct, FLASH_LATENCY_4);\r\n if (ret != HAL_OK)\r\n {\r\n asm(\"bkpt 255\");\r\n }\r\n \r\n \/*activate CSI clock mondatory for I\/O Compensation Cell*\/ \r\n __HAL_RCC_CSI_ENABLE();\r\n \r\n \/* Enable SYSCFG clock mondatory for I\/O Compensation Cell *\/\r\n __HAL_RCC_SYSCFG_CLK_ENABLE();\r\n \r\n \/* Enables the I\/O Compensation Cell *\/ \r\n HAL_EnableCompensationCell(); \r\n}<\/pre>\n<\/li>\n- Switch the project to the Release configuration and enable both RTOS tracing and function tracing via VisualGDB Project\u00a0Properties -> Dynamic\u00a0Analysis:
\n![\"06-addref\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/06-addref.png\")
<\/a>Proceed with referencing the profiler framework as VisualGDB suggests.<\/li>\n- If you try building the project now, the profiler framework will report\u00a0a missing USE_FREERTOS macro.\u00a0This happens because the sample project we used in this tutorial comes directly from the ST samples and does not contain VisualGDB-specific macros:
\n![\"07-mising\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/07-mising.png\")
<\/a><\/li>\n- Add the USE_FREERTOS macro via VisualGDB Project Properties -> MSBuild and the project will build successfully:
![\"08-macro\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/08-macro.png\")
<\/a><\/li>\n- Finally replace the main() function\u00a0and the 2 thread\u00a0functions with the following code:\n
#include \"main.h\"\r\n#include \"cmsis_os.h\"\r\n#include <math.h>\r\n\r\nosThreadId LEDThread1Handle, LEDThread2Handle;\r\nstatic void SenderThread(void const *argument);\r\nstatic void ReceiverThread(void const *argument);\r\nstatic void SystemClock_Config(void);\r\nstatic void CPU_CACHE_Enable(void);\r\n\r\nosSemaphoreDef(s_Semaphore);\r\nosSemaphoreId(s_SemaphoreId);\r\n\r\nint main(void)\r\n{\r\n \/* Enable the CPU Cache *\/\r\n CPU_CACHE_Enable();\r\n\r\n HAL_Init();\r\n SystemClock_Config();\r\n\r\n osThreadDef(LED1, SenderThread, osPriorityNormal, 0, configMINIMAL_STACK_SIZE);\r\n osThreadDef(LED2, ReceiverThread, osPriorityHigh, 0, configMINIMAL_STACK_SIZE);\r\n\r\n LEDThread1Handle = osThreadCreate(osThread(LED1), NULL);\r\n LEDThread2Handle = osThreadCreate(osThread(LED2), NULL);\r\n\r\n s_SemaphoreId = osSemaphoreCreate(osSemaphore(s_Semaphore), 32);\r\n osKernelStart();\r\n}\r\n\r\nvoid __attribute__((noinline)) TestSinf()\r\n{\r\n volatile float in = 0.1234F, out;\r\n for (int i = 0; i < 100; i++)\r\n out = sinf(in);\r\n}\r\n\r\nvoid __attribute__((noinline)) EmptyFunction()\r\n{\r\n asm(\"nop\");\r\n}\r\n\r\nstatic void SenderThread(void const *argument)\r\n{\r\n uint32_t count = 0;\r\n\r\n for (;;)\r\n {\r\n TestSinf();\r\n \r\n osSemaphoreRelease(s_SemaphoreId);\r\n osDelay(100);\r\n }\r\n}\r\n\r\nstatic void ReceiverThread(void const *argument)\r\n{\r\n for (;;)\r\n {\r\n osSemaphoreWait(s_SemaphoreId, osWaitForever);\r\n EmptyFunction();\r\n }\r\n}<\/pre>\nThe code above creates 2 threads:<\/p>\n
- \n
- The sender thread will call the\u00a0TestSinf() function that calls sinf() 100 times and returns.\u00a0Then it will release a semaphore and\u00a0sleep for 100 milliseconds.\u00a0We will measure the time taken by the TestSinf() function\u00a0on different devices to do a basic comparison of their floating-point performance.<\/li>\n
- The receiver thread will wait on the semaphore released by the sender thread and then will immediately call EmptyFunction(). We will measure the time before\u00a0the call to osSemaphoreWait() and the subsequent invocation of EmptyFunction() to\u00a0estimate the FreeRTOS thread switching latency.<\/li>\n<\/ul>\n<\/li>\n
- Build the code, set a breakpoint inside SenderThread() and start\u00a0debugging by pressing F5. Once the breakpoint hits, open the Debug->Windows->Real-time Watch window:
\n![\"09-led\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/09-led.png\")
<\/a><\/li>\n- Add both threads (LED1 and LED2) and the TestSinf(), EmptyFunction() and osSemaphoreRelease() functions to real-time watch, then press F5 to continue debugging. VisualGDB will display short bursts of activity every 100\u00a0milliseconds corresponding\u00a0to the thread activity:
![\"10-run\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/10-run.png\")
<\/a><\/li>\n- Zoom\u00a0into one of the bursts.\u00a0Hover the mouse over the TestSinf() call to see its run time:
![\"11-sinf\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/11-sinf.png\")
<\/a><\/li>\n- Similarly check the runtime of EmptyFunction():
![\"12-func\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/12-func.png\")
<\/a><\/li>\n- Select the time between the call to osSemaphoreRelease and the invocation of EmptyFunction() in the second thread. This\u00a0represents the time required for the\u00a0threads to switch (that includes changing the thread state, selecting the next thread to run, etc):
![\"13-latency\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/13-latency.png\")
<\/a><\/li>\n- Finally add\u00a0a new C++ source file to the project\u00a0with the following contents:\n
#include <list>\r\n#include <stdlib.h>\r\n\r\nextern \"C\" void TestList()\r\n{\r\n std::list<int> lst;\r\n for (int i = 0; i < 100; i++)\r\n lst.push_back(rand());\r\n\r\n lst.sort();\r\n}<\/pre>\nThen call TestList() from main().<\/li>\n
- Set a breakpoint in TestList() and once it hits, add the list::sort() method to real-time watch:
![\"14-sort\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/14-sort.png\")
<\/a><\/li>\n- Once the sort method\u00a0finishes running, check its runtime in Real-time watch:
![\"14-sorttime\"](\"https:\/\/visualgdb.com\/w\/wp-content\/uploads\/2018\/05\/14-sorttime.png\")
<\/a><\/li>\n- You can repeat those measurements on other boards to compare their relative performance. Below are the figures we measured:
\n - Add both threads (LED1 and LED2) and the TestSinf(), EmptyFunction() and osSemaphoreRelease() functions to real-time watch, then press F5 to continue debugging. VisualGDB will display short bursts of activity every 100\u00a0milliseconds corresponding\u00a0to the thread activity:
- Select “New Project -> Embedded Binary” and click “Next”: