Using the example code from this site (and all other examples appear to be similar):

Cycle Counting on ARM Cortex-M with DWT

The DWT counter on the STM32F411 appears to increment correctly when running a debug session but not in a release session. Here’s an example function that uses DWT to execute a timeout on the I2C bus:

tKernelReturnValue 
OSp_I2C_Master_Send_Data(
    I2C_TypeDef * I2C, 
    char data) 
{
    uint32_t timeout = KIN1_DWT_CYCCNT;
    tKernelReturnValue result;
    
    I2C->DR = data;
    result.error = O$_I2C_TRANSMIT_ERROR;
    while(  (!(I2C->SR1 & I2C_SR1_TXE)) &&
            ((KIN1_DWT_CYCCNT - timeout) < I2C_BUS_TIMEOUT) ) {
        OS_ClrWDT();
    } // end while
    if (I2C->SR1 & I2C_SR1_TXE) {
        result.error = O$_SUCCESS;
    } // end if
    return result;

} // end OSp_I2C_Master_Send_Data

In a live debug session I can verify the DWT is running, but the code won’t break out of the while loop during a release build session. Normally I would have assumed that a clock was incorrectly configured but the F411 reference manual doesn’t indicate that the DWT connects to any clock source that has to be independently enabled. I have used this same example code on the LPC55S69 and not had this issue. I have also set the compiler to not optimize anything just in case. Not sure what else to look at for this issue.