stm32精确延时问题

有时候程序需要让CPU死等一会，一般用定时器或者systick定时器可以实现精确的微秒级延时，但是那样白白占用了定时器，太浪费资源了，且在有os的系统，systick多用于系统心跳。

最近网上看到国外高人使用了这样的延时函数：

rt_err_t rt_hw_delay_init(void) 
{
#if !defined(STM32F0xx)
        uint32_t c;
        
    /* Enable TRC */
    CoreDebug->DEMCR &= ~0x01000000;
    CoreDebug->DEMCR |=  0x01000000;
        
    /* Enable counter */
    DWT->CTRL &= ~0x00000001;
    DWT->CTRL |=  0x00000001;
        
    /* Reset counter */
    DWT->CYCCNT = 0;
        
        /* Check if DWT has started */
        c = DWT->CYCCNT;
        
        /* 2 dummys */
        __ASM volatile ("NOP");
        __ASM volatile ("NOP");
        
        /* if result is zero, DWT has not started */
        if (DWT->CYCCNT - c == 0)
        return RT_ERROR;
#else
        /* Return OK */
        return RT_EOK;
#endif
}
void rt_delay_us(__IO uint32_t micros) 
{
#if !defined(STM32F0xx)
        uint32_t start = DWT->CYCCNT;
        
        /* Go to number of cycles for system */
        micros *= (HAL_RCC_GetHCLKFreq() / 1000000);
        
        /* Delay till end */
        while ((DWT->CYCCNT - start) < micros);
#else
        /* Go to clock cycles */
        micros *= (SystemCoreClock / 1000000) / 5;
        
        /* Wait till done */
        while (micros--);
#endif
}
void rt_delay_ms(__IO uint32_t millis) 
{
    rt_delay_us(millis*1000);
}