STM32G030 Interrupt latency

[ferrario]

I am experiencing a vey long external interrupt latency up to 2 microseconds.

Clock is internal PLL, 64MHz.

void HAL_GPIO_EXTI_Rising_Callback(uint16_t GPIO_Pin)

{

   if (GPIO_Pin == TRIGGER_Pin)

      {

           __disable_irq();

            LL_GPIO_ResetOutputPin(S2_GPIO_Port, S2_Pin);         

            __enable_irq();

      }

}

how can I speed it up, I need it to be <1 microsecond

Thanks

[Javier1]

For reseting your pin you use LL_GPIO_ResetOutputPin(S2_GPIO_Port, S2_Pin);       

You could do it faster by writting the corresponding bit in the BSRR register https://electronics.stackexchange.com/questions/343538/how-does-the-bsrr-register-work

I used BSRR for softwarepwm here

Also you could set the GPIO SPEED to very high

I dont know how much time this will save you....

[ferrario]

Dear Javier,

this is indeed what the LL call does.

__STATIC_INLINE void LL_GPIO_ResetOutputPin(GPIO_TypeDef *GPIOx, uint32_t PinMask)

{

 WRITE_REG(GPIOx->BRR, PinMask);

}

I think the problem lies somewhere else, but I can't find it

[Javier1]

I dont really know if 2us delay is a lot to serve an interruption.

you could place the vector table in sram instead of Flash?

[ferrario]

Vector Table in RAM Crashes

 .isr_vector :

 {

   . = ALIGN(4);

   KEEP(*(.isr_vector)) /* Startup code */

   . = ALIGN(4);

 } >RAM

Thanks

[Tesla DeLorean]

>>how can I speed it up, I need it to be <1 microsecond

Run with less clutter. Implement code IN the IRQHandler

Not clear why interrupts need disabling. Lower priority ones are implicitly blocked, and the write to the BRR is singular and atomic. Callbacks are done in interrupt context

[ferrario]

Thanks a lot. Disabling interrupts has been removed with minor improvement.

WRITE_REG is a macro, actually writes the BRR

Implementing in the ISR let me save only 200ns.

Confused

[TDK]

HAL_GPIO_EXTI_Rising_Callback is not the IRQ handler. In this case, EXTI4_15_IRQHandler is the handler, so write your code in there instead. Don't use HAL if you can't live with a small delay.

<2us is more than doable at 64 MHz clock speed.

[ferrario]

Upon interrupt the void HAL_GPIO_EXTI_IRQHandler(uint16_t GPIO_Pin) is called.

This in turns calls the callbacks.

How can I get rid of HAL in this case, this is the direct interrupt callback?

[TDK]

Like I said, the handler in the vector table is EXTI4_15_IRQHandler. You can view the call stack during the interrupt to verify this. Find that function in your program and put the code there instead.

[ferrario]

Thanks to all

void EXTI0_1_IRQHandler(void)

{

 /* USER CODE BEGIN EXTI0_1_IRQn 0 */

      if (__HAL_GPIO_EXTI_GET_RISING_IT(GPIO_PIN_0 ) != 0x00u)

       {

       // LL_GPIO_ResetOutputPin(S2_GPIO_Port, S2_Pin);

         S2_GPIO_Port->BRR= S2_Pin;

        // delay_us(Delays[WAIT]);

         // LL_GPIO_SetOutputPin(S1_GPIO_Port, S1_Pin); //HOLD

         S1_GPIO_Port->BSRR=S1_Pin;

         __HAL_GPIO_EXTI_CLEAR_RISING_IT(GPIO_PIN_0);

       }

       if (__HAL_GPIO_EXTI_GET_FALLING_IT(GPIO_PIN_0) != 0x00u)

       {

         __HAL_GPIO_EXTI_CLEAR_FALLING_IT(GPIO_PIN_0);

         // HAL_GPIO_EXTI_Falling_Callback(GPIO_Pin);

         // LL_GPIO_ResetOutputPin(S1_GPIO_Port, S1_Pin); //HOLD

         S1_GPIO_Port->BRR=S1_Pin;

         // LL_GPIO_ResetOutputPin(S2_GPIO_Port, S2_Pin);

         S2_GPIO_Port->BRR=S2_Pin;

           HAL_ADC_Start_IT(&hadc1);

       }

 /* USER CODE END EXTI0_1_IRQn 0 */

 // HAL_GPIO_EXTI_IRQHandler(GPIO_PIN_0);

 /* USER CODE BEGIN EXTI0_1_IRQn 1 */

 /* USER CODE END EXTI0_1_IRQn 1 */

}

So

1- code written in the ISR

2- removed even LL calls, wrote the BRR

latency is now 760ns.

Thanks once again